eCAMI

eCAMI leverages an amino acid k-mer-based approach and a bipartite network algorithm to classify carbohydrate-active enzymes (CAZymes), identify conserved peptide motifs, and enable enzyme family/subfamily annotation and EC number prediction for Swiss-Prot enzymes.

Key Features:

• Amino Acid k-Mer-Based Classification: eCAMI uses short amino acid k-mers to classify enzymes into families and subfamilies and to identify distinguishing peptides.
• Motif Identification: The tool detects conserved short peptide sequences (k-mers) that serve as characteristic motifs for enzyme subfamilies.
• Bipartite Network Algorithm: A bipartite network algorithm maps relationships between enzymes and their k-mer peptides to support classification and motif discovery.
• Generalization to Swiss-Prot and EC Prediction: The methodology has been generalized to Swiss-Prot enzymes classified by EC numbers to facilitate enzyme commission (EC) number prediction.
• Benchmarking and Performance: eCAMI was benchmarked against homology-based and deep-learning methods and demonstrated superior accuracy and memory efficiency for CAZyme and EC classification and annotation.
• Implementation: The method is implemented as a Python package for integration into bioinformatics workflows.

Scientific Applications:

• Bioenergy Research: Classification and annotation of CAZymes to identify enzymatic pathways relevant to biomass degradation and bioenergy production.
• Human Gut Microbiome Studies: Detailed enzyme annotations to elucidate microbial enzyme functions within the human gut microbiome.
• Plant Pathogen Research: Identification and classification of plant pathogenic enzymes to support studies of disease mechanisms and control strategies.

Methodology:

eCAMI employs an amino acid k-mer-based approach and identification of conserved k-mer peptides, uses a bipartite network algorithm to map enzyme–k-mer relationships, has been generalized to Swiss-Prot EC-classified enzymes, is implemented as a Python package, and was benchmarked against homology-based and deep-learning tools showing superior accuracy and memory efficiency.