update curation scripts

src/02.2_assemble_K50_dG_dataset.R

@@ -57,19 +57,25 @@ dataset1 |>
 #    G0: Good (wild-type ΔG values below 4.75 kcal mol^−1), 325,132 ΔG measurements at 17,093 sites in 365 domains
 #    G1: Good but WT outside dynamic range
 
-dataset23 <- readr::read_csv(
+dataset2 <- readr::read_csv(
     file = "data/Processed_K50_dG_datasets/Tsuboyama2023_Dataset2_Dataset3_20230416.csv",
     show_col_types = FALSE)
 # 776,298 rows
 
-dataset23 |>
+dataset2 |>
     arrow::write_parquet(
-        "intermediate/dataset23.parquet")
+        "intermediate/dataset2.parquet")
 
-dataset3 <- dataset23 |>
+dataset3 <- dataset2 |>
     dplyr::filter(ddG_ML != "-")
 
-dataset3_single_mutant <- dataset3 |>
+dataset3 |>
+    arrow::write_parquet(
+        "intermediate/dataset3.parquet")
+
+
+
+dataset3_single <- dataset3 |>
     dplyr::filter(!(mut_type |> stringr::str_detect("(ins|del|[:])")))
         
 
@@ -77,38 +83,16 @@ ThermoMPNN_splits |> dplyr::group_by(split_name) |>
     dplyr::do({
         split <- .
         split_name <- split$split_name[1]
-        mutant_set <- dataset3_single_mutant |>
+        mutant_set <- dataset3_single |>
             dplyr::filter(mut_type != "wt") |>
             dplyr::semi_join(split, by = c("WT_name" = "id"))
         cat("Writing out split ", split_name, ", nrow: ", nrow(mutant_set), "\n", sep = "")
         
         arrow::write_parquet(
             x = mutant_set,
-            sink = paste0("intermediate/dataset3_ThermoMPNN_", split_name, ".parquet"))
+            sink = paste0("intermediate/dataset3_single_", split_name, ".parquet"))
         data.frame()
     })
-        
-
-dataset3_single_mutant_train <- dataset3_single_mutant |>
-    dplyr::filter(mut_type != "wt") |>
-    dplyr::semi_join(
-        ThermoMPNN_splits |>
-        dplyr::filter(split_name == "train"),
-        by = c("WT_name" = "id"))
-
-dataset3_single_mutant_val <- dataset3_single_mutant |>
-    dplyr::filter(mut_type != "wt") |>
-    dplyr::semi_join(
-        ThermoMPNN_splits |>
-        dplyr::filter(split_name == "val"),
-        by = c("WT_name" = "id"))
-
-dataset3_single_mutant_test <- dataset3_single_mutant |>
-    dplyr::filter(mut_type != "wt") |>
-    dplyr::semi_join(
-        ThermoMPNN_splits |>
-        dplyr::filter(split_name == "test"),
-        by = c("WT_name" = "id"))
 
 
 ####

src/02.2_check_assembled_datasets.R

@@ -0,0 +1,44 @@
+
+
+# consistency between models and function predictions
+source("product/MPI/src/summarize_map.R")
+
+
+
+check_id_consistency <- function(
+    dataset_tag,
+    split,
+    verbose = FALSE) {
+
+    if (verbose) {
+        cat("Loading model ids...\n")
+    }
+    ids <- arrow::read_parquet(
+        paste0("intermediate/", dataset_tag, "_", split, ".parquet"),
+        col_select = "id")
+
+    if (verbose) {
+        cat("Loading function prediction ids...\n")
+    }    
+    ids_anno <- arrow::read_parquet(
+        paste0("intermediate/", dataset_tag, "_function_predictions.parquet"),
+        col_select = "id") |>
+        dplyr::distinct(id)
+
+    problems <- dplyr::full_join(
+        ids_model |>
+        dplyr::mutate(model_id = id),
+        ids_anno |>
+        dplyr::mutate(anno_id = id),
+        by = "id") |>
+        summarize_map(
+            x_cols = model_id,
+            y_cols = anno_id,
+            verbose = verbose)
+    problems
+}
+
+check_id_consistency("rosetta_high_quality")
+check_id_consistency("rosetta_low_quality")
+check_id_consistency("dmpfold_high_quality")
+check_id_consistency("dmpfold_low_quality")

src/03.1_upload_data.py

@@ -16,12 +16,6 @@
 import datasets
 
 
-# Dataset1
-# Dataset2
-# Dataset3
-#  Single Mutants
-#    
-
 # dataset1
 # dataset2
 # dataset3
@@ -30,6 +24,53 @@ import datasets
 
