+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+Predicting antibody binders and generating synthetic antibodies using deep learning
+Yoong Wearn Lim
+2022/02/18
+ +
+
+Goals
+-
+
- To predict binder and non-binder antibodies to CTLA-4 and PD-1, using Convolutional Neural Network (CNN). +
- To generate synthetic antibody sequences (CDR3K + CDR3H) to CTLA-4 and PD-1, using Generative Adversarial Network (GAN). +
+
+Settings
+Ensure that python3 and tensorflow version 2.4.0 are used.
+knitr::opts_chunk$set(warning=FALSE, message=FALSE)
+
+library(reticulate)
+use_virtualenv("py3-virtualenv")
+py_config()## python: /home/ywlim/.virtualenvs/py3-virtualenv/bin/python
+## libpython: /usr/lib/python3.6/config-3.6m-x86_64-linux-gnu/libpython3.6.so
+## pythonhome: /home/ywlim/.virtualenvs/py3-virtualenv:/home/ywlim/.virtualenvs/py3-virtualenv
+## virtualenv: /home/ywlim/.virtualenvs/py3-virtualenv/bin/activate_this.py
+## version: 3.6.9 (default, Jul 17 2020, 12:50:27) [GCC 8.4.0]
+## numpy: /home/ywlim/.virtualenvs/py3-virtualenv/lib/python3.6/site-packages/numpy
+## numpy_version: 1.19.2
+##
+## python versions found:
+## /home/ywlim/.virtualenvs/py3-virtualenv/bin/python
+## /home/ywlim/.virtualenvs/r-reticulate/bin/python
+## /home/ywlim/miniconda3/envs/r-reticulate/bin/python
+## /usr/bin/python3
+## /usr/bin/python
+## /home/ywlim/miniconda3/envs/ame/bin/python
+## /home/ywlim/miniconda3/envs/ice_env/bin/python
+## /home/ywlim/miniconda3/bin/pythonlibrary(keras)
+library(tidyverse)
+library(PepTools)
+library(mltools)
+library(caret)
+library(pROC)
+
+filter = dplyr::filter
+select = dplyr::select
+theme_set(theme_bw())
+
+Predicting binders vs. non-binders (CNN)
+
+
+Prepare sequences
+Here we prepare CDR3K + CDR3H amino acid sequences for both targets:
+-
+
- Load all pre- and post-FACS sorted sequences. +
- Randomize the order of the sequences. +
- Pad CDR3K and CDR3H with “-” so all sequences have the same length. +
- Define binders and non-binders. +
-
+
- Binders: post/pre fold change >= 1.8, post-sort frequency >= 0.01 +
- Non-binders: post/pre fold change < 1, pre-sort frequency >= 0.01 +
-
+
- Split sequences into training and testing sets. +
all <- read_delim("CNN/all_ab_pre_post.txt", delim = "\t")
+
+# shuffle order of data frame
+set.seed(42)
+rows <- sample(nrow(all))
+all <- all[rows, ]
+
+# padding
+max_k <- max(nchar(all$CDR3K)) # 11
+max_h <- max(nchar(all$CDR3H)) # 25
+
+all$paddedh <- gsub("\\s", "-", format(all$CDR3H, width = max_h))
+all$paddedk <- gsub("\\s", "-", format(all$CDR3K, width = max_k))
+all$lengthh <- nchar(all$CDR3H)
+all$lengthk <- nchar(all$CDR3K)
+all$padded <- paste0(all$paddedk, all$paddedh)
+
+# filter for only unique sequences
+all2 <- all %>%
+ group_by(padded) %>%
+ top_n(1, wt = post)
+all2 <- all2[!duplicated(all2$padded), ]
+
+# define binders/non-binders
+all2 <- all2 %>%
+ ungroup() %>%
+ mutate(enriched = ifelse((fc >= 1.8 & post >= 0.01), 1,
+ ifelse(fc < 1 & pre >= 0.01, 0, "ambi")))
+
+all2 <- filter(all2, enriched != "ambi")
+
+# split test train
+split_train_test <- function(antigenx) {
+ mini <- all2 %>% filter(antigen == antigenx)
+ set.seed(22)
+ trainIndex <- createDataPartition(mini$enriched, p = .8,
+ list = FALSE,
+ times = 1)
+ df_train <- mini[ trainIndex, ]
+ df_test <- mini[-trainIndex, ]
+
+ my_list <- list(train = df_train, test = df_test)
+ return(my_list)
+}
+c1 <- split_train_test("CTLA-4")
+p1 <- split_train_test("PD-1")
+
+saveRDS(c1, file = "CNN/c1.RDS")
+saveRDS(p1, file = "CNN/p1.RDS")
+
+Visualize test train split
+c1$train$type <- "Train"
+c1$test$type <- "Test"
+c1_both <- bind_rows(c1$train, c1$test)
+
+p1$train$type <- "Train"
+p1$test$type <- "Test"
+p1_both <- bind_rows(p1$train, p1$test)
+
+both <- bind_rows(c1_both, p1_both)
+count <- data.frame(table(both$antigen, both$type, both$enriched))
+names(count) <- c("antigen", "type", "binder", "freq")
+count$type <- factor(count$type, levels = c("Train", "Test"))
+
+count$binder <- as.numeric(as.character(count$binder))
+count$binder[count$binder == 0] <- "Non-binder"
+count$binder[count$binder == 1] <- "Binder"
+
+ggplot(count, aes(x = type, y = freq, fill = binder, label = freq)) +
+ facet_wrap(~ antigen, scales = "free_y") +
+ geom_col(position = "stack") +
+ geom_text(position = position_stack(vjust = .5)) +
+ theme_bw() +
+ labs(y = "# sequences")
+
+Encoding
+Encode sequences into numerical matrices (images) using BLOSUM62 scores.
