DARSI Predicts Gene Expression And Binding Sites
Karshenas, Röschinger and Garcia publish DARSI on April 1, 2026 in PLoS Computational Biology, introducing a convolutional neural network that predicts gene expression from raw regulatory DNA using MPRA training data. The model localizes transcription factor binding sites at single-base resolution, validates predictions against curated databases, and releases code and trained models on GitHub to enable experimental follow-up.
Key Points
- 1Introduces DARSI, a CNN trained on MPRA data to predict expression from raw DNA.
- 2Demonstrates single-base localization of transcription factor binding sites and benchmarks against curated databases.
- 3Enables experimentally actionable predictions and supplies code and models on GitHub for follow-up.
Scoring Rationale
Peer-reviewed PLoS Computational Biology paper introducing a novel CNN (DARSI) trained on MPRA datasets with validated, single-base binding-site localization and public code, boosting novelty, credibility, actionability, and relevance. Scope is focused on regulatory genomics rather than industry-wide impact, which lowered the scope score slightly; published today, so no freshness penalty applied.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems