Predictive Modelling of Multimodal Single Cell Genomic Data with Machine Learning Algorithms
摘要
This paper demonstrates the use of machine learning to model the multimodal nature of a single cell. Using machine learning to predict RNA from DNA, that is, using chromatin accessibility data to predict the RNA gene expression and to predict surface protein from RNA, that is, using RNA sequence data to predict surface protein levels in a single cell. Predicting the proportions of DNA, RNA, and surface proteins in a cell can provide valuable insights into the molecular composition and characteristics of different cell types, which could further help in building the Human Cell Atlas (HCA). This information on proportion also helps in identifying specific cell markers and signatures that define different cell populations within various tissues and organs. In this paper, we have summarized how machine learning can be helpful in understanding the modalities of a single cell and have created a pipeline for the same.