removing md5,adding mindel,converting to more familiar file formats and updating readme

README.md

@@ -3,165 +3,54 @@ license: mit
 pretty_name: BrentLab Yeast Genome Resources
 language:
   - en
-dataset_info:
-  features:
-    - name: start
-      dtype: int32
-      description: Start coordinate (1-based, **inclusive**)
-    - name: end
-      dtype: int32
-      description: End coordinate (1-based, **inclusive**)
-    - name: strand
-      dtype: string
-      levels:
-        - +
-        - "-"
-      description: Strand of feature
-    - name: type
-      dtype: string
-      levels:
-        - gene
-        - ncRNA_gene
-        - tRNA_gene
-        - snoRNA_gene
-        - transposable_element_gene
-        - pseudogene
-        - telomerase_RNA_gene
-        - snRNA_gene
-        - rRNA_gene
-        - blocked_reading_frame
-      description: classification of feature
-    - name: locus_tag
-      dtype: string
-      description: Systematic ID of feature
-    - name: symbol
-      dtype: string
-      description: Common name of feature
-    - name: alias
-      dtype: string
-      description: Alternative names of feature, typically alternative symbols
-    - name: source
-      dtype: string
-      description: Annotation file version/origin of the feature
-    - name: note
-      dtype: string
-      description: Additional feature information, typically the description from the
-        SGD gff/gtf
-  partitioning:
-    keys:
-      - name: chr
-        dtype: string
-        levels:
-          - chrI
-          - chrII
-          - chrVII
-          - chrV
-          - chrIII
-          - chrIV
-          - chrVIII
-          - chrVI
-          - chrX
-          - chrIX
-          - chrXI
-          - chrXIV
-          - chrXII
-          - chrXIII
-          - chrXV
-          - chrXVI
-          - chrM
-configs:
-  - config_name: features
-    default: true
-    data_files:
-      - split: train
-        path:
-          - features/*/part-0.parquet
 ---
 # BrentLab Yeast Genome Resources
 
-This Dataset stores resources meant to make data exploration of yeast -omic data,
-curated by the Brent Lab, easier and more harmonious.  
-To cite, [cite
-SGD](https://sites.google.com/view/yeastgenome-help/about#h.p_Qck4DmfRP5zN)
-
-This repo provides 4 datasets:
-
-- **brentlab_features**: BrentLab Yeast Genome Resources.
-- **gal_tss_sgd-5-1_verified_orf**: Genomic coordinates of Transcription Start Sites
-  (TSS) for verified ORFs under Galactose culture conditions.
-- **median_across_conds_tss_sgd-5-1_verified_orf**: Median Transcription Start Site
+This Dataset stores resources meant to aid in the exploration of yeast -omic data,
+curated by the Brent Lab.  
+
+## Terminology
+
+Across all datasets in the BrentLab collection, we use the following terms consistently
+
+- **locus_tag**: The systematic ID of an ORF. Eg, YKL038W
+- **symbol**: The common name of an ORF. Eg, RGT1
+- **target**: when the genomic locus is the 'target' of location of measurement, then
+  it is referred to as a 'target'. Eg, in RNAseq, the column `target_locus_tag` would
+  store the counts over that gene.
+- **regulator**: This collection is made up of binding location assays, and perturbation
+  of transcription factors and chromatin interacting proteins, which I refer to generally
+  as 'regulators'
+
+This repo provides the following:
+
+You can find these by clicking on the [files and versions](https://huggingface.co/datasets/BrentLab/yeast_genome_resources/tree/main) tab. The file format is noted.
+
+- **brentlab_features** (csv): This is a simplified version of the
+  SGD S288C-R64-3-1 annotations. We have used this file to standardize target
+  and regulator `locus_tag` and `symbol` across the other datasets in this collection.
+- **yiming_promoters** (bed): The promoter regions used in
+  [Kang et al in the Dual threshold optimization paper](https://genome.cshlp.org/content/30/3/459).
+  These promoter regions are used for callingcards.
+- **mindel_promoters** (csv): The promoters used by the Barkai lab to evaluate the
+  overlap between perturbation and binding in the Mahendrawada 2025 set in
+  [this preprint](https://www.biorxiv.org/content/10.1101/2025.10.12.681120v1.abstract).
+  See scripts/create_promoter_bed_from_mindel.R.
+- **gal_tss_sgd-5-1_verified_orf** (bed): Genomic coordinates of Transcription Start Sites
+  (TSS) for verified ORFs under Galactose culture conditions. These are derived from
+  annotation version S288C-R64-5-1. See scripts/create_tss_files.R
+- **median_across_conds_tss_sgd-5-1_verified_orf** (bed): Median Transcription Start Site
   (TSS) coordinates for verified ORFs aggregated across different experimental
-  conditions.
-- **ypd_tss_sgd-5-1_verified_orf**: Genomic coordinates of Transcription Start Sites
+  conditions. See scripts/create_tss_files.R
+- **ypd_tss_sgd-5-1_verified_orf** (bed): Genomic coordinates of Transcription Start Sites
   (TSS) for verified ORFs under standard YPD (rich medium) culture conditions.
+  See scripts/create_tss_files.R
+- **intergenic_regions_5_1** (bed and fasta): Both the genomic coordinates and sequences
+  of the intergenic regions in version S288C-R64-5-1. See scripts/parse_intergenic_regions.R
+- **chrmap** (csv): This file provides a mapping between chromosome names between, eg
+  UCSC, ensembl, etc.
 
