A pipeline for constructing reference genomes for large cohort-specific metagenome compression

Metagenomic data compression is very important as metagenomic projects are facing the challenges of larger data volumes per sample and more samples nowadays. The reference-based compression is a promising method to obtain a high compression ratio. However, existing microbial reference genome databases are not suitable to be directly used as references for compression due to their large size and redundancy, and different metagenomic cohorts often have various microbial compositions. We presented a novel pipeline that generated simplified and tailored reference genomes for large metagenomic cohorts, enabling reference-based compression for metagenomic data. We constructed customized reference genomes, ranging from 2.4 to 3.9GB, for 29 real metagenomic datasets, and evaluated their compression performance. Reference-based compression achieved an impressive compression ratio of over 20 for human whole-genome data and up to 33.8 for all samples, demonstrating a remarkably 4.5 times improvement than the standard Gzip compression. Our method provides new insights into reference-based metagenomic data compression and has a broad application potential for faster and cheaper data transfer, storage, and analysis.