 
 
+##### dataset1 #######
+dataset = datasets.load_dataset(
+    "parquet",
+    name = "dataset",
+    data_dir = "./intermediate",
+    data_files = {
+        "train" : "dataset1.parquet"},
+    cache_dir = "/scratch/maom_root/maom0/maom",
+    keep_in_memory = True)
+
+dataset.push_to_hub(
+    repo_id = "maom/MegaScale",
+    config_name = "dataset1",
+    data_dir = "dataset1/data")
+
+
+##### dataset2 #######
+dataset = datasets.load_dataset(
+    "parquet",
+    name = "dataset",
+    data_dir = "./intermediate",
+    data_files = {
+        "train" : "dataset2.parquet"},
+    cache_dir = "/scratch/maom_root/maom0/maom",
+    keep_in_memory = True)
+
+dataset.push_to_hub(
+    repo_id = "maom/MegaScale",
+    config_name = "dataset2",
+    data_dir = "dataset2/data")
+
+
+##### dataset3 #######
+dataset = datasets.load_dataset(
+    "parquet",
+    name = "dataset",
+    data_dir = "./intermediate",
+    data_files = {
+        "train" : "dataset3.parquet"},
+    cache_dir = "/scratch/maom_root/maom0/maom",
+    keep_in_memory = True)
+
+dataset.push_to_hub(
+    repo_id = "maom/MegaScale",
+    config_name = "dataset3",
+    data_dir = "dataset3/data")
+
 
 ##### dataset3_single #######
 dataset = datasets.load_dataset(
@@ -37,9 +78,9 @@ dataset = datasets.load_dataset(
     name = "dataset3_single",
     data_dir = "./intermediate",
     data_files = {
-        "train" : "dataset3_ThermoMPNN_train.parquet",
-        "val" : "dataset3_ThermoMPNN_val.parquet",
-        "test" : "dataset3_ThermoMPNN_test.parquet"},
+        "train" : "dataset3_single_train.parquet",
+        "val" : "dataset3_single_val.parquet",
+        "test" : "dataset3_single_test.parquet"},
     cache_dir = "/scratch/maom_root/maom0/maom",
     keep_in_memory = True)
 
@@ -49,31 +90,31 @@ dataset.push_to_hub(
     data_dir = "dataset3_single/data")
 
 
-##### dataset3_single #######
+##### dataset3_single_CV #######
 dataset = datasets.load_dataset(
     "parquet",
     name = "dataset3_single_CV",
     data_dir = "./intermediate",
     data_files = {
-        "train_0" : "dataset3_ThermoMPNN_train_0.parquet",
-        "train_1" : "dataset3_ThermoMPNN_train_1.parquet",        
-        "train_2" : "dataset3_ThermoMPNN_train_2.parquet",
-        "train_3" : "dataset3_ThermoMPNN_train_3.parquet",
-        "train_4" : "dataset3_ThermoMPNN_train_4.parquet",        
-        "val_0" : "dataset3_ThermoMPNN_val_0.parquet",
-        "val_1" : "dataset3_ThermoMPNN_val_1.parquet",        
-        "val_2" : "dataset3_ThermoMPNN_val_2.parquet",
-        "val_3" : "dataset3_ThermoMPNN_val_3.parquet",
-        "val_4" : "dataset3_ThermoMPNN_val_4.parquet",        
-        "test_0" : "dataset3_ThermoMPNN_test_0.parquet",
-        "test_1" : "dataset3_ThermoMPNN_test_1.parquet",        
-        "test_2" : "dataset3_ThermoMPNN_test_2.parquet",
-        "test_3" : "dataset3_ThermoMPNN_test_3.parquet",
-        "test_4" : "dataset3_ThermoMPNN_test_4.parquet"},
+        "train_0" : "dataset3_single_CV_train_0.parquet",
+        "train_1" : "dataset3_single_CV_train_1.parquet",        
+        "train_2" : "dataset3_single_CV_train_2.parquet",
+        "train_3" : "dataset3_single_CV_train_3.parquet",
+        "train_4" : "dataset3_single_CV_train_4.parquet",        
+        "val_0" : "dataset3_single_CV_val_0.parquet",
+        "val_1" : "dataset3_single_CV_val_1.parquet",        
+        "val_2" : "dataset3_single_CV_val_2.parquet",
+        "val_3" : "dataset3_single_CV_val_3.parquet",
+        "val_4" : "dataset3_single_CV_val_4.parquet",        
+        "test_0" : "dataset3_single_CV_test_0.parquet",
+        "test_1" : "dataset3_single_CV_test_1.parquet",        
+        "test_2" : "dataset3_single_CV_test_2.parquet",
+        "test_3" : "dataset3_single_CV_test_3.parquet",
+        "test_4" : "dataset3_single_CV_test_4.parquet"},
     cache_dir = "/scratch/maom_root/maom0/maom",
     keep_in_memory = True)
 