 ## Usage
 
-The python package `tfbpapi` provides an interface to this data which eases
-examining the datasets, field definitions and other operations. You may also 
-download the parquet datasets directly from hugging face by clicking on
-"Files and Versions", or by using the huggingface_cli and duckdb directly.
-In both cases, this provides a method of retrieving dataset and field definitions.
-
-### `tfbpapi`
-
-After [installing tfbpapi](https://github.com/BrentLab/tfbpapi/?tab=readme-ov-file#installation),
-you can adapt this [tutorial](https://brentlab.github.io/tfbpapi/tutorials/hfqueryapi_tutorial/)
-in order to explore the contents of this repository.
-
-### huggingface_cli/duckdb
-
-You can retrieves and displays the file paths for each configuration of
-the "BrentLab/yeast_genome_resources" dataset from Hugging Face Hub.
-
-```python
-from huggingface_hub import ModelCard
-from pprint import pprint
-
-card = ModelCard.load("BrentLab/yeast_genome_resources", repo_type="dataset")
-
-# cast to dict
-card_dict = card.data.to_dict()
-
-# Get partition information
-dataset_paths_dict = {d.get("config_name"): d.get("data_files")[0].get("path") for d in card_dict.get("configs")}
-
-pprint(dataset_paths_dict)
-```
-
-You may access just the Dataset metadata like this:
-```python
-from huggingface_hub import ModelCard
-
-card = ModelCard.load("BrentLab/yeast_genome_resources", repo_type="dataset")
-
-# cast to dict
-card_dict = card.data.to_dict()
-
-# Get partition information
-card_dict.get("dataset_info").get("partitioning").get("keys")
-```
-
-If you wish to pull the entire repo, due to its size you may need to use an
-[authentication token](https://huggingface.co/docs/hub/en/security-tokens).
-If you do not have one, try omitting the token related code below and see if
-it works. Else, create a token and provide it like so:
-
-```python
-repo_id = "BrentLab/yeast_genome_resources"
-
-hf_token = os.getenv("HF_TOKEN")
-
-# Download entire repo to local directory
-repo_path = snapshot_download(
-    repo_id=repo_id,
-    repo_type="dataset",
-    token=hf_token
-)
-
-print(f"\n✓ Repository downloaded to: {repo_path}")
-
-# Construct path to the rossi_annotated_features parquet file
-parquet_path = os.path.join(repo_path, "brentlab_features.parquet")
-print(f"✓ Parquet file at: {parquet_path}")
-```
+Currently, we expect that this will be used for its raw files, eg by downloading a given
+file and opening it in your favorite program.

