Potential Phage-derived hypothetical proteins in clinical Mycobacterium tuberculosis isolates (cluster-19).

Potential Phage-derived hypothetical proteins in clinical Mycobacterium tuberculosis isolates (cluster-19). This repository contains 971 hypothetical protein sequences identified through large-scale in silico deep mining of the Mycobacterium tuberculosis proteome, their UMAP embedding coordinates (YZ coordinates were used in the manuscript), and cluster annotations. Contents1) hypotheticals_cluster19_v1: FASTA sequences of 971 hypothetical protein sequences. FASTA headers match the CSV field `original_id` (e.g., ERR/SRR accessions with locus tags).2) UMAP_Kmeans_v1.csv: UMAP coordinates and k-means cluster labels for each sequence. original_id: sequence identifier; matches FASTA headers; sequence_length; UMAP coordinates; annotation. 3) cluster_annotation_v1: Mapping from cluster ID to brief functional annotation. Methodology and brief description of results. De novo assemblies of publicly available M. tuberculosis whole-genome sequencing datasets were annotated with BAKTA. The resulting hypothetical proteins were binned based on sequence length. The majority of bins showed sequence similarity with M. tuberculosis proteins. The bin of 100-300 amino acid-long hypothetical proteins, upon embedding, showed a sub-cluster of 971 proteins, whose medoids showed no sequence similarity with M. tuberculosis proteins. Subsequent analysis showed the presence of Phage-derived domains and foreign host-like sequences in this subcluster. This repository contains those 971 sequences. Version: v1.0 (2026-04-13)