ENCODE preprint: neural nets predict super-additive GATA motif epistasis (r=0.55)
anshulkundaje · x · 2026-11-19
The ENCODE Project Consortium posted a new preprint of "The Encyclopedia of DNA Elements" on bioRxiv; Anshul Kundaje's team discusses neural-network validation details.
- MPRA library of 697 genomic loci with two GATA motifs at variable spacing, each assayed as wild-type plus combinatorial motif-scramble variants.
- BPNet (TAL1/GATA1/GATA2), ChromBPNet (DNase/ATAC) and an MPRA variant model predict first-order activity with 0.5–0.6 correlation to measured MPRA.
- Models capture super-additive epistasis: scrambling one GATA motif reduces the apparent contribution of its intact partner; predicted DNase epistasis correlates r=0.55 with MPRA measurements.
- Controlled analyses are being replicated at scale across the full compendium; small super-additive effects on binding can translate to large effects on accessibility and expression.
More from Research
- New TLAPS-Bench blog: formally proving correctness of real-world system specs with TLA+ — tianyin_xu · 2026-11-19
- Astra's proof sidesteps three deep Mathlib results by inventing a new 'capacity' quantity — IgorCarron · 2026-11-19
- Stanford's Kundaje: ChIP-seq resolution limits motif-level deep learning analysis — anshulkundaje · 2026-11-19
- Mahina 4.0 LoRA Adapters Claim Fewer Hallucinations, Devs Seek Stress-Tests — External-Fee-8920 · 2026-11-19
- The biggest advantage of a top lab isn't money — it's saving you years — madeofAjala · 2026-11-19
- LEV: Looped decision model lifts 2.6B accuracy 63.2% to 76.2% with adaptive compute — jiqizhixin · 2026-11-19