brentlab_features.csv.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8af3315ab44818c28f009fcc879d9d950b4fc5f6eb446cb7d43370c984fd305
+size 690751

brentlab_features.parquet.md5

@@ -1 +0,0 @@
-d1d88c50a03846b23894d1c20a86b231  brentlab_features.parquet

gal_tss_sgd-5-1_verified_orf.bed.md5

@@ -1 +0,0 @@
-fba1b195e3ba61575ff63d8f7e332173  gal_tss_sgd-5-1_verified_orf.bed

intergenic_regions_5_1.bed.md5

@@ -1 +0,0 @@
-328a943c457125e070d760f135077cad  intergenic_regions_5_1.bed

intergenic_regions_5_1.fasta.gz.md5

@@ -1 +0,0 @@
-a6406d6da363177fe2be496035e24f33  intergenic_regions_5_1.fasta.gz

intergenic_regions_metadata_5_1.csv.md5

@@ -1 +0,0 @@
-b665e559400eaa0f2e08e6dd4f807c21  intergenic_regions_metadata_5_1.csv

median_across_conds_tss_sgd-5-1_verified_orf.bed.md5

@@ -1 +0,0 @@
-d997a383e9eed6ccf6a3fa4e035fafcd  median_across_conds_tss_sgd-5-1_verified_orf.bed

mindel_promoters.csv.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab77901f7c665f3c95e8ae5723c04bfa57d6070df490e6698568fce52be54d14
+size 1434190

mindel_promoters.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52c702ef6d75e87d5b193589bbf633506ff5b728c7ea85700008f2f1becf8b46
+size 2244143