+# function to encode peptides
+load(file = "BLOSUM62_with_deletion.Rdata")
+
+pep_encode_blosum <- function(pep) {
+ bl62_prob <- BLOSUM62
+ p_mat <- pep %>% pep_mat
+ n_peps <- length(pep)
+ l_peps <- nchar(pep[1])
+ l_enc <- ncol(bl62_prob)
+ o_tensor <- array(data = NA, dim = c(n_peps, l_peps, l_enc))
+ for (i in 1:n_peps) {
+ pep_i_residues <- p_mat[i, ]
+ pep_img <- bl62_prob[pep_i_residues, ]
+ o_tensor[i, , ] <- pep_img
+ }
+ return(o_tensor)
+}# CTLA-4
+c1_train <- pep_encode_blosum(c1$train$padded)
+c1_test <- pep_encode_blosum(c1$test$padded)
+
+# reshape
+c1_train <- array_reshape(c1_train, c(nrow(c1_train), ncol(c1_train), 22, 1))
+c1_test <- array_reshape(c1_test, c(nrow(c1_test), ncol(c1_test), 22, 1))
+
+num_classes <- 2
+
+c1_train_y <- c1$train %>% pull(enriched) %>% array %>% keras::to_categorical(num_classes = num_classes)
+c1_test_y <- c1$test %>% pull(enriched) %>% array %>% keras::to_categorical(num_classes = num_classes)
+
+# PD-1
+p1_train <- pep_encode_blosum(p1$train$padded)
+p1_test <- pep_encode_blosum(p1$test$padded)
+
+# reshape
+p1_train <- array_reshape(p1_train, c(nrow(p1_train), ncol(p1_train), 22, 1))
+p1_test <- array_reshape(p1_test, c(nrow(p1_test), ncol(p1_test), 22, 1))
+
+num_classes <- 2
+
+p1_train_y <- p1$train %>% pull(enriched) %>% array %>% keras::to_categorical(num_classes = num_classes)
+p1_test_y <- p1$test %>% pull(enriched) %>% array %>% keras::to_categorical(num_classes = num_classes)
+
+saveRDS(p1_train, file = "CNN/p1_train.RDS")
+saveRDS(p1_test, file = "CNN/p1_test.RDS")
+saveRDS(p1_train_y, file = "CNN/p1_train_y.RDS")
+saveRDS(p1_test_y, file = "CNN/p1_test_y.RDS")
+saveRDS(c1_train, file = "CNN/c1_train.RDS")
+saveRDS(c1_test, file = "CNN/c1_test.RDS")
+saveRDS(c1_train_y, file = "CNN/c1_train_y.RDS")
+saveRDS(c1_test_y, file = "CNN/c1_test_y.RDS")
+
+CNN models
+Here we build CNN models to predict binders vs. non-binders. We train the CTLA-4 model and PD-1 model separately.
+model <- keras_model_sequential() %>%
+ layer_conv_2d(filters = 64, kernel_size = c(3, 3),
+ activation = "relu", input_shape = c(36, 22, 1), padding = "same") %>%
+ layer_dropout(rate = 0.5) %>%
+ layer_max_pooling_2d(pool_size = c(2, 2), padding = "same") %>%
+ layer_conv_2d(filters = 32, kernel_size = c(3, 4),
+ activation = "relu", padding = "same") %>%
+ layer_dropout(rate = 0.4) %>%
+ layer_max_pooling_2d(pool_size = c(2, 2), padding = "same") %>%
+ layer_conv_2d(filters = 32, kernel_size = c(4, 4),
+ activation = "relu", padding = "same") %>%
+ layer_dropout(rate = 0.3) %>%
+ layer_flatten() %>%
+ layer_dense(units = 64, activation = "relu") %>%
+ layer_dropout(rate = 0.4) %>%
+ layer_dense(units = 2, activation = "softmax")
+summary(model)## Model: "sequential"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## conv2d_2 (Conv2D) (None, 36, 22, 64) 640
+## ________________________________________________________________________________
+## dropout_3 (Dropout) (None, 36, 22, 64) 0
+## ________________________________________________________________________________
+## max_pooling2d_1 (MaxPooling2D) (None, 18, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_1 (Conv2D) (None, 18, 11, 32) 24608
+## ________________________________________________________________________________
+## dropout_2 (Dropout) (None, 18, 11, 32) 0
+## ________________________________________________________________________________
+## max_pooling2d (MaxPooling2D) (None, 9, 6, 32) 0
+## ________________________________________________________________________________
+## conv2d (Conv2D) (None, 9, 6, 32) 16416
+## ________________________________________________________________________________
+## dropout_1 (Dropout) (None, 9, 6, 32) 0
+## ________________________________________________________________________________
+## flatten (Flatten) (None, 1728) 0
+## ________________________________________________________________________________