 dataset.push_to_hub(
     repo_id = "MaomLab/MegaScale",
     config_name = "dataset3_single_CV",
-    data_dir = "datase3_single/data")
+    data_dir = "datase3_single_CV/data")

src/03.2_check_uploaded_data.py

@@ -0,0 +1,42 @@
+
+
+import datasets
+import pyarrow
+
+def test_local_hf_match(
+        dataset_tag,
+        split):
+    print(f"For dataset '{dataset_tag}' and split '{split}' testing if local and remote ids match ...")
+    ids_hf = datasets.load_dataset(
+        path = "maom/MegaScale",
+        name = dataset_tag,
+        data_dir = dataset_tag,
+        cache_dir = "/scratch/maom_root/maom0/maom",
+        keep_in_memory = True).data[split].select(['id']).to_pandas()
+    ids_local = pyarrow.parquet.read_table(
+        source = f"intermediate/{dataset_tag}_{split}.parquet",
+        columns = ["id"]).to_pandas()
+    assert ids_local.equals(ids_hf)
+    
+
+test_local_hf_match("dataset3_single", "train")
+test_local_hf_match("dataset3_single", "val")
+test_local_hf_match("dataset3_single", "test")
+
+test_local_hf_match("dataset3_single_CV", "train_0")
+test_local_hf_match("dataset3_single_CV", "train_1")
+test_local_hf_match("dataset3_single_CV", "train_2")
+test_local_hf_match("dataset3_single_CV", "train_3")
+test_local_hf_match("dataset3_single_CV", "train_4")
+
+test_local_hf_match("dataset3_single_CV", "val_0")
+test_local_hf_match("dataset3_single_CV", "val_1")
+test_local_hf_match("dataset3_single_CV", "val_2")
+test_local_hf_match("dataset3_single_CV", "val_3")
+test_local_hf_match("dataset3_single_CV", "val_4")
+
+test_local_hf_match("dataset3_single_CV", "test_0")
+test_local_hf_match("dataset3_single_CV", "test_1")
+test_local_hf_match("dataset3_single_CV", "test_2")
+test_local_hf_match("dataset3_single_CV", "test_3")
+test_local_hf_match("dataset3_single_CV", "test_4")

