CaMELS

CaMELS predicts calmodulin (CaM) binding proteins and their binding sites from amino acid sequence to identify CaM–protein interactions and their loci.

Key Features:

• Algorithm Suite: Uses machine learning approaches including a large-margin classifier and multiple-instance learning with a custom optimization algorithm.
• CaM Interaction Prediction: Predicts CaM–protein interactions from protein sequence features using a large-margin classifier.
• Binding Site Prediction: Identifies CaM-binding sites using multiple-instance learning combined with a custom optimization algorithm to handle imprecise site annotations.
• Benchmarking and Validation: Validated using mutagenic studies, proteome-wide Gene Ontology enrichment analyses, and protein structures, and shown to outperform motif-based search methods for interaction and binding-site prediction.
• Sequence Importance: Considers the entire protein sequence rather than only putative binding motifs when predicting CaM interactions.
• Feature Identification: Extracts sequence-derived features and characteristic amino acid subsequences indicative of CaM interaction and binding-site localization.

Scientific Applications:

• Calmodulin-binding protein discovery: Supports identification and prioritization of calmodulin-binding proteins across proteomes for experimental validation.
• Binding site mapping: Provides putative binding-site loci to guide mutagenesis and structural studies.
• Experimental design and prioritization: Aids design and optimization of wet-lab experiments by prioritizing candidate proteins and sites for validation.

Methodology:

CaMELS extracts sequence features, applies a large-margin classifier for interaction prediction, and employs multiple-instance learning with a custom optimization algorithm for binding-site identification.