+## dense_1 (Dense) (None, 64) 110656
+## ________________________________________________________________________________
+## dropout (Dropout) (None, 64) 0
+## ________________________________________________________________________________
+## dense (Dense) (None, 2) 130
+## ================================================================================
+## Total params: 152,450
+## Trainable params: 152,450
+## Non-trainable params: 0
+## ________________________________________________________________________________# duplicate model
+# note we have to use "clone_model", can't just do model_c1 <- model (this is just a pointer)
+model_c1 <- clone_model(model)
+model_p1 <- clone_model(model)
+
+# need to compile each one separately
+model_c1 %>% compile(
+ loss = 'categorical_crossentropy',
+ optimizer = 'adam',
+ metrics = 'accuracy'
+)
+
+model_p1 %>% compile(
+ loss = 'categorical_crossentropy',
+ optimizer = 'adam',
+ metrics = 'accuracy'
+)
+
+epochs <- 30
+batch_size <- 50
+validation_split <- 0.2
+
+# fit models
+c1_hist <- model_c1 %>% fit(
+ c1_train, c1_train_y,
+ epochs = epochs,
+ batch_size = batch_size,
+ validation_split = validation_split)
+
+p1_hist <- model_p1 %>% fit(
+ p1_train, p1_train_y,
+ epochs = epochs,
+ batch_size = batch_size,
+ validation_split = validation_split)
+
+model_c1 %>% save_model_tf("CNN/model_c1")
+model_p1 %>% save_model_tf("CNN/model_p1")
+
+Prediction
+Now that the models are fully trained, we evaluate them using the reserved 20% test sets.
+
+
+CTLA-4 model
+pred_c1 <- model_c1 %>% predict_proba(c1_test) %>% as.data.frame()
+names(pred_c1) <- c("nb_prob", "b_prob")
+y_real_c1 <- c1_test_y %>% apply(1,function(x){ return( which(x==1) - 1) })
+y_pred_c1 <- model_c1 %>% keras::predict_classes(c1_test)
+
+perf_c1 <- model_c1 %>% evaluate(c1_test, c1_test_y)
+acc_c1 <- perf_c1[[2]] %>% round(3) * 100
+mcc_c1 <- mltools::mcc(preds = as.vector(y_pred_c1), actuals = as.vector(y_real_c1)) %>% round(2)
+
+results_c1 <- tibble(y_real = y_real_c1,
+ y_pred = y_pred_c1,
+ b_prob = pred_c1$b_prob)
+
+results_c1 <- results_c1 %>%
+ mutate(class = ifelse(y_real == 0 & y_pred == 0, "TN",
+ ifelse(y_real == 0 & y_pred == 1, "FP",
+ ifelse(y_real == 1 & y_pred == 1, "TP",
+ ifelse(y_real == 1 & y_pred == 0, "FN", NA)))))
+
+res_c1 <- table(results_c1[, 1:2]) %>%
+ as.data.frame() %>%
+ mutate(y_real = as.character(y_real),
+ y_pred = as.character(y_pred))
+
+res_c1$y_real[res_c1$y_real == 0] <- "Non-binder"
+res_c1$y_pred[res_c1$y_pred == 0] <- "Non-binder"
+res_c1$y_real[res_c1$y_real == 1] <- "Binder"
+res_c1$y_pred[res_c1$y_pred == 1] <- "Binder"
+res_c1$class <- c("TN", "FN", "FP", "TP")
+
+# generate confusion matrix
+ggplot(res_c1, aes(x = y_pred, y = y_real, label = Freq, color = class)) +
+ geom_text(size = 7) +
+ geom_tile(color = "black", fill = NA) +
+ theme(axis.text.x = element_text(angle = 45, hjust = 1, vjust = 1, color = "black"),
+ axis.text.y = element_text(color = "black"),
+ axis.ticks = element_blank(),
+ panel.grid.major = element_blank(),
+ panel.grid.minor = element_blank()) +
+ labs(y = "Measured (FACS)", x = "Predicted", title = "CTLA-4",
+ subtitle = paste0("Accuracy=", acc_c1, "\nMCC=", mcc_c1)) +
+ scale_x_discrete(expand = c(0, 0)) +
+ scale_y_discrete(expand = c(0, 0)) +
+ guides(color = FALSE)
# density plot showing binder probabilities
+ggplot(results_c1, aes(x = b_prob, color = class)) +
+ geom_density(fill = NA) +
+ theme(axis.text = element_text(color = "black"),
+ panel.grid.major = element_blank(),
+ panel.grid.minor = element_blank()) +
+ labs(y = "Density", x = "Binder probability", title = "CTLA-4", color = "Type")
# roc curve
+roc_obj_c1 <- roc(y_real_c1, results_c1$b_prob)
+plot.roc(roc_obj_c1, print.auc = TRUE, main = "CTLA-4 ROC curve")
+
+PD-1 model
+pred_p1 <- model_p1 %>% predict_proba(p1_test) %>% as.data.frame()
+names(pred_p1) <- c("nb_prob", "b_prob")
