AllTheBacteria: a community resource empowers biology and discovers novel peptide antibiotics

Technical Summary of "AllTheBacteria: a community resource empowers biology and discovers novel peptide antibiotics"

Problem Statement
Public microbial genomes represent an immense record of biological diversity, evolution, and molecular function. However, the utility of this data is currently constrained by significant heterogeneity in its processing. Raw sequencing data from public repositories are not uniformly assembled, quality-controlled, annotated, or searchable at scale. This lack of standardization renders much of the available genomic information difficult to reuse for large-scale discovery applications.

Methodology
To address these limitations, the authors developed AllTheBacteria, an open, community-built resource designed to transform public bacterial short-read whole-genome sequencing reads into a uniformly processed discovery platform. The pipeline involves:

• Data Curation: Aggregating and processing raw reads into high-quality, uniformly assembled genomes.
• Standardized Annotation: Applying consistent taxonomic assignments, genome annotations, antimicrobial resistance (AMR) calls, and antiphage-defence annotations across the dataset.
• Structural and AI Integration: Generating protein structure predictions and creating AI-ready sequence tables.
• Experimental Validation: The resource was utilized to mine proteomes for novel antimicrobial peptides. Specifically, the authors employed their deep learning model, APEX 1.1, to analyze 3,919,096 encrypted peptide fragments derived from the AllTheBacteria proteomes.

Key Contributions
The paper presents the current release of AllTheBacteria, which contains:

• 2,440,377 high-quality bacterial and archaeal genomes.
• Representation of 11,273 distinct species.
• A comprehensive suite of standardized data layers, including taxonomy, annotations, resistance profiles, defence systems, and structural predictions.

Results
The study demonstrates the utility of this infrastructure through both computational and experimental validation:

1. Computational Discovery: Using the APEX 1.1 model on the AllTheBacteria dataset, the authors identified 1,867 peptide candidates with predicted antimicrobial activity.
2. Experimental Screening: Twenty-four representative peptides were synthesized and tested against 20 clinically relevant bacterial strains, including antibiotic-resistant pathogens.
3. Biological Characterization: Multiple peptides exhibited low-micromolar activity, membrane-responsive conformational transitions, and selective envelope perturbation.
4. In Vivo Efficacy: A lead molecule, ATB20, demonstrated efficacy in a murine skin abscess model infected with Acinetobacter baumannii. ATB20 reduced the bacterial burden with potency comparable to polymyxin B and showed no overt toxicity.

Significance
The paper establishes AllTheBacteria as a foundational community resource for microbiology and a renewable engine for AI-guided antimicrobial discovery. The authors assert that this infrastructure enables applications that were previously impractical, including:

• Global sequence search and outbreak contextualization.
• Pangenome method development.
• Mapping of antimicrobial resistance reservoirs.
• Exploration of antiphage-defence ecology.

By transforming fragmented public data into a unified, searchable, and AI-ready platform, AllTheBacteria facilitates the discovery of novel therapeutic agents, as evidenced by the identification and validation of the lead peptide ATB20.