ProbioSML: Machine Learning-Derived Genomic Features from Probiotic Bacteria

1,072 non-redundant protein-coding sequences derived from comparative genomic analysis of bacteria frequently reported as probiotics. The dataset, named ProbioSML, was created by Diego Lucas Neres Rodrigues using pangenomic analysis and supervised machine learning models like Random Forest. It was last updated on April 22, 2026, and is available under a CC-BY-4.0 license.

Use Cases

• Exploratory analysis of genomic patterns associated with probiotic taxa based on the 1,072 discriminative features.
• Comparative genomics studies using the provided gene presence-absence matrices and functional annotations.
• Benchmarking machine learning models for feature extraction in microbial genomics based on the described Random Forest, SVM, and Logistic Regression approaches.

Strengths

• Dataset contains 1,072 non-redundant protein-coding sequences, providing a specific feature set.
• Includes gene presence-absence matrices and functional annotations, as described in the study.
• All data and analysis scripts are publicly available in an open-access repository.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific modeling tasks.
• The 782.9 KB file size indicates a relatively small-scale dataset.

Provenance

Source

figshare, with data available via Zenodo DOI 10.5281/zenodo.14181443.

Collection Method

Generated via pangenomic analysis combined with supervised machine learning (Random Forest, SVM, Logistic Regression) on bacterial genomes.

Freshness

Last updated 2026-04-22 05:46:20; freshness should be verified.