+y_real_p1 <- p1_test_y %>% apply(1,function(x){ return( which(x==1) - 1) })
+y_pred_p1 <- model_p1 %>% keras::predict_classes(p1_test)
+
+perf_p1 <- model_p1 %>% evaluate(p1_test, p1_test_y)
+acc_p1 <- perf_p1[[2]] %>% round(3) * 100
+mcc_p1 <- mltools::mcc(preds = as.vector(y_pred_p1), actuals = as.vector(y_real_p1)) %>% round(2)
+
+results_p1 <- tibble(y_real = y_real_p1,
+ y_pred = y_pred_p1,
+ b_prob = pred_p1$b_prob)
+
+results_p1 <- results_p1 %>%
+ mutate(class = ifelse(y_real == 0 & y_pred == 0, "TN",
+ ifelse(y_real == 0 & y_pred == 1, "FP",
+ ifelse(y_real == 1 & y_pred == 1, "TP",
+ ifelse(y_real == 1 & y_pred == 0, "FN", NA)))))
+
+res_p1 <- table(results_p1[, 1:2]) %>%
+ as.data.frame() %>%
+ mutate(y_real = as.character(y_real),
+ y_pred = as.character(y_pred))
+
+res_p1$y_real[res_p1$y_real == 0] <- "Non-binder"
+res_p1$y_pred[res_p1$y_pred == 0] <- "Non-binder"
+res_p1$y_real[res_p1$y_real == 1] <- "Binder"
+res_p1$y_pred[res_p1$y_pred == 1] <- "Binder"
+res_p1$class <- c("TN", "FN", "FP", "TP")
+
+# generate confusion matrix
+ggplot(res_p1, aes(x = y_pred, y = y_real, label = Freq, color = class)) +
+ geom_text(size = 7) +
+ geom_tile(color = "black", fill = NA) +
+ theme(axis.text.x = element_text(angle = 45, hjust = 1, vjust = 1, color = "black"),
+ axis.text.y = element_text(color = "black"),
+ axis.ticks = element_blank(),
+ panel.grid.major = element_blank(),
+ panel.grid.minor = element_blank()) +
+ labs(y = "Measured (FACS)", x = "Predicted", title = "PD-1",
+ subtitle = paste0("Accuracy=", acc_p1, "\nMCC=", mcc_p1)) +
+ scale_x_discrete(expand = c(0, 0)) +
+ scale_y_discrete(expand = c(0, 0)) +
+ guides(color = FALSE)
# density plot showing binder probabilities
+ggplot(results_p1, aes(x = b_prob, color = class)) +
+ geom_density(fill = NA) +
+ theme(axis.text = element_text(color = "black"),
+ panel.grid.major = element_blank(),
+ panel.grid.minor = element_blank()) +
+ labs(y = "Density", x = "Binder probability", title = "PD-1", color = "Type")
# roc curve
+roc_obj_p1 <- roc(y_real_p1, results_p1$b_prob)
+plot.roc(roc_obj_p1, print.auc = TRUE, main = "PD-1 ROC curve")
+
+
+
+
+Generating synthetic sequences (GAN)
+Here we build GAN models to generate anti-CTLA-4 and anti-PD-1 CDR3 sequences. We will train a separate model for each V gene, and we will train light and heavy chain separately. In total, we will train 15 models.
+
+
+Prepare sequences
+Previously, we annotated full length light and heavy chain sequences using ANARCI.
+We used the AHo numbering scheme, which is based on the structural alignments of the 3D structures of the antibody variable regions. A quote from https://plueckthun.bioc.uzh.ch/antibody/Numbering/NumFrame.html:
+++An important difference to the numbering scheme presented here (AHo) is that in the IMGT scheme insertions and deletions “grow” unidirectionally, as in the original Chothia definition (Chothia & Lesk, 1987), while in the AHo scheme, insertions and deletions are placed symmetrically around the key position marked in yellow. Furthermore, length variations in CDR 1 and CDR 2 are represented by a single gap in IMGT and by one or two gaps in AHo.
+
After AHo annotation, the CDR3 regions were extracted from the sequences. All sequences now have the same sequence length.
+
+
+Encoding
+Here we encode the pre-processed CDR3 sequences into “images”.
+# pre-processed CDR3 sequences, by target, chain, and V gene
+seq_all <- readRDS(file = "GAN/seq_all.RDS")
+
+encoded <- list()
+length(encoded) <- 15
+names(encoded) <- names(seq_all)
+
+for (i in 1:15) {
+ seq_encoded <- pep_encode_blosum(seq_all[[i]])
+ # reshape
+ encoded[[i]] <- array_reshape(seq_encoded, c(nrow(seq_encoded), ncol(seq_encoded), 22, 1))
+}
+saveRDS(encoded, file = "GAN/seq_all_encoded.RDS")
+
+head(str(encoded))## List of 15
+## $ 1 CTLA4 heavy IGHV3-33*01 : num [1:307, 1:32, 1:22, 1] 0.267 0.533 0.533 0.533 0.533 ...
