7713a08da69691ba499d5b9d44379c43e8cdb608 markd Fri Sep 18 09:27:52 2026 -0700 Adding Transcription Start container with ENCODE 4 PRO-cap and ProCapNet tracks. refs #35528 New superTrack transcriptionStart in the rna group, holding two faceted composites: encode4ProCap with PRO-cap measurements and proCapNet with the model predictions and sequence-contribution scores. hg38 has all three data types, hs1 the predictions only. ENCODE 4 PRO-cap comes from the portal rather than the submitter's hub copy. For each of the six experiments only the plus and minus strand signal of unique reads files of that experiment's default analysis are taken, which drops files superseded by a later reprocessing. ENCODE publishes no pooled file, so the per-replicate files are summed per cell line and strand, the same merge the ProCapNet models were trained on. Total signal is conserved exactly. The published ProCapNet prediction bigWigs store one bedGraph interval per base and hold a literal NaN at every unresolved (N) base on hg38, which makes autoScale and every summary statistic NaN. They are re-encoded into fixedStep sections, a third smaller with no value changed, dropping 164,268,582 NaN bases of 3,088,269,832 on hg38 and none on hs1. Losslessness verified against the originals on random windows across five chromosomes. The composites are faceted rather than plain because a container multiWig under a plain composite is flattened away by hgTrackUi and never drawn. Each cell line is one row with a checkbox per data type, a Sample class facet, ENCODE accession links and a Files column linking each bigWig on hgdownload. Scripts and the cell line configuration are in makeDb/outside/proCapNet; the trackDb stanzas and the faceted metadata tables are generated, not hand edited. Claude-Session: https://claude.ai/code/session_01LAB6jWshLvB7eNXQKWVuW5 diff --git src/hg/makeDb/trackDb/human/proCapNet.html src/hg/makeDb/trackDb/human/proCapNet.html new file mode 100644 index 00000000000..09ac9b6eb21 --- /dev/null +++ src/hg/makeDb/trackDb/human/proCapNet.html @@ -0,0 +1,227 @@ +<h2>Description</h2> + +<p> +ProCapNet is a neural network trained to predict PRO-cap signal from DNA +sequence alone. PRO-cap is a run-on assay that captures the 5' end of each +nascent RNA, so it reports the exact base and strand at which RNA polymerase II +started transcribing, including at enhancers and at unstable transcripts that +RNA-seq and CAGE miss. Six separate models were trained, one on each of six cell +lines with ENCODE PRO-cap data. +</p> + +<p> +This track holds two kinds of output from those models: +</p> + +<ul> +<li><b>Predicted PRO-cap</b>: what the model expects the PRO-cap signal to be, +at every base of the genome, on both strands. The sequence rules that govern +where initiation happens are largely shared between cell types, so any one model +highlights sequence capable of driving initiation, including at regions where no +PRO-cap experiment has been done.</li> +<li><b>Sequence contribution scores</b>: how much each individual base pushed +the model's prediction up or down. Bases inside a functional element such as a +TATA box or an initiator carry high scores, and the pattern of high-scoring +bases often spells out the recognition sequence of a promoter-associated +transcription factor. These are computed only around MANE Select transcription +start sites, and exist for GRCh38 only.</li> +</ul> + +<p> +Predictions are not measurements: they say what the sequence looks capable of, +not what a given cell is doing. The matching experimental data is in the +<a href="hgTrackUi?g=encode4ProCap">PRO-cap</a> track. +</p> + +<h2>Display Conventions and Configuration</h2> + +<p> +Cell lines are listed in the table on this page, one row each, with a checkbox +per data type. Use the Sample class facet to narrow the list, and the Group +tracks by buttons to order the browser by sample or by data type. +</p> + +<p> +Each predicted PRO-cap track is an overlay of the two strands: plus strand +predictions are drawn upward and minus strand predictions downward. The y axis +is the predicted number of PRO-cap reads at that base. +</p> + +<p> +Contribution scores are drawn as a sequence logo when zoomed in far enough to +show individual bases: the letter of the reference base is scaled by its score, +so a run of tall letters is a motif the model relied on. At lower zoom the same +values are drawn as a wiggle. Scores can be negative, meaning the base argued +against initiation being placed