src/summarize_map.R

@@ -0,0 +1,346 @@
+
+#'  Diagnostics for messy joins
+#'
+#'  given a data frame with two ways to group rows,
+#'  summarize and give examples of situations where the mapping is not 1-1
+#'
+#' @param x_cols tidyselect specification of a set of columns defining objects
+#' @param y_cols tidyselect specification of a set of columns defining objects
+#'
+#' data <- data.frame(
+#' 	x=c(1,2,NA,3,4,5,6,6,6,7,7),
+#' 	y=c("a",NA,"c","d","d","d","e","f","g","h","h"))
+#'
+#' data |> summarize_map(
+#'   x_cols = x),
+#'   y_cols = y))
+#' X<-[x]:
+#'   |X|: 7                          # number of groups
+#'   |is.na.X|: 1                    # number of groups with NA in atleaset 1 col
+#'   range(|x|:X): 1, 3              # size range of groups
+#' Y<-[y]:
+#'   |Y|: 7
+#'   |is.na.Y|: 1
+#'   range(|y|:Y): 1, 3
+#' [X U Y]:                          # grouping by the union of xcols and ycols
+#'   |X U Y|: 8
+#'   |is.na.XUY|: 2
+#'   range(|z|:X U Y): 1, 2
+#' [X @ Y]:
+#'   |X ~ Y|: 5
+#'   |X:X < Y|, |Y:Y < X|: 1, 1
+#'   |X:X > Y|, |Y:Y < X|: 3, 3
+#' $is.na.X
+#'    x y
+#' 1 NA c
+#'
+#' $is.na.Y
+#'   x    y
+#' 1 2 <NA>
+#'
+#' $dup.XUY
+#'   x y
+#' 1 7 h
+#' 2 7 h
+#'
+#' $dup.X
+#'   x y
+#' 1 6 e
+#' 2 6 f
+#' 3 6 g
+#'
+#' $dup.Y
+#'   y x
+#' 1 d 3
+#' 2 d 4
+#' 3 d 5
+#' @export
+summarize_map <- function(
+  data,
+  x_cols,
+  y_cols,
+  n_examples = 4,
+  verbose = FALSE) {
+
+    # convert column selections named vectors of column indices into data
+    x_cols <- tidyselect::eval_select(rlang::enquo(x_cols), data)
+    y_cols <- tidyselect::eval_select(rlang::enquo(y_cols), data)
+    xUy_cols <- union(x_cols, y_cols)
+    names(xUy_cols) <- names(data[xUy_cols])
+    
+    if(verbose) {
+        cat("The following is a report of the relationship between two different ways of identifying instances\n")
+    }
+
+    # example rows
+    problems <- list()
+
+    count_xUy <- data |>
+        dplyr::count(dplyr::across(tidyselect::all_of(xUy_cols))) |>
+        dplyr::ungroup()
+    count_x <- count_xUy |>
+        dplyr::count(dplyr::across(tidyselect::all_of(names(x_cols))), name = "size") |>
+        dplyr::ungroup()
+    count_y <- count_xUy |>
+        dplyr::count(dplyr::across(tidyselect::all_of(names(y_cols))), name = "size") |>
+        dplyr::ungroup()
+    
+    if (verbose) {
+        cat("\nProperties of X identifiers:\n")
+    }
+    cat("X<-[", paste(names(x_cols), collapse = ", "), "]:\n", sep = "")
+    cat("  |X|: ", count_x |> stats::na.omit(method = "r") |> nrow(), sep = "")
+    
+    na_count <- data |>
+        dplyr::select(tidyselect::all_of(x_cols)) |>
+        stats::complete.cases() |>
+        magrittr::not() |>
+        sum()
+    cat(
+        ifelse(
+            na_count == 0,
+            "",
+            paste0(" (", na_count, " NA)")),
+        "\n", sep = "")
+
+    size_dist <- count_x |>
+        stats::na.omit(method = "r") |>
+        dplyr::count(size) |>
+        dplyr::ungroup()
+    if (nrow(size_dist) < 12) {
+        cat("  count*size: ",
+            paste(size_dist$n, size_dist$size, sep = "*", collapse = ", "),
+            "\n", sep = "")
+	} else {
+            top <- 1:6
+            bottom <- (nrow(size_dist) - 6+1):nrow(size_dist)
+            cat("  count*size: ",
+                paste(
+                    size_dist$n[top],
+                    size_dist$size[top], sep = "*", collapse = ", "),
+                ", ... ",
+                paste(
+                    size_dist$n[bottom],
+                    size_dist$size[bottom], sep = "*", collapse = ", "),
+                "\n", sep="")
+	}
+
+    if (verbose) {
+        cat("\nProperties of the Y identifiers:\n")