+## $ 2 CTLA4 heavy IGHV1-18*04 : num [1:70, 1:32, 1:22, 1] 0.533 0.2 0.533 0.533 0.267 ...
+## $ 3 CTLA4 heavy IGHV3-20*01 : num [1:39, 1:32, 1:22, 1] 0.267 0.533 0.533 0.533 0.2 ...
+## $ 4 CTLA4 heavy IGHV4-39*01 : num [1:36, 1:32, 1:22, 1] 0.533 0.533 0.533 0.533 0.533 ...
+## $ 5 CTLA4 light IGKV3-20*01 : num [1:292, 1:32, 1:22, 1] 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 ...
+## $ 6 CTLA4 light IGKV1D-39*01: num [1:54, 1:32, 1:22, 1] 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 ...
+## $ 7 CTLA4 light IGKV1-17*01 : num [1:37, 1:32, 1:22, 1] 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 ...
+## $ 8 CTLA4 light IGKV1-16*01 : num [1:34, 1:32, 1:22, 1] 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 ...
+## $ 9 PD1 heavy IGHV4-4*07 : num [1:81, 1:32, 1:22, 1] 0.533 0.533 0.533 0.533 0.533 ...
+## $ 10 PD1 heavy IGHV3-33*03 : num [1:80, 1:32, 1:22, 1] 0.267 0.533 0.533 0.533 0.267 ...
+## $ 11 PD1 heavy IGHV1-18*04 : num [1:67, 1:32, 1:22, 1] 0.533 0.533 0.533 0.533 0.533 ...
+## $ 12 PD1 light IGKV1-17*01 : num [1:104, 1:32, 1:22, 1] 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 ...
+## $ 13 PD1 light IGKV1-6*02 : num [1:43, 1:32, 1:22, 1] 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 ...
+## $ 14 PD1 light IGKV3-15*01 : num [1:38, 1:32, 1:22, 1] 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 ...
+## $ 15 PD1 light IGKV1-9*01 : num [1:32, 1:32, 1:22, 1] 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 ...## NULL# encoded is a list with 15 items
+str(encoded[[1]])## num [1:307, 1:32, 1:22, 1] 0.267 0.533 0.533 0.533 0.533 ...# num [1:307, 1:32, 1:22, 1] 0.267 0.533 0.533 0.533 0.533 ...
+
+GAN models
+32 positions (row), 22 possible amino acids (including X and -) (columns).
+h <- 32; w <- 22; c <- 1; l <- 100
+loss_list <- list()
+
+# 1 model for each target/chain/V gene
+# 15 models total
+for (modelx in 1:15) {
+ # generator network
+ gi <- layer_input(shape = l)
+
+ go <- gi %>% layer_dense(units = 16 * 11 * 128) %>%
+ layer_activation_leaky_relu() %>%
+ layer_reshape(target_shape = c(16, 11, 128)) %>%
+ layer_conv_2d(filters = 64,
+ kernel_size = c(2, 2),
+ padding = "same") %>%
+ layer_activation_leaky_relu() %>%
+ layer_conv_2d_transpose(filters = 32,
+ kernel_size = c(2, 2),
+ strides = 2,
+ padding = "same") %>%
+ layer_activation_leaky_relu() %>%
+ layer_conv_2d(filters = 128,
+ kernel_size = c(5, 6),
+ padding = "same") %>%
+ layer_activation_leaky_relu() %>%
+ layer_conv_2d(filters = 1,
+ kernel_size = c(6, 6),
+ activation = "tanh",
+ padding = "same")
+ g <- keras_model(gi, go)
+ summary(g)
+
+ # discriminator network
+ di <- layer_input(shape = c(h, w, c))
+
+ do <- di %>%
+ layer_conv_2d(filters = 96, kernel_size = 3) %>%
+ layer_activation_leaky_relu() %>%
+ layer_flatten() %>%
+ layer_dropout(rate = 0.3) %>%
+ layer_dense(units = 1, activation = "sigmoid")
+
+ d <- keras_model(di, do)
+
+ summary(d)
+
+ # compile discriminator
+ d %>% compile(optimizer = "rmsprop",
+ loss = "binary_crossentropy")
+
+ # freeze weights and compile
+ freeze_weights(d)
+
+ gani <- layer_input(shape = l)
+ gano <- gani %>% g %>% d
+ gan <- keras_model(gani, gano)
+ gan %>% compile(optimizer = "rmsprop",
+ loss = "binary_crossentropy")
+ summary(gan)
+
+ start <- 1; dloss <- NULL; gloss <- NULL
+
+ # use batch size = 20, train 100 rounds
+ b <- 20
+ for (i in 1:100) {
+ # random noise
+ noise <- matrix(rnorm(b * l), nrow = b, ncol = l)
+
+ fake <- g %>% predict(noise)
+
+ # combine real and fake images
+ stop <- start + b - 1
+ # 50 random real images
+ rowx <- sample(nrow(encoded[[modelx]]), size = b)
+ real <- encoded[[modelx]][rowx,,, , drop = FALSE] # no need to reshape since we use drop=FALSE
+
+ rows <- nrow(real)
+ both <- array(0, dim = c(rows * 2, dim(real)[-1]))
+ both[1:rows,,,] <- fake
+ both[(rows + 1):(rows * 2),,,] <- real
+ labels <- rbind(matrix(runif(b, 0.9, 1), nrow = b, ncol = 1),
+ matrix(runif(b, 0, 0.1), nrow = b, ncol = 1)) # note the noise in truth labels
+
+ # train discriminator
+ dloss[i] <- d %>% train_on_batch(both, labels)
+
+ # train generator using gan
+ fakeAsReal <- array(runif(b, 0, 0.1), dim = c(b, 1))
+ gloss[i] <- gan %>% train_on_batch(noise, fakeAsReal)
+ }
+
+ # save the generator
+ save_model_tf(g, filepath = paste0("GAN/GAN_model_", modelx))
+
+ # monitor loss
+ res <- data.frame(iteration = seq(1:100), dloss = dloss, gloss = gloss)
+ res <- res %>%
+ gather(type, value, 2:3)
+ res$id <- modelx
+ loss_list[[modelx]] <- res
+}## Model: "model"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_1 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_2 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_3 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_5 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_2 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose (Conv2DTranspose) (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_1 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_4 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_3 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_1"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_2 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_6 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_4 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_1 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_4 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_3 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_2"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_3 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_1 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_3"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_4 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_4 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_8 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_1 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_9 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_7 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_1 (Conv2DTranspose (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_6 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_8 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_5 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_7 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_4"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_5 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_10 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_9 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_2 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_5 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_5 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_5"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_6 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_3 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_4 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_6"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_7 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_6 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_13 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_2 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_13 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_12 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_2 (Conv2DTranspose (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_11 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_12 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_10 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_11 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_7"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_8 