FeatureDCA Enables Controllable Protein Sequence Generation
Caredda et al. (published Feb 19, 2026) introduce FeatureDCA, an autoregressive extension of Direct Coupling Analysis that conditions sequence generation on principal components derived from MSAs. The model matches or surpasses unconditioned Potts and autoregressive baselines in reproducing higher-order sequence statistics, preserves sequence diversity, and yields structures consistent with targets using AlphaFold and ESMFold. FeatureDCA enables interpretable, targeted sampling toward functional and structural subtypes for protein design.
Key Points
- 1Introduces FeatureDCA, an autoregressive DCA conditioned on principal components for guided sequence generation
- 2Demonstrates matched or improved generative accuracy and structural realism validated by AlphaFold and ESMFold
- 3Enables targeted sampling toward subtypes, preserving diversity and interpretability for protein design applications
Scoring Rationale
High novelty and peer-reviewed validation across families, supported by code release but scope remains focused on protein design.
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
