Kai Cheng Group

Computational metaproteomics turns mass spectrometry measurements into a functional readout of microbial activity. But microbiome-scale datasets are large, heterogeneous, and often difficult to analyse consistently across studies. Our group builds the software, statistical frameworks, and AI models needed to make metaproteomics scalable, reproducible, and biologically interpretable.

We develop and maintain open, user-friendly pipelines that connect raw MS data → peptide/protein evidence → taxa and function → cohort-level biological insights. A major focus is enabling robust analysis in both DDA and DIA settings, with strong quality control and transparent reporting so results can be trusted, shared, and reused.

Our work is highly collaborative. We partner with experimental and clinical teams to apply these methods to real questions—such as diet–microbiome interactions, xenobiotic responses, and functional shifts across cohorts—while also using public datasets to benchmark methods and discover general principles.

Long-term, we aim to move microbiome science from descriptive catalogues toward predictive, testable models. By combining harmonised data resources with learning-based inference, we want to help researchers generate better hypotheses, prioritise mechanisms, and design experiments more efficiently.