+    }
+    cat("Y<-[", paste(names(y_cols), collapse = ", "), "]:\n", sep = "")
+    cat("  |Y|: ", count_y |> stats::na.omit(method = "r") |> nrow(), sep = "")
+    na_count <- data |>
+        dplyr:::select(tidyselect::all_of(y_cols)) |>
+        stats::complete.cases() |>
+        magrittr::not() |>
+        sum()
+    cat(ifelse(na_count == 0, "", paste0(" (", na_count, " NA)")), "\n", sep = "")
+    
+    size_dist <- count_y |>
+        stats::na.omit(method = "r") |>
+        dplyr::count(size) |>
+        dplyr::ungroup()
+    if (nrow(size_dist) < 12) {
+        cat("  count*size: ",
+            paste(size_dist$n, size_dist$size, sep = "*", collapse = ", "),
+            "\n", sep = "")
+    } else {
+        top <- 1:6
+        bottom <- (nrow(size_dist) - 6+1):nrow(size_dist)
+        cat("  count*size: ",
+            paste(
+                size_dist$n[top],
+                size_dist$size[top], sep = "*", collapse = ", "),
+            ", ... ",
+            paste(
+                size_dist$n[bottom],
+                size_dist$size[bottom], sep = "*", collapse = ", "),
+            "\n", sep="")
+    }
+    
+    if (verbose) {
+        cat("\nProperties of the intersection of union of the X and Y identifiers:\n")
+    }
+    cat("[X U Y]:\n")
+    cat("  |X U Y|: ", count_xUy |> stats::na.omit(method = "r") |> nrow(), sep = "")
+    na_count <- data |>
+        dplyr:::select(!!!xUy_cols) |>
+        stats::complete.cases() |>
+        magrittr::not() |>
+        sum()
+    cat(ifelse(na_count == 0, "", paste0(" (", na_count, " NA)")), "\n", sep = "")
+    
+    size_dist <- count_xUy |>
+        stats::na.omit(method = "r") |>
+        dplyr::rename(size = n) |>
+        dplyr::count(size) |>
+        dplyr::ungroup()
+    if (nrow(size_dist) < 12) {
+        cat("  count*size: ",
+            paste(size_dist$n, size_dist$size, sep = "*", collapse = ", "),
+            "\n", sep="")
+    } else {
+        top <- 1:6
+        bottom <- (nrow(size_dist) - 6+1):nrow(size_dist)
+        cat("  count*size: ",
+            paste(
+                size_dist$n[top],
+                size_dist$size[top], sep = "*", collapse = ", "),
+            ", ... ",
+            paste(
+                size_dist$n[bottom],
+                size_dist$size[bottom], sep = "*", collapse = ", "),
+            "\n", sep = "")
+    }
+    
+
+    count_xUy <- count_xUy |> stats::na.omit(method = "r")
+    
+    if (verbose) {
+        cat("Properties of the intersection of the X and Y identifiers:\n")
+    }
+    cat("[X @ Y]:\n")
+    if (verbose) {
+        cat("  Number of X and Y identifiers that are 1 to 1:\n")
+    }
+    cat("  |X ~ Y|: ",
+        count_xUy |>
+        dplyr::semi_join(
+            count_x |> dplyr::filter(size == 1),
+            by = names(x_cols)) |>
+        dplyr::semi_join(
+            count_y |> dplyr::filter(size == 1),
+            by = names(y_cols)) |>
+        nrow(),
+        "\n", sep = "")
+    
+    if (verbose) {
+        cat("  Number of X and Y identifiers where an X identifier maps to multiple Y identifiers:\n")
+    }
+    cat(
+        "  |X:X < Y|, |Y:Y < X|: ",
+        count_xUy |>
+        dplyr::semi_join(
+            count_x |> dplyr::filter(size > 1),
+            by = names(x_cols)) |>
+        nrow(),
+        ", ",
+        count_xUy |>
+        dplyr::count(
+            dplyr::across(tidyselect::all_of(names(x_cols))),
+            name = "size") |>
+        dplyr::filter(size > 1) |>
+        nrow(),
+        "\n", sep = "")
+    
+    if (verbose) {
+        cat(
+            "  Number of X and Y identifiers where a Y identifier maps to ",