where it was. Scores exist only in the roughly +2 kb window around each MANE Select transcription start site, about 38.7 Mb of +the genome. Everywhere else the track is empty, which is not the same as a score +of zero. +</p> + +<p> +Read depth differs between the six PRO-cap experiments the models were trained +on, and both predicted signal and contribution scores scale with it, so the +y axis is not comparable between cell lines. Tracks are colored by the cell line +the model was trained on: +</p> +<ul> +<li><span style="display:inline-block; background-color:#0072B2; width:18px; height:12px; vertical-align:middle;"></span> <b>A673</b> Ewing sarcoma</li> +<li><span style="display:inline-block; background-color:#D55E00; width:18px; height:12px; vertical-align:middle;"></span> <b>Caco-2</b> colorectal adenocarcinoma</li> +<li><span style="display:inline-block; background-color:#009E73; width:18px; height:12px; vertical-align:middle;"></span> <b>Calu3</b> lung adenocarcinoma</li> +<li><span style="display:inline-block; background-color:#CC79A7; width:18px; height:12px; vertical-align:middle;"></span> <b>HUVEC</b> umbilical vein endothelial cells</li> +<li><span style="display:inline-block; background-color:#E69F00; width:18px; height:12px; vertical-align:middle;"></span> <b>K562</b> chronic myelogenous leukemia</li> +<li><span style="display:inline-block; background-color:#56B4E9; width:18px; height:12px; vertical-align:middle;"></span> <b>MCF10A</b> non-tumorigenic breast epithelium</li> +</ul> + +<h2>Methods</h2> + +<h3>Model</h3> + +<p> +ProCapNet adapts the BPNet architecture. It reads 2,114 bp of one-hot encoded +sequence and outputs both a base-resolution profile over the central 1,000 bp, +covering both strands as a single softmax so the model can learn strand +asymmetry, and a scalar giving the log total number of initiation events in that +window. Training used ENCODE PRO-cap read alignments in which the first read of +each pair was discarded and only the single 5'-most base of the second read was +kept, merged across replicates and kept separate by strand. Each model was +trained on all PRO-cap peaks in its cell line plus randomly sampled +DNase-hypersensitive sites from the same cell line at a 7:1 peak to background +ratio, with 7-fold cross-validation split by chromosome. Bases that are not +uniquely mappable by 36-mer reads were given zero loss weight during training. +Full details are in Cochran <em>et al</em>. +</p> + +<h3>Predicted PRO-cap</h3> + +<p> +Genome-wide predictions were generated by the Kundaje lab by applying the model +to every 2,114 bp window at a stride of 250 bp, so each base is the average of +four overlapping predictions, then averaging across the seven cross-validation +models and across the forward and reverse-complemented sequence. On hg38 no +prediction was made where most of a window was unresolved (N) in the reference. +</p> + +<h3>Sequence contribution scores</h3> + +<p> +Scores were computed with DeepSHAP, which estimates each base's contribution by +contrasting the model's output on the real sequence against its output on a set +of reference sequences, here 25 dinucleotide shuffles of the sequence being +scored. Because DeepSHAP needs a single scalar to explain, the base-resolution +profile output was summarized by mean-normalizing the pre-softmax logits and +taking their dot product with the post-softmax profile, which weights each +base's logit by its predicted probability of being used and sums over the +1,000 bp output window and both strands. This is the profile or TSS-positioning +task; ProCapNet can also produce scores for its read-count task, which are not +shown here. Each scored sequence was run through all seven cross-validation +models and in both orientations, and the scores averaged. +</p> + +<h3>Processing at UCSC</h3> + +<p> +The bigWig files were downloaded from +<a href="https://mitra.stanford.edu/kundaje/kcochran/ucsc_track_hubs_public/ProCapNet/" +target="_blank">the Kundaje lab server</a>. The contribution score files are used +unaltered. In the prediction files the per-base values were re-encoded from one +interval per base into fixedStep sections, which cuts the file size by about a +third and changes no value. Bases carrying a literal NaN in the published hg38 +files, all of them in unresolved (N) reference sequence, were dropped so that the +track autoscales and summarizes correctly; this removed 164,268,582 of the +3,088,269,832 bases on the primary hg38 chromosomes, leaving 