Snekmer

Snekmer converts protein sequences into amino acid recoding (AAR) kmer vectors to improve sensitivity and coverage for assessing evolutionary and functional relationships among proteins.

Key Features:

• Amino Acid Recoding (AAR): Employs similarity groupings based on the chemical properties of amino acids to form reduced character sets that recode proteins into simplified representations.
• Kmer Vector Construction: Transforms AAR-recoded protein sequences into kmer vectors for downstream classification and clustering analyses.
• Dual Operational Modes: Supports supervised classification models trained on specific input protein families and clustering-based de novo family determination using unsupervised methods.
• Improved Detection of Distant Similarities: Links sequences with distant sequence similarity that are challenging to detect using exact-match kmer methodologies.
• Scalability and Implementation: Implemented in Python with Snakemake and demonstrated to scale to large protein sequence datasets.

Scientific Applications:

• Nitrogen cycling protein differentiation: Differentiates protein sequences within nitrogen cycling families for refined functional assignment.
• Uniprot-scale protein classification: Applies to broader protein sets sourced from Uniprot for classification and clustering at scale.
• Evolutionary and functional inference: Facilitates analysis of evolutionary relationships and functional genomics in microbial ecology and related fields.

Methodology:

Groups amino acids by chemical similarity to form reduced character sets (AAR), recodes protein sequences accordingly, constructs kmer vectors from the recoded sequences, and applies supervised classification or clustering (unsupervised learning) to those vectors.