Loading...
Loading...
Contributors:
Dataset URI:
10.5281/zenodo.22764023Data and code accompanying the manuscript "A sequence-to-function model predicts evolutionary constraint in human promoters beyond motif occupancy" by Düzgün Z, Yüzüak H and Demirtaş Korkmaz F. The study asks whether the per-base regulatory sensitivity assigned by AlphaGenome predicts evolutionary constraint measured from a 100-way vertebrate alignment, after controlling for sequence composition, distance to the transcription start site, and a deliberately strong transcription factor motif baseline. Sensitivity was quantified from precomputed saturation mutagenesis in the AlphaGenome Atlas over 4 kb windows centred on the transcription start sites of 144 literature-curated unfolded protein response promoters and 299 randomly sampled protein-coding promoters, in two cell contexts and for five scorers, with the gene as the unit of analysis. Contents. 31 analysis scripts in run order; the gene panels; per-position derived tables for AlphaGenome sensitivity, phyloP and phastCons conservation, gnomAD v4.1 variation and JASPAR motif occupancy; 22 result files, one per analysis; the five final figures as 400 dpi PNG and vector PDF; and documentation, including a map from every value quoted in the paper to the file it comes from, a provenance note listing the pitfalls encountered and how each was caught, and a record of the end-to-end reproduction test. Reproducibility. From a clean working directory built only from this archive, with no AlphaGenome API key and no network access, all 22 result files regenerate with a largest numeric difference of zero and all five figures regenerate byte-identical. The run takes about fourteen minutes. Four upstream collection steps cannot be reproduced from the archive alone because they retrieve external data (GENCODE v46 gene models, the AlphaGenome API, the UCSC REST API and gnomAD v4.1); their outputs are included here. See docs/reproduction.md. Licensing, please read before reuse. The code, the documentation and the non-AlphaGenome content are released under CC BY 4.0. The tables under data/derived/alphagenome/, and every result and figure computed from them, are Output Derivatives of AlphaGenome and are additionally subject to the AlphaGenome Output Terms of Use, which permit non-commercial use only and prohibit use for training variant-effect models. A copy of those terms is included at the root of the archive as LEGALLY_BINDING_TERMS_OF_USE.txt. JASPAR matrices are not redistributed; the archive instead lists the 120 matrix identifiers used, so that the scan can be rebuilt with pyjaspar. Funding. Scientific Research Projects Coordination Unit of Giresun University, project SAĞ-BAP-A-250626-64.
Keyword(s):
AlphaGenome; deep learning; evolutionary constraint; JASPAR; phastCons.
MeSH Term(s):
Grants and Funding:
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
No Supplementary Data Available
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)