2,924,001,250 bases +with a prediction. The hs1 files had no such bases and all 3,117,275,501 were +kept. Nothing else was altered: every retained value is bit for bit the published +value. The steps are recorded in the +<a href="https://github.com/ucscGenomeBrowser/kent/blob/master/src/hg/makeDb/doc/hg38/transcriptionStart.txt" +target="_blank">makeDoc</a> and the scripts are in +<a href="https://github.com/ucscGenomeBrowser/kent/tree/master/src/hg/makeDb/outside/proCapNet" +target="_blank">the kent source tree</a>. +</p> + +<h2>Data Access</h2> + +<p> +The data can be explored interactively in table format with the +<a href="../cgi-bin/hgTables">Table Browser</a> or the +<a href="../cgi-bin/hgIntegrator">Data Integrator</a> and exported from there to +spreadsheet or tab-sep tables. From scripts, the data can be accessed through our +<a href="https://api.genome.ucsc.edu" target="_blank">API</a>, track=<i>proCapNet</i>. +</p> + +<p> +The Files column of the table on this page links each cell line's bigWigs +directly, so a single file can be fetched without working out its path. +</p> + +<p> +For automated download and analysis, the genome annotation is stored in bigWig +files that can be downloaded from +<a href="http://hgdownload.soe.ucsc.edu/gbdb/hg38/proCapNet/" target="_blank">our +download server</a>. Predictions are under <tt>pred/</tt> and are named for the +cell line, the model and the strand, for example +<tt>K562.proCapNet.pos.bw</tt> and <tt>K562.proCapNet.neg.bw</tt>. Contribution +scores are under <tt>contrib/</tt>, for example +<tt>K562.proCapNet-contrib.bw</tt>. The T2T-CHM13 predictions are in the same +layout under <tt>/gbdb/hs1/proCapNet/pred/</tt>; there are no contribution scores +for that assembly. Individual regions or the whole genome annotation +can be obtained using our tool <tt>bigWigToBedGraph</tt>, which can be compiled +from the source code or downloaded as a precompiled binary for your system. +Instructions for downloading source code and binaries can be found +<a href="http://hgdownload.soe.ucsc.edu/downloads.html#utilities_downloads" target="_blank">here</a>. +The tool can also be used to obtain features within a given range, e.g. +<tt>bigWigToBedGraph http://hgdownload.soe.ucsc.edu/gbdb/hg38/proCapNet/pred/K562.proCapNet.pos.bw +-chrom=chr21 -start=0 -end=100000000 stdout</tt> +</p> + +<p> +The original files can be downloaded from +<a href="https://mitra.stanford.edu/kundaje/kcochran/ucsc_track_hubs_public/ProCapNet/" +target="_blank">the Kundaje lab server</a>. The trained models are on the ENCODE +portal, linked from the Experiment column of the table on this page. +</p> + +<h2>Credits</h2> + +<p> +ProCapNet was developed by Kelly Cochran in the Kundaje lab at Stanford +University. The genome-wide predictions and contribution scores were generated by +Kelly Cochran in collaboration with the GENCODE consortium. Thanks to Kelly +Cochran and Anshul Kundaje for making the data available. +</p> + +<h2>References</h2> + + +<p> +Cochran K, Yin M, Mantripragada A, Schreiber J, Marinov GK, Shah SR, Yu H, Lis JT, Kundaje A. +<a href="https://www.ncbi.nlm.nih.gov/pubmed/38853896" target="_blank"> +Dissecting the cis-regulatory syntax of transcription initiation with deep learning</a>. +<em>bioRxiv</em>. 2024 Nov 21;. +PMID: <a href="https://www.ncbi.nlm.nih.gov/pubmed/38853896" target="_blank">38853896</a>; PMC: <a +href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11160661/" target="_blank">PMC11160661</a> +</p> + + + +<p> +Avsec Ž, Weilert M, Shrikumar A, Krueger S, Alexandari A, Dalal K, Fropf R, McAnany C, Gagneur J, +Kundaje A <em>et al</em>. +<a href="https://www.ncbi.nlm.nih.gov/pubmed/33603233" target="_blank"> +Base-resolution models of transcription-factor binding reveal soft motif syntax</a>. +<em>Nat Genet</em>. 2021 Mar;53(3):354-366. +PMID: <a href="https://www.ncbi.nlm.nih.gov/pubmed/33603233" target="_blank">33603233</a>; PMC: <a +href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8812996/" target="_blank">PMC8812996</a> +</p> + + + +<p> +Kwak H, Fuda NJ, Core LJ, Lis JT. +<a href="https://www.ncbi.nlm.nih.gov/pubmed/23430654" target="_blank"> +Precise maps of RNA polymerase reveal how promoters direct initiation and pausing</a>. +<em>Science</em>. 2013 Feb 22;339(6122):950-3. +PMID: <a href="https://www.ncbi.nlm.nih.gov/pubmed/23430654" target="_blank">23430654</a>; PMC: <a +href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3974810/" target="_blank">PMC3974810</a> +</p> +