+            "multiple X identifiers:\n")
+    }
+    cat(
+        "  |X:X > Y|, |Y:Y > X|: ",
+        count_xUy |>
+        dplyr::semi_join(
+            count_y |>
+            dplyr::filter(size > 1),
+            by = names(y_cols)) |>
+        nrow(),
+        ", ",
+        count_xUy |>
+        dplyr::count(
+            dplyr::across(tidyselect::all_of(names(y_cols))),
+            name = "size") |>
+        dplyr::filter(size > 1) |>
+        nrow(),
+        "\n", sep = "")
+    
+    #is.na.X
+    ex_rows <- data |>
+        dplyr:::select(tidyselect::all_of(x_cols)) |>
+        stats::complete.cases() |>
+        magrittr::not() |>
+        which()
+    if (length(ex_rows)) {
+        if (!is.null(n_examples) && (n_examples < length(ex_rows))) {
+            ex_rows <- ex_rows |> sample(n_examples, replace = FALSE)
+        }
+        problems$is.na.X <- data |>
+            dplyr::slice(ex_rows) |>
+            dplyr::arrange(dplyr::across(tidyselect::all_of(names(x_cols))))
+    }
+    
+    #is.na.Y
+    ex_rows <- data |>
+        dplyr:::select(tidyselect::all_of(y_cols)) |>
+        stats::complete.cases() |>
+        magrittr::not() |>
+        which()
+    if (length(ex_rows)) {
+        if (!is.null(n_examples) && (n_examples < length(ex_rows))) {
+            ex_rows <- ex_rows |> sample(n_examples, replace = FALSE)
+        }
+        problems$is.na.Y <- data |>
+            dplyr::slice(ex_rows) |>
+            dplyr::arrange(dplyr::across(tidyselect::all_of(names(y_cols))))
+    }
+    
+    #dup.X
+    dup.X <- count_xUy |>
+        dplyr::filter(n == 1) |>
+        dplyr::count(
+            dplyr::across(tidyselect::all_of(names(x_cols))),
+            name = "size") |>
+        dplyr::filter(size > 1) |>
+        dplyr::ungroup() |>
+        dplyr:::select(-size)
+    if (nrow(dup.X) > 1) {
+        if (!is.null(n_examples) && (n_examples < nrow(dup.X))) {
+            dup.X <- dup.X |> dplyr::sample_n(n_examples, replace = FALSE)
+        }
+        problems$dup.X <- dup.X |>
+            dplyr::left_join(data, by = names(x_cols)) |>
+            dplyr::arrange(dplyr::across(tidyselect::all_of(names(x_cols))))
+    }
+    
+    #dup.Y
+    dup.Y <- count_xUy |>
+        dplyr::filter(n == 1) |>
+        dplyr::count(
+            dplyr::across(tidyselect::all_of(names(y_cols))),
+            name = "size") |>
+        dplyr::filter(size > 1) |>
+        dplyr::ungroup() |>
+        dplyr:::select(-size)
+    if (nrow(dup.Y) > 1) {
+        if (!is.null(n_examples) && (n_examples < nrow(dup.Y))) {
+            dup.Y <- dup.Y |> dplyr::sample_n(n_examples, replace = FALSE)
+        }
+        problems$dup.Y <- dup.Y |>
+            dplyr::left_join(data, by = names(ycols)) |>
+            dplyr::arrange(dplyr::across(tidyselect::all_of(names(y_cols))))
+    }
+    
+    #dup.XUY
+    dup.XUY <- count_xUy |>
+        dplyr::filter(n > 1) |>
+        dplyr:::select(-n)
+    if (nrow(dup.XUY) > 1) {
+        if (!is.null(n_examples) && (n_examples < nrow(dup.XUY))) {
+            dup.XUY <- dup.XUY |> dplyr::sample_n(n_examples, replace = FALSE)
+        }
+        problems$dup.XUY <- dup.XUY |>
+            dplyr::left_join(data, by = names(xUy_cols)) |>
+            dplyr::arrange(dplyr::across(tidyselect::all_of(names(xUy_cols))))
+    }
+    if (verbose) {
+        cat("Returned instances where:\n")
+        cat("\tis.na.X: The X identifier is NA\n")
+        cat("\tis.na.Y: The Y identifier is NA\n")
+        cat("\tdup.X:   The X identifier is not unique\n")
+        cat("\tdup.Y:   The Y identifier is not unique\n")
+        cat("\tdup.XUY: The X and Y identifiers together are not unique\n")
+    }
+    problems
+}