UPEC Genomic Database & Machine Learning Dataset
🧬 Overview
This repository provides a curated genomic dataset and high-dimensional machine learning feature matrices for 1,184 human-specific Uropathogenic Escherichia coli (UPEC) isolates. By integrating pangenomic profiling with supervised machine learning, this framework enables the identification of lineage-specific genomic signatures, mobile genetic architecture, and metabolic markers driving UPEC pathoadaptation.
🔗 Access the Interactive Database https://zaaraishaq.github.io/UPEC-ML-data/
📂 Dataset Contents 1. Metadata (UPEC_Metadata_Complete.csv) A detailed epidemiological and genomic metadata file containing: Identifiers: BioProject, BioSample, and Assembly Accessions. Clinical Data: Host disease pathology (UTI, Cystitis, Pyelonephritis). Epidemiology: Collection year and geographical location. Assembly Stats: Genome size, GC content, and scaffold counts.
2. Machine Learning Matrix (Available in Repository) File: Statistically_Significant_Dataset.csv Format: Comprehensive binary matrix (0/1) tailored for machine learning. Dimensions: 838 isolates × 12,012 statistically significant genomic features.
3. Genome Assemblies Genomes were retrieved from NCBI. To access the raw sequence data for any isolate in this study, please refer to the Accession IDs (BioProject/Assembly) provided in the UPEC_Metadata_Complete.csv file.
🔬 Methodology
The dataset was constructed using a standardized bioinformatics pipeline: Data Retrieval: Public genomes were retrieved from NCBI on July 25, 2024. Annotation: Genomes were annotated using Prokka (v1.14.6). Pangenome Analysis: Core and accessory genes were defined using Roary (v3.13.0). MLST: Sequence Types determined via the Achtman scheme. Serotyping: Predicted using ecOH. Resistome: ResFinder, CARD, AMRFinderPlus and PointFinder. Virulome: VFDB & E. coli_VF databases. Mobilome: Predicted through PlasmidFinder, IntegronFinder, ISEScan and PHASTEST.
📚 Citation The manuscript regarding this database and dataset is under review.