scripts/create_promoter_bed_from_mindel.R

@@ -0,0 +1,137 @@
+library(tidyverse)
+library(here)
+library(Biostrings)
+
+# see https://scholar.google.com/citations?view_op=view_citation&hl=en&user=cW9tmRMAAAAJ&sortby=pubdate&citation_for_view=cW9tmRMAAAAJ:mlAyqtXpCwEC
+#
+# and the associated github repo
+# https://github.com/vmindel/ChEC_Target_Selection/tree/main
+#
+# I am using data/prom_seqs
+
+chrmap = read_csv("~/code/hf/yeast_genome_resources/chrmap.csv.gz")
+
+prom_seq = read_csv("~/projects/ChEC_Target_Selection/data/promoter_sequences.csv") %>%
+    mutate(norm_seq = ifelse(strand == "positive", tolower(seq), tolower(rseq)))
+# note that this has the orf coordinates. Additionally, the start/stop are not
+# oriented correctly by strand, so a negative strand start is greater than
+# its stop. Not useful.
+prom_orf_df = read_csv("~/projects/ChEC_Target_Selection/data/promoter_definitions.csv") %>%
+    select(-`...1`) %>%
+    left_join(
+        chrmap %>%
+            select(numbered, ucsc) %>%
+            mutate(numbered = as.integer(numbered)) %>%
+            dplyr::rename(chr_loc = numbered,
+                          chr = ucsc)) %>%
+    select(-chr_loc) %>%
+    dplyr::relocate(chr)
+
+# at this point, I wrote the seqs out to a fasta
+
+# Write sequences to FASTA
+seqs <- DNAStringSet(toupper(prom_seq$seq))
+names(seqs) <- prom_seq$name
+writeXStringSet(seqs, here("data/prom_seqs.fa"))
+
+# installed blat in a conda
+# env with conda create -n blat -c bioconda blat
+
+# and then ran blat on the seqs
+# blat sacCer3.2bit prom_seqs.fa output.pslx -out=pslx -minScore=700
+
+pslx = read_tsv(here("data/output.pslx"), skip = 5, col_names = c(
+    "matches", "misMatches", "repMatches", "nCount",
+    "qNumInsert", "qBaseInsert", "tNumInsert", "tBaseInsert",
+    "strand", "qName", "qSize", "qStart", "qEnd",
+    "tName", "tSize", "tStart", "tEnd",
+    "blockCount", "blockSizes", "qStarts", "tStarts", "qSeq", "tSeq")) %>%
+    mutate(qSeq = trimws(str_remove(qSeq, ",")))
+
+# then I read in the blat output and joined it to the prom_seqs and prom_orf_df
+# in order to filter down to the correct promoter alignments
+# NOTE: i verified that after all of this, the qSeq == norm_seq for all
+promoter_coordinates_df = pslx %>%
+    # join on the orf coordinates -- these will be used to filter out among
+    # promoters with multiple perfect full length alignments (there are some!)
+    left_join(prom_orf_df, by = c('qName' = 'name', 'tName' = 'chr')) %>%
+    # accept only perfect alignments on the same chr as the orf with
+    # no mismatches or inserts
+    filter(complete.cases(.), misMatches == 0, qNumInsert==0, qBaseInsert==0) %>%
+    # there are some loci with multiple perfect alignments (probably due to
+    # repeats). count these and add as a column
+    group_by(qName) %>%
+    mutate(promoter_exact_aligns = n()) %>%
+    ungroup() %>%
+    # the correct alignment should start at the 5' end of the gene
+    arrange(qName) %>%
+    mutate(corr_align = case_when(
+        strand == "-" & tStart == start ~ TRUE,
+        strand == "+" & tEnd == start ~ TRUE,
+        .default = FALSE)) %>%
+    # only accept those alignments that begin at the 5' end of the orf
+    filter(corr_align) %>%
+    arrange(desc(promoter_exact_aligns)) %>%
+    ungroup() %>%
+    select(tName, tStart, tEnd, qName, qSize, strand, qSeq, promoter_exact_aligns)
+
+
+# map the features to the brent and mahendrawada features
+mahendrawada_features = arrow::read_parquet("~/code/hf/mahendrawada_2025/features_mahendrawada_2025.parquet")
+brent_features = arrow::read_parquet("~/code/hf/yeast_genome_resources/brentlab_features.parquet")
+
+brent_features_map = promoter_coordinates_df %>%
+    select(qName) %>%
+    # try to join by symbol first
+    left_join(select(brent_features, locus_tag, symbol),
+              by = c('qName' = 'symbol')) %>%
+    # next, where the symbol is actually a locus tag, join by locus tag
+    left_join(
+        select(brent_features, locus_tag) %>%
+            mutate(tmp = locus_tag) %>%
+            dplyr::rename(qName = locus_tag)) %>%
+    # where there is a tmp (meaning it is actually a locus_tag), replace
+    # the NA in locus_tag with that
+    mutate(locus_tag = ifelse(is.na(locus_tag), tmp, locus_tag)) %>%
+    # drop the tmp column
+    select(-tmp) %>%
+    # next, search the aliases for a match to the qName. If one is found, then
+    # use the corresponding locus_tag. Like before, where the locus_tag column
+    # is NA and an alias was found, replace locus_tag with the corresponding
+    # value in tmp and then drop the tmp column
+    mutate(tmp = map_chr(qName, function(q) {
+        idx <- str_detect(brent_features$alias, fixed(q))
+        idx[is.na(idx)] <- FALSE
+        if (any(idx)) brent_features$locus_tag[which(idx)[1]] else NA_character_})) %>%
+    mutate(locus_tag = ifelse(is.na(locus_tag), tmp, locus_tag)) %>%
+    select(-tmp) %>%
+    # remove this -- it is a -1 frameshift from AAD6. The fact the coordinates
+    # are different in this, and the two therefore had different promoters
+    # makes me question the value of these loci. I removed both
+    filter(!qName %in% c("AAD6", "AAD16")) %>%
+    # after this, only AAD16, ADE5_7, ARG5_6 and DUR1_2 are left. These were
+    # identified by searching SGD
+    mutate(locus_tag = case_when(
+        qName == "ADE5_7" ~ "YGL234W",
+        qName == "ARG5_6" ~ "YER069W",
+        qName == "DUR1_2" ~ "YBR208C",
+        .default = locus_tag)) %>%
+    left_join(
+        select(brent_features, locus_tag, symbol)) %>%
+    mutate(in_mahendrawada_features = locus_tag %in% mahendrawada_features$gene_id)
+
+promoter_coordinates_final = promoter_coordinates_df %>%
+    left_join(brent_features_map) %>%
+    dplyr::rename(
+        chr = tName,
+        start = tStart,
+        end = tEnd,
+        mindel_name = qName,
+        target_locus_tag = locus_tag,
+        target_symbol = symbol,
+        promoter_sequence = qSeq) %>%
+    select(chr, start, end, mindel_name, target_locus_tag, target_symbol,
+           strand, promoter_sequence,
+           in_mahendrawada_features, promoter_exact_aligns)
+
+# write_csv(promoter_coordinates_final, "~/code/hf/yeast_genome_resources/mindel_promoters.csv.gz")

ypd_tss_sgd-5-1_verified_orf.bed.md5

@@ -1 +0,0 @@
-a97586b6a73b5ff0d368786b72227186  ypd_tss_sgd-5-1_verified_orf.bed