Find in Library

Summary: Expectations of machine learning (ML) are high for discovering new patterns in high-throughput biological data, but most such practices are accustomed to relying on existing knowledge conditions to design experiments. Investigations of the power and limitation of ML in revealing complex patterns from data without the guide of existing knowledge have been lacking. In this study, we conducted systematic experiments on such ab initio knowledge discovery with ML methods on single-cell RNA-sequencing data of early embryonic development. Results showed that a strategy combining unsupervised and supervised ML can reveal major cell lineages with minimum involvement of prior knowledge or manual intervention, and the ab initio mining enabled a new discovery of human early embryonic cell differentiation. The study illustrated the feasibility, significance, and limitation of ab initio ML knowledge discovery on complex biological problems. The Bigger Picture: Machine learning (ML) has been shown to be powerful in many artificial intelligence tasks, so people expect it to be able to reveal patterns that even human experts may have difficulty discovering. Scientists are enthusiastic in using ML to analyze the complex biology underlying various single-cell genomics data, but most existing studies of this type are accustomed to relying on existing knowledge to design experiments. Such practices may miss important discoveries and leave the question open as to how far ML and data may go beyond the sphere of existing knowledge.This study uses the example of cell lineages in early embryonic development to investigate the feasibility of machine-learning discovery of biological knowledge from data with minimum use of prior knowledge. We call the tasks ab initio knowledge discovery. The strategy and observations can act as a baseline for future efforts of discovering new knowledge from single-cell genomics data.