Unraveling city-specific signature and identifying sample origin locations for the data from CAMDA MetaSUB challenge

Abstract Background Composition of microbial communities can be location-specific, and the different abundance taxon within location could help us to unravel city-specific signature predict sample origin locations accurately. In this study, whole genome shotgun (WGS) metagenomics data from samples across 16 cities around world another 8 were provided as main mystery datasets respectively part CAMDA 2019 MetaSUB “Forensic Challenge”. The feature selecting, normalization, three methods machine learning, PCoA (Principal Coordinates Analysis) ANCOM (Analysis composition microbiomes) conducted for both datasets. Results Features combined with machines learning methods, revealed that combination common features was effective predicting samples. average error rates 11.93 30.37% obtained respectively. Using dataset labels dataset, nearly 89.98% test correctly labeled “mystery” showed 60% total variability explained by first two axes. Although many overlapped, separation some found in PCoA. results ANCOM, importance score Random Forest, indicated “family”, “order” main-dataset most efficient information prediction Conclusions classification suggested microbiomes distinctive cities, which used identify origins. This also supported RF. addition, accuracy improved more better sequencing depth.