What is it about?
Single cells behave in complex and sometimes seemingly random ways. We have applied Generative Adversarial Networks (GANs), a new type of artificial intelligence, to make sense of the way genes are controlled to make up all the different types of in skin cells in our bodies. A GAN involves two separate neural networks, a generator which simulates the behaviour of cells, and a discriminator that rates the quality of the simulation. The two networks compete against each other - with the generator trying to trick the discriminator into thinking that it’s seeing real cells - and quickly improve at their tasks. Looking at how the networks achieve their tasks provide new insights into the way cells behave.
Featured Image
Why is it important?
With advanced experimental techniques, we are now able to measure which genes are switched on in individual cells; but it is still hard to understand how and when these events happen. Without this knowledge, it is difficult to understand how cells differentiate into different cell types and how these processes can go wrong in diseases. Our new approach allows us to understand the relationship between different genes and how this contributes to cell behaviour.
Perspectives
Last year, over 400 new single-cell datasets were released and the pace of data production is ever-increasing. In this study we have shown that we can get even more out of this kind of data. Neural networks are usually thought of as ‘black boxes’, because they can be difficult to interpret. We have shown that this is not always the case, and that GANs can be used to improve our current understanding of the way cells behave in the skin.
Arsham Ghahramani
Francis Crick Institute
Last year over 400 new single cell datasets were released and the pace of data production is ever-increasing, but there are still many basic challenges to looking at these datasets. For instance, it is still difficult to combine data from multiple laboratories produced under different conditions. Here we have shown that we can get even more out of existing data; our GANs are able to integrate three distinct datasets without any prior normalisation. Most importantly, neural networks are usually considered to be black boxes that are difficult to interpret; hopefully we have shown that this is not always the case and that GANs can be used to improve our current understanding of the way cells behave in the skin.
Professor Nicholas Luscombe
Francis Crick Institute
Read the Original
This page is a summary of: Generative adversarial networks uncover epidermal regulators and predict single cell perturbations, February 2018, Cold Spring Harbor Laboratory Press,
DOI: 10.1101/262501.
You can read the full text:
Resources
Contributors
The following have contributed to this page