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_14 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_14 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_3 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_6 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_7 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_8"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_9 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_6 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_7 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_9"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_10 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_8 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_18 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_3 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_17 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_17 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_3 (Conv2DTranspose (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_16 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_16 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_15 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_15 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_10"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_11 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_18 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_19 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_4 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_7 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_9 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_11"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_12 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_9 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_10 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_12"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_13 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_10 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_23 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_4 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_21 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_22 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_4 (Conv2DTranspose (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_21 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_20 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_20 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_19 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_13"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_14 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_22 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_24 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_5 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_8 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_11 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_14"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_15 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_12 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_13 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_15"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_16 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_12 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_28 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_5 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_25 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_27 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_5 (Conv2DTranspose (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_26 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_24 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_25 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_23 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_16"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_17 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_26 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_29 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_6 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_9 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_13 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_17"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_18 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_15 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_16 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_18"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_19 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_14 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_33 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_6 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_29 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_32 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_6 (Conv2DTranspose (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_31 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_28 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_30 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_27 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_19"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_20 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_30 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_34 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_7 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_10 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_15 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_20"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_21 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_18 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_19 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_21"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_22 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_16 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_38 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_7 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_33 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_37 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_7 (Conv2DTranspose (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_36 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_32 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_35 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_31 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_22"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_23 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_34 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_39 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_8 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_11 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_17 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_23"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_24 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_21 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_22 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_24"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_25 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_18 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_43 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_8 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_37 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_42 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_8 (Conv2DTranspose (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_41 