Gene expression equivalent of polygenic risk score ?

Question

[Picasa]

Hi

In clinical bioinformatics, polygenic scores are commonly used to assess an individual’s risk of disease by aggregating information from many variants.

Is there something similar that can be done using gene expression data instead of variants?

For example, a standardized way to combine expression values into a risk or diagnostic score that can be applied to individual patients ?

Answer 1

[GouthamAtla]

I don't think something exists out of the box. With variants, we have a 'weight' for each variant on disease risk (obtained through large scale GWAS analysis). To do that with gene expression, one should train a model on extremely large cohort with gene expression data, obtain weights for each gene for a disease . Then to apply this, the target cohort should have measured gene expression using exactly same technology (tissues, platform etc etc) and use the same normalisation approaches, so that the predicted weights can be used to compute disease risk. Something that might be easier is to work with ratios of gene expression values (known disease marker genes) or a gene set signature (relative levels of group of genes) to mitigate issues with platforms or analysis methods.

This is a nice paper on proteomics risk scores, which illustrates on how to approach the problem. https://www.nature.com/articles/s41591-024-03142-z