Niche-Based Microbial Community Assemblage in Urban Transit Systems and the Influence of City Characteristics

Microbiota residing on the urban transit systems (UTSs) can be shared by travelers and have niche-specific assemblage. However, it remains unclear how the assemblages are influenced by city characteristics, rendering city-specific and microbial-aware urban planning challenging. Here, we analyzed 3,359 UTS microbial samples collected from 16 cities around the world. We found the stochastic process dominated in all UTS microbiota assemblages, with the explanation rate (R-2) of the neutral community model (NCM) higher than 0.7. Moreover, city characteristics predominantly drove such assemblage, largely responsible for the variation in the stochasticity ratio (50.1%). Furthermore, by utilizing an artificial intelligence model, we quantified the ability of UTS microbes in discriminating between cities and found that the ability was also strongly affected by city characteristics, especially climate and continent. From these, we found that although the NCM R-2 of the New York City UTS microbiota was 0.831, the accuracy of the microbial-based city characteristic classifier was higher than 0.9. This is the first study to demonstrate the effects of city characteristics on the UTS microbiota assemblage, paving the way for city-specific and microbial-aware applications.IMPORTANCE We analyzed the urban transit system microbiota assemblage across 16 cities. The stochastic process was dominant in the urban transit system microbiota assemblage. The urban transit system microbe's ability in discriminating between cities was quantified using transfer learning based on random forest (RF) methods. Certain urban transit system microbes were strongly affected by city characteristics. We analyzed the urban transit system microbiota assemblage across 16 cities. The stochastic process was dominant in the urban transit system microbiota assemblage. The urban transit system microbe's ability in discriminating between cities was quantified using transfer learning based on random forest (RF) methods. Certain urban transit system microbes were strongly affected by city characteristics.