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_36 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_40 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_35 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_25"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_26 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_38 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_44 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_9 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_12 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_19 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_26"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_27 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_24 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_25 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_27"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_28 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_20 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_48 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_9 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_41 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_47 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_9 (Conv2DTranspose (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_46 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_40 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_45 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_39 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_28"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_29 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_42 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_49 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_10 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_13 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_21 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_29"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_30 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_27 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_28 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_30"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_31 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_22 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_53 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_10 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_45 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_52 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_10 (Conv2DTranspos (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_51 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_44 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_50 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_43 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_31"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_32 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_46 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_54 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_11 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_14 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_23 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_32"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_33 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_30 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_31 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_33"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_34 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_24 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_58 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_11 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_49 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_57 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_11 (Conv2DTranspos (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_56 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_48 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_55 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_47 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_34"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_35 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_50 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_59 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_12 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_15 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_25 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_35"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_36 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_33 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_34 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_36"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_37 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_26 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_63 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_12 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_53 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_62 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_12 (Conv2DTranspos (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_61 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_52 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_60 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_51 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_37"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_38 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_54 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_64 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_13 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_16 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_27 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_38"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_39 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_36 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_37 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_39"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_40 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_28 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_68 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_13 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_57 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_67 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_13 (Conv2DTranspos (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_66 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_56 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_65 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_55 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_40"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_41 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_58 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_69 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_14 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_17 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_29 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_41"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_42 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_39 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_40 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+## Model: "model_42"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_43 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## dense_30 (Dense) (None, 22528) 2275328
+## ________________________________________________________________________________
