N. Hemachandra

This paper predicts fold-switching in proteins using handcrafted and language model features, finding binary classification outperforms 3-class and handcrafted features avoid sequence memorization.

The paper evaluates BLASTp and profile HMMs for fine-grained functional annotation, finding BLASTp struggles with remote homologs while HMMs improve sensitivity but may miss some family members due to functional divergence.

This study evaluates nine clustering algorithms for unsupervised protein function annotation using pseudo-amino acid composition features, finding some methods effective for functional family segregation but less so for fine-grained differences.

The paper uses sequence composition features to classify paralogous proteins and identifies key amino acids driving classification.