VarChAMP sqY2H protein-protein interaction screen: scored colony measurements, manual rescore, plate maps and sequence confirmation

Input files of the sqY2H (semi-quantitative yeast two-hybrid) scoring pipeline of VarChAMP (GitHub: varchamp-ppi-sqy2h). The screen tests disease-gene baits, wild type and missense variants from the VarChAMP 1% allele collection, against a panel of interactor preys. Four replicate spots are made per bait-prey pair, on a mating-control medium and on selective media of increasing stringency, in four retest batches. The pipeline builds from these files: a consensus growth score per bait, prey and medium; a log2fc per variant edge against its WT bait; an edge-level perturbation log-likelihood ratio (two-component mixture model); a variant-level profile distance (norm_dist_rms). Alleles are annotated from the VarChAMP allele-metadata table (doi:10.5281/zenodo.23376506). The pipeline checks every file against the sha256 pinned in its configs/data.json. The colony measurements come from an automated imaging workflow (FIJI registration, grid alignment, CellProfiler segmentation) by Ahmad Kamal Hamid, Cimini Lab, Imaging Platform, Broad Institute of MIT and Harvard. Its output table is the spot table below; the imaging workflow is not included. Version 1.1 adds the original day-4 plate photographs, from which the pipeline's image mode scores the colonies directly instead of using that table's automated measures. Files File Content sqY2H_manual_automated_annotated_scores_size.csv The spot table: one row per spot, 306,431 rows × 20 columns. Batches VUSAPWT1B1, VUSAPWT1B2, VUSAPWT2B1 and VUSAPWT6B1; 329 DB baits and 963 AD preys. Columns below. sqy2h_rescore_20251219.tar.gz Manual rescore of the scoring plates (2025-12-19): 532 headerless TSVs, one per scoring plate (format below). The pipeline uses these scores in place of the spot table's manual_score_growth. sqy2h_day4_plate_images.tar The 138 day-4 photographs of the megaplates, the day of the automated and manual scores: one 3456 × 3456 JPG per large plate, camera originals (Canon EOS Rebel SL1). Format below. mapping_y2h.csv Cherry-picking map of the retest plates: 12,587 rows × 16 columns. Per retest well: the AD and DB ORF ids and symbols, the mutation ids, the retest plate and position (96-well), the retest and standard batch, the assay, and the source plates and wells of both constructs. The pipeline takes the prey symbols from it. vus_ad_interactors.csv Sequence confirmation of the AD preys: 1,056 rows (prey × assay). ad_orf_id, symbol, source and consolidated plate and well, assay_id, swim_seq_validated. vus_db_alleles.csv Sequence confirmation of the DB variant baits: 2,545 rows. orf_id, mut_id, nt_change, original_project, orf_var_id, the plate and well of the clone in each DB batch (Y8Dh_*, VUSB1_mut_db_* … VUSB6_mut_db_*, empty-well flags), swim_seq_validated. wt_disease_nodes_summary.csv Sequence confirmation and summary of the DB wild-type baits: 342 rows. disease_orf_id, disease_symbol, number of alleles, interactor counts and ids (gene and ORF level, per AD assay), source collection and consolidated entry plate and well, swim_seq_validated. In the three sequence-confirmation files, swim_seq_validated is 1 (confirmed), 0 (failed) or empty (no result). Spot table columns Column Meaning scoring_pla, scoring_pos Scoring plate (<batch>r<NN>_<plate>, NN 01–06 by medium) and well (A01–P24, 384-well) large_plate_name Megaplate image: six 384-well scoring plates retest_batch, assay_id Retest batch; assay (1 for VUSAPWT1B1 and 1B2, 2 for 2B1, 6 for 6B1) condition, condition_name Medium code and name: LW (mating control, no selection); LWH1, LWH10, LWH25 (selective, 1, 10 and 25 mM 3-AT); LWA and LWAH1 (adenine media, not used by the pipeline) media_control True for the media-control wells of the first plate of each megaplate, False for assay wells, empty for empty wells ad_orf_id, ad_mut_id AD prey ORF (0 = empty AD, the autoactivation control); ad_mut_id is empty db_orf_id, db_mut_id DB bait ORF and mutation id (0 = wild type, 999999 = RRS control bait). Variants join the allele-metadata table on (orf_id, mut_id). db_source_plate, db_source_well, ad_source_plate, ad_source_well Source plates and wells of the two constructs per_well_integrated_intensity_gray Automated growth measure: integrated grey-level intensity of the colony objects assigned to the well (CellProfiler) total_colony_area Area of those colony objects, in pixels Children_reExpandedColonies_Count Number of colony objects assigned to the well manual_score_growth Manual visual score, 0 (no growth) to 4 (strong growth); empty: not scorable (contamination). Superseded by the rescore Rescore files Paths: sqy2h_rescore_20251219/<batch>/<medium dir>/<batch>a<NN>_<plate>.tsv. Format: one line per well, tab-separated, no header. Columns: well (A01–P24); manual score 0–4 (NA = not scorable, treated as contamination); an iscore column (often NA, not used); a growth label. Medium codes (NN): 07 or 13 = LW; 09 or 15 = LWH1; 10 or 16 = LWH10; 11 or 17 = LWH25. Batches VUSAPWT1B1 and 1B2 use 07–11; 2B1 and 6B1 use 13–17. Joining to the spot table: replace a with r in the file name and code NN with (NN − 7) mod 6 + 1; that gives scoring_pla. Then join on (scoring_pla, scoring_pos). Reproducible archive: members are sorted by name, with fixed timestamps and permissions, so re-packing the same files gives the same checksum. Plate images Paths: sqy2h_day4_plate_images/<large_plate_name>.JPG, e.g. VUSAPWT1B1r01_001_002_003_004_005_006.JPG: batch, medium code r<NN> (as in scoring_pla) and the six scoring plates of the megaplate. The name joins the spot table's large_plate_name. Taken on 2020-06-30 (VUSAPWT1B1), 2020-07-10 (VUSAPWT1B2) and 2020-07-20 (VUSAPWT2B1 and VUSAPWT6B1). The same photographs are in doi:10.5281/zenodo.15020115 (sqY2H images for CellProfiler), but 114 of the 138 there are re-encoded copies; the files here are the originals, byte for byte, which the pipeline pins by SHA-256. Plain tar (JPGs do not compress), packed like the rescore archive. Credits Colony quantification workflow: Ahmad Kamal Hamid, Cimini Lab, Imaging Platform, Broad Institute of MIT and Harvard.