+## leaky_re_lu_73 (LeakyReLU) (None, 22528) 0
+## ________________________________________________________________________________
+## reshape_14 (Reshape) (None, 16, 11, 128) 0
+## ________________________________________________________________________________
+## conv2d_61 (Conv2D) (None, 16, 11, 64) 32832
+## ________________________________________________________________________________
+## leaky_re_lu_72 (LeakyReLU) (None, 16, 11, 64) 0
+## ________________________________________________________________________________
+## conv2d_transpose_14 (Conv2DTranspos (None, 32, 22, 32) 8224
+## ________________________________________________________________________________
+## leaky_re_lu_71 (LeakyReLU) (None, 32, 22, 32) 0
+## ________________________________________________________________________________
+## conv2d_60 (Conv2D) (None, 32, 22, 128) 123008
+## ________________________________________________________________________________
+## leaky_re_lu_70 (LeakyReLU) (None, 32, 22, 128) 0
+## ________________________________________________________________________________
+## conv2d_59 (Conv2D) (None, 32, 22, 1) 4609
+## ================================================================================
+## Total params: 2,444,001
+## Trainable params: 2,444,001
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_43"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_44 (InputLayer) [(None, 32, 22, 1)] 0
+## ________________________________________________________________________________
+## conv2d_62 (Conv2D) (None, 30, 20, 96) 960
+## ________________________________________________________________________________
+## leaky_re_lu_74 (LeakyReLU) (None, 30, 20, 96) 0
+## ________________________________________________________________________________
+## flatten_15 (Flatten) (None, 57600) 0
+## ________________________________________________________________________________
+## dropout_18 (Dropout) (None, 57600) 0
+## ________________________________________________________________________________
+## dense_31 (Dense) (None, 1) 57601
+## ================================================================================
+## Total params: 58,561
+## Trainable params: 58,561
+## Non-trainable params: 0
+## ________________________________________________________________________________
+## Model: "model_44"
+## ________________________________________________________________________________
+## Layer (type) Output Shape Param #
+## ================================================================================
+## input_45 (InputLayer) [(None, 100)] 0
+## ________________________________________________________________________________
+## model_42 (Functional) (None, 32, 22, 1) 2444001
+## ________________________________________________________________________________
+## model_43 (Functional) (None, 1) 58561
+## ================================================================================
+## Total params: 2,502,562
+## Trainable params: 2,444,001
+## Non-trainable params: 58,561
+## ________________________________________________________________________________
+
+Generating sequences
+Use the fully trained GAN models to generate 100 sequences for each model.
+myseq_all <- list()
+for (j in 1:15) {
+ # load previously saved generator
+ g <- load_model_tf(paste0("GAN/GAN_model_", j))
+
+ myseq <- list()
+ for (i in 1:100) {
+ noise <- matrix(rnorm(b * l), nrow = b, ncol = l)
+
+ # generate fake image using the generator
+ fake <- g %>% predict(noise)
+ f <- fake[1,,,]
+ dim(f) <- c(32, 22, 1)
+
+ myseq[[i]] <- f
+ }
+ myseq_all[[j]] <- myseq
+}
+
+Decode images
+Here we decode the generated CDR3 images back into amino acid sequences.
+aa_order <- c("A", "R", "N", "D", "C", "Q", "E", "G", "H", "I", "L", "K", "M", "F", "P", "S", "T", "W", "Y", "V", "X", "-")
+
+decode_seq_cdr3 <- function(seq_list) {
+ aa_list <- list()
+ for (i in 1:length(seq_list)) {
+ peptide1 <- as.data.frame(seq_list[[i]])
+ colnames(peptide1) <- aa_order
+
+ # remove column for amino acid X
+ # it was previously added for padding so that we have an even shape
+ peptide1 <- peptide1 %>% select(-X)
+
+ # which amino acid has the highest score (decoding)
+ aa <- colnames(peptide1)[apply(peptide1, 1, which.max)]
+
+ # find where the gaps are
+ gaps <- which(aa %in% "-")
+ # first and last gap in CDR3 region
+ gap_first <- gaps[gaps > 1][1]
+ gap_last <- gaps[gaps < 32]
+ gap_last <- gap_last[length(gap_last)]
+
+ # replace spurious aa flanked by gaps in CDR3
+ aa[gap_first:gap_last] <- "-"
+
+ aa_string <- paste(aa, collapse = "")
+
+ # remove gaps
+ aa_string <- gsub("-", "", aa_string)
+
+ aa_list[[i]] <- aa_string
+ }
+ all_aa <- data.frame(aa = do.call(rbind, aa_list))
+ return(all_aa)
+}gen_seq <- list()
+
+for (i in 1:15) {
+ temp <- decode_seq_cdr3(myseq_all[[i]])
+ gen_seq[[i]] <- temp$aa
+}
+
+names(gen_seq) <- names(seq_all)Example generated sequences:
+for (i in 1:15) {
+ print(names(gen_seq)[i])
+ print(gen_seq[[i]][!duplicated(gen_seq[[i]])] %>% head(10))
+} ## [1] "1 CTLA4 heavy IGHV3-33*01"
+## [1] "ARGHWFGYFDY" "ARGHIFGYFDY" "ARGAIMGQGYFDY" "ARGEDMGHFDY"
+## [5] "AKGAQHKGYFDY" "ARGEIWGYFDY" "ARGGFWGYFDY" "ARGGGLGGYFDY"
+## [9] "ARKGKLGGPFDY" "ARGGHIGYYDY"
+## [1] "2 CTLA4 heavy IGHV1-18*04"
+## [1] "ARIGPSNFDY" "ARVGRSYWFDY" "ARVGRSKFDY" "ARVGIPWFDY" "ARVSGFNGIDY"
+## [6] "ARVGANFYFDY" "ARIGASWFFDY" "ARIGSGYFDY" "ARVGPHGDFDY" "ARVGRYGQFDY"
+## [1] "3 CTLA4 heavy IGHV3-20*01"
+## [1] "ARDGWKLLDDY" "ARDGWEILDDY" "ARDGWKRNLLDDY" "ARDGWKRNALLDDY"
+## [5] "ARDGWRLLDDY" "ARDGWELLDDY" "ARDGWRLGDDY" "ARDGWEYYYLDDY"
+## [9] "ARDGWDILDDY" "ARDGWKYYYLGDDY"
+## [1] "4 CTLA4 heavy IGHV4-39*01"
+## [1] "AKAAYYYGNYVWYFDY" "ARHNISFGSSYYFDY" "AHAYYNRGSWHYDY"
+## [4] "ARIAHHYANRYGSWHFDY" "ARIAVYYWYFDY" "AQKSYRFSSYYYYYFFDY"
+## [7] "ARKKIQYSSWYFDY" "AREEVEYSWHFDY" "AGKKYYYVSVYSSYYFDY"
+## [10] "ARIYYHHGYYGYYYFEY"
+## [1] "5 CTLA4 light IGKV3-20*01"
+## [1] "QQYGSSPWT" "QQYGSSPFT" "QQYGASPWT" "QQYGNSPWT" "QQYGSSPYT" "QQYGNSPFT"
+## [7] "QQYGTSPFT"
+## [1] "6 CTLA4 light IGKV1D-39*01"
+## [1] "QQSYSTPFT" "QQSYSTPYT" "QQAYSTPYT" "QQNYSTPYT" "QQAYSTPFT" "QQDYSTPYT"
+## [7] "QQNYSTPFT"
+## [1] "7 CTLA4 light IGKV1-17*01"
+## [1] "LQHNSYPWT" "MQHNSYPWT" "LQHNQYPWT" "LQYHSYPWT" "LQYNSYPWT" "LQHHSYPWT"
+## [7] "LQHHSFPWT" "LQRNSYPWT" "LQHGSYPWT" "LQHNNYPWT"
+## [1] "8 CTLA4 light IGKV1-16*01"
+## [1] "QQYKSYPYT" "QQYNSYPYT" "QQYRSYPYT"
+## [1] "9 PD1 heavy IGHV4-4*07"
+## [1] "ARDWYGSGRNRYYGMDV" "ARDYYGAGQDRYYGMDV" "ARDYDGSGRRYYGMDV"
+## [4] "ARDYYGSGRRYYGMDV" "ARDYYGSGRDRYYGMDV" "ARDFFGSGNRYYGMDV"
+## [7] "ARDYDSSGSNSYHYGMDV" "AKDHYGSGRRYYGMDV" "ARDYDGSGQRYYGMDV"
+## [10] "ARDFYGSGRKYYGMDV"
+## [1] "10 PD1 heavy IGHV3-33*03"
+## [1] "AGGPATYGDY" "AGGGGYFGDY" "AGGGNYYADY" "AGGGYWDY" "AGGGKYYGDY"
+## [6] "AGGGHYFGDY" "AGGPSYYFDY" "AGGGGYYWDY" "AGGGNYYGDY" "AGGGNYYYDY"
+## [1] "11 PD1 heavy IGHV1-18*04"
+## [1] "ARGTGYYQGGMDY" "ARDYYDVLTGHRGIFDY" "ARDHHDILYYGHIFDY"
+## [4] "ARWVFYSIWFDY" "EQEHYESSRGVVYYSMDI" "ARDHHEILRGYHHHIFDY"
+## [7] "AREHSRIMDY" "ARQHFSIAQEIIIYDY" "AREHHYSNADYYYYGFDY"
+## [10] "AQGTGDTWYDY"
+## [1] "12 PD1 light IGKV1-17*01"
+## [1] "LQYNSYPYT" "LQHNNYPWT" "LQHNSYPYT" "LQYTSYPYT" "LQHTSYPYT" "LQHGSYPYT"
+## [7] "LQHNSYPFT" "MQHNSYPWT" "LQHESYPYT" "LQYSSYPYT"
+## [1] "13 PD1 light IGKV1-6*02"
+## [1] "LQDYNYPWT" "LQDFNYPWT" "LQDYNYPYT" "LQDYNYPFT" "LQDHNYPWT"
+## [1] "14 PD1 light IGKV3-15*01"
+## [1] "QQYNNWPFT" "QQYNNWPYT" "QQYNNWPWT" "QQYNSWPWT" "QQYNNWPKT" "QQYNTWPYT"
+## [7] "QQYNNWPST" "QQYNNWPIT" "QQYNNWPRT" "QQYNNWPLT"
+## [1] "15 PD1 light IGKV1-9*01"
+## [1] "QQLNSYPYT" "QQVNSYPYT" "QQLNSYPWT" "QQLNSYPFT" "QQLNSYPMT" "QQLNNYPYT"
+## [7] "QQMNSYPYT"