Multi-proteins similarity-based sampling to select representative genomes from large databases.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genome databases are expanding rapidly, leading to high redundancy and variable data quality.
• Selecting representative genome subsets is crucial for genomic studies but current methods are often biased and slow for large datasets.

Purpose of the Study:

• To introduce MPS-Sampling (Multiple-Protein Similarity-based Sampling), a novel method for fast, scalable, and efficient genome sampling.
• To address the limitations of existing biased and time-consuming sampling approaches for large genomic datasets.

Main Methods:

• MPS-Sampling utilizes homologous protein families to group similar genomes.
• It employs a two-step clustering process to delineate homogeneous genome groups.
• Representative genomes are selected based on user-defined criteria within these groups.

Main Results:

• MPS-Sampling was applied to 178,203 bacterial genomes using 48 ribosomal protein families.
• It successfully generated representative genome sets, reducing dataset size significantly (0.3% to 32.17%).
• Selected genomes demonstrated strong taxonomic and phylogenetic representativeness of the complete dataset.

Conclusions:

• MPS-Sampling offers an efficient, fast, and scalable solution for sampling large genome collections within practical computational limits.
• The method avoids biases associated with taxonomic information and phylogenetic tree inference.
• MPS-Sampling meets the growing demand for reliable genome sampling tools in bioinformatics.