Relation Therapeutics Unveils PEP-seq, Generating Training Data for DNA Foundation Models
anshulkundaje · x · 2026-10-31
Relation Therapeutics released a bioRxiv preprint introducing prime editing perturb-seq (PEP-seq), a protocol answering what data DNA foundation models need. Since Perturb-seq is limited in installable variants and prime editing in readouts, PEP-seq links programmed mutations to transcriptomic states. The team designed 2,164 prime editing guides targeting common eQTLs, premature termination codons and positional variants to scale variant-to-function training data. An accompanying figure outlines a three-tier data framework: base models on public reference data, fine-tuning on paired WGS+scRNA-seq personal genomes, and synthetic designed edits for counterfactual testing to improve OOD performance on VUS and rare disease interpretation.
More from Research
- Working Paper Argues AI Can Break the Coasian Ceiling on Firm Size — zetalyrae · 2026-10-31
- Molence Bio Open-Sources Mol-1, Matching Closed SOTA at Antibody Structure Prediction — AllThingsApx · 2026-10-31
- Revisiting why transformers can't do zero-shot recursion, via Transformer Circuits math — akbirthko · 2026-10-31
- AI-assisted proof: first high-probability regret bounds for Feel-good Thompson sampling — QuanquanGu · 2026-10-31
- Vitória Piai's free psycholinguistics book Speaking joins ChapterPal's open-access library with an AI tutor — burkov · 2026-10-31
- Neuron Lens: A database of every neuron and attention head in Qwen3-8B and Gemma 4 12B — CatAstro_Piyush · 2026-10-31