Transfer learning reveals sequence determinants of the quantitative response to transcription factor dosage

<p>Processed data and code for "Transfer learning reveals sequence determinants of the quantitative response to transcription factor dosage," Naqvi et al 2025.</p> <p>Directory is organized into the following subfolders, each tar'ed and gzipped:</p> <p><strong>data_analysis.tar.gz - Processed data for modulation of TWIST1 levels and calculation of RE responsiveness to TWIST1 dosage</strong></p> <ul> <li>atac_design.txt - design matrix for ATAC-seq TWIST1 titration samples</li> <li>all.sub.150bpclust.greater2.500bp.merge.TWIST1.titr.ATAC.counts.txt - ATAC-seq counts from all samples over all reproducible ATAC-seq peak regions, as defined in Naqvi et al 2023</li> <li>atac_deseq_fitmodels_moded50.R - R code for calculating new version of ED50 and response to full depletion from TWIST1 titration data (note, uses drm.R function from <a href="https://doi.org/10.5281/zenodo.7689948">10.5281/zenodo.7689948</a>,&nbsp;install drc() with this version to avoid errors)</li> </ul> <p><strong>baseline_models.tar.gz - Code and data for training baseline models to predict RE responsiveness to SOX9/TWIST1 dosage</strong></p> <ul> <li>{sox9|twist1}.{0v100|ed50}.{train|valid|test}.txt - Training/testing/validation data (ED50 or full TF depletion effect for SOX9 or TWIST1), split into train/test/validation folds</li> <li>HOCOMOCOv11_core_HUMAN_mono_jaspar_format.all.sub.150bpclust.greater2.500bp.merge.minus300bp.p01.maxscore.mat.cpg.gc.basemean.txt.gz - matrix of predictors for all REs. Quantitative encoding of PWM match for all HOCOMOCO motifs + CpG + GC content, plus unperturbed ATAC-seq signal</li> <li>train_baseline.R - R code to train baseline (LASSO regression or random forest) models using predictor matrix and the provided training data. <ul> <li>Note: training the random forest to predict full TF depletion is computationally intensive because it is across all REs, if doing this run on CPU for ~6 hrs.&nbsp;</li> </ul> </li> </ul> <p><strong>chrombpnet_models.tar.gz - Remainder of code, data, and models for fine-tuning and interpreting ChromBPNet mdoels to </strong><strong>predict RE responsiveness to SOX9/TWIST1 dosage</strong></p> <ul> <li>Fine-tuning code, data, models <ul> <li>{all|sox9.direct|twist1.bound.down}.{train|valid|test}.{ed50|0v100.log2fc}.txt - Training/testing/validation data (ED50 or full TF depletion effect for SOX9 or TWIST1), split into train/test/validation folds</li> <li>pretrained.unperturbed.chrombpnet.h5 - Pretrained model of unperturbed ATAC-seq signal in CNCCs, obtained by running ChromBPNet (https://github.com/kundajelab/chrombpnet) on DMSO-treated SOX9/TWIST1-tagged ATAC-seq data</li> <li>finetune_chrombpnet.py - code for fine-tuning the pretrained model for any of the relevant prediction tasks (ED50/ effect of full TF depletion for SOX9/TWIST1)</li> <li>best.model.chrombpnet.{0v100|ed50}.{sox9|twist1}.h5 - output of finetune_chrombpnet.py, best model after 10 training epochs for the indicated task</li> <li>chrombpnet.{0v100|ed50}.{sox9|twist1}.contrib.{h5|bw} - contribution scores for the indicated predictive model, obtained by running chrombpnet contribs_bw on the corresponding model h5 file.</li> <li>chrombpnet.{0v100|ed50}.{sox9|twist1}.contrib.modisco.{h5|bw} - TF-MoDIsCo output from the corresponding contribution score file</li> </ul> </li> <li>Interpretation code, data, models <ul> <li>contrib_h5_to_projshap_npy.py - code to convert contrib .h5 files into .npy files containing projected SHAP scores (required because the CWM matching code takes this format of contribution scores)</li> <li>sox9.direct.10col.bed, twist1.bound.down.10col.uniq.bed - regions over which CWMs will be matched (likely direct targets of each TF)</li> <li>match_cwms.py - Python code to match individual CWM instances. Takes as input: modisco .h5 file, SHAP .npy file, bed file of regions to be matched. Output is a bed file of all CWM matches (not pruned, contains many redundant matches).</li> <li>chrombpnet.ed50.{sox9|twist1}.contrib.perc05.matchperc10.allmatch.bed - output of match_cwms.py&nbsp;</li> <li>take_max_overlap.py - code to merge output of match_cwms.py into clusters, and then take the maximum (length-normalized) match score in each cluster as the representative CWM match of that cluster. Requires upstream bedtools commands to be piped in, see example usage in file.&nbsp;</li> <li>chrombpnet.ed50.{sox9|twist1}.contrib.perc05.matchperc10.allmatch.maxoverlap.bed - output of &nbsp;take_max_overlap.py. These CWM instances are the ones used throughout the paper.</li> </ul> </li> </ul> <p><strong>modisco_reports.zip -</strong><strong> TF-MoDIsCo reports from running on the fine-tuned ChromBPNet models</strong></p> <ul> <li>modisco_report_{sox9|twist1}_{0v100|ed50}: folders containing images of discovered CWMs and HTMLs/PDFs of summarized reports from running TF-MoDisCo on the indicated fine-tuned ChromBPNet model</li> </ul> <p><strong>chrombpnet_models_supp.tar.gz - Alternative ChromBPNet mdoels to </strong><strong>predict SOX9/TWIST1 ED50 using varying definitons of direct targets</strong></p> <ul> <li> <p>best.model.chrombpnet.ed50.twist1.3hdn.h5 - TWIST1 direct targets defined using response to full 3h depletion (as was done for SOX9 throughout the rest of the paper)</p> </li> <li> <p>best.model.chrombpnet.ed50.sox9.v5chip.h5 - SOX9 direct targets defined using V5 ChIP-seq from SOX9-tagged lines (as was done for TWIST1 throughout the paper)</p> </li> </ul> <p><strong>mirny_model.tar.gz - Code and data for analyzing and fitting Mirny model of TF-nucleosome competition to observed RE dosage response curves</strong></p> <ul> <li>twist1.strong.multi.only.ed50.cutoff.true.hill.txt - ED50 and signed hill coefficients for all TWIST1-dependent REs with only buffering Coordinators (mostly one or two) and no other TFs' binding sites. "ed50_new" is the ED50 calculation used in this paper.&nbsp;</li> <li>twist1.strong.weak{1|2|3}.ed50.cutoff.true.hill.txt - ED50 and signed hill coefficients for all TWIST1-dependent REs with only buffering Coordinators (mostly one or two) and the indicated number of sensitizing (weak) Coordinators and no other TFs' binding sites. "ed50_new" is the ED50 calculation used in this paper.&nbsp;</li> <li>MirnyModelAnalysis.py - Python code for analysis of Mirny model of TF-nucleosome competition. Contains implementations of analytic solutions, as well as code to fit model to observed ED50 and hill coefficients in the provided data files.</li> </ul> <p><strong>nucleoatac.tar.gz - Output files from running NucleoATAC on merged ATAC-seq from each of 5 TWIST1 dosages</strong></p> <ul> <li>TWIST1_{dosage}_merge.nucmap_combined.bed.gz - see NucleoATAC docs for output format</li> </ul>