MultiMAP: dimensionality reduction and integration of multimodal data
Mika Sarkin Jain, Krzysztof Polański, Cecilia Domínguez Conde, Xi Chen, Jong-Eun Park, Lira Mamanova, Andrew Knights, Rachel A. Botting, Emily Stephenson, Muzlifah Haniffa, Austen Lamacraft, Mirjana Efremova, Sarah A. Teichmann
Abstract
Multimodal data is rapidly growing in many fields of science and engineering, including single-cell biology. We introduce MultiMAP, a novel algorithm for dimensionality reduction and integration. MultiMAP can integrate any number of datasets, leverages features not present in all datasets, is not restricted to a linear mapping, allows the user to specify the influence of each dataset, and is extremely scalable to large datasets. We apply MultiMAP to single-cell transcriptomics, chromatin accessibility, methylation, and spatial data and show that it outperforms current approaches. On a new thymus dataset, we use MultiMAP to integrate cells along a temporal trajectory. This enables quantitative comparison of transcription factor expression and binding site accessibility over the course of T cell differentiation, revealing patterns of expression versus binding site opening kinetics.