Get Started
Home
Topics
Search
Library
Research questionDo gains from fine-tuned genomic variant-effect prediction transfer to zero-shot functional sequence inpainting?Genomic models can score variants after supervised fine-tuning yet fail to reconstruct biologically functional sequence in zero-shot settings. This leaves unclear whether a masking prior that helps one objective captures the constraints needed for the other.
AI
Diffusion Models
Evaluation & Benchmarks
Machine Learning
Research Paper
Latest papersRecent research connected to this question, newest first.When Genomic Masking Priors Fail to Transfer: Strong Variant Prediction, Weak Functional GenerationThe evidence concerns a 202M-parameter bidirectional discrete diffusion model using entropy-guided span masking, evaluated on ClinVar single-nucleotide variant prediction and functional inpainting in promoters, enhancers, exon boundaries, and intron boundaries. Results compare entropy-guided spans with random-span and composition-preserving controls for 50–500 bp gaps; the findings do not establish that entropy guidance improves either objective or that variant-prediction gains transfer to functional generation.research paper · Sep 4, 2026
Related questions
How can protein language models predict mutation effects accurately without the cost of larger dense models?How can neural fields reconstruct missing signal regions from sparse observations while retaining reusable task priors?How can self-supervised representations transfer reliably across microscopy datasets with scarce labels and mismatched staining or channels?Can reducing the complexity of a linear generative prior improve expected reconstruction error in noiseless Gaussian compressed sensing?