Mapping biology from mouse to human

Using machine learning to understand human biology

Project status



Many animals have similar sets of genes, but these genes are expressed in very different ways. This project will use methods from transfer learning to map biology between species, and so better understand what makes humans unique.

Explaining the science

The basic idea of transfer learning is to mimic the human ability to learn new concepts from very few examples by associating new information with prior understanding. For example, we only need to see one image of a new mode of transport - a Segway for instance - to realise that, although different from something we have seen before, it has similarities to other forms of transport that we do know and thereby realise its function, and retain this knowledge for future reference. Transfer learning seeks to develop ways to formalise this idea in machine learning process. 

Project aims

Almost all cells in our bodies carry the same set of genes, but they express these genes in different ways. For example, bone cells are different from brain cells because they express different portions of the genome. The patterns of gene expression that characterise different cell types map partially, but not entirely, between the species. This project aims to use methods from transfer learning - a a set of machine learning methods that takes information learnt from one setting and uses it to improve understanding in another setting - to better understand how cell biology maps from one species to another. 


This project aims to help us better understand what makes us different to other species. This knowledge is of scientific interest and could be of practical use, for example in helping design better drugs. 


Professor Ben MacArthur

Director of AI for Science and Government, Deputy Programme Director for Health and Medical Sciences, and Turing Fellow