Eytan Ruppin: We democratize precision oncology through accurate response prediction

Eytan Ruppin, Chief of the Cancer Data Science Laboratory, shared on X:

“Can we democratize precision oncology through accurate response prediction based solely on tumor H&E images?

Our ENLIGHT-DeepPT framework can help make it happen! We are ecstatic to share our latest work in digital pathology, just out in Nature Cancer!

While Next Generation Sequencing (NGS) testing for cancer patients is becoming common in the US and Europe, most patients in low- to middle-income countries do not yet enjoy its benefits and consequently those of Precision Medicine.

Furthermore, NGS often takes 4-6 weeks to return a result. Many patients with advanced cancers require immediate treatment, and Digital Pathology can potentially offer Cancer Treatment options within a much shorter time frame.

To address this challenge, we developed a two-step indirect prediction approach:

First, we apply a novel deep learning method called DeepPT that infers Gene Expression from standard H&E-stained histopathology slides.

Subsequently, these inferred gene expression values are utilized for predicting cancer therapy response using our previously developed ENLIGHT algorithm.

As a starting point, DeepPT infers gene expression from H&E slides with markedly improved performance compared to published tools.

It predicts thousands of genes in each cancer type with significant positive correlations!

DeepPT models trained on TCGA samples were successfully applied to independent tissue sample datasets of breast cancer (TransNEO, Sammut et al.) and brain cancer (NCI) [1,753 and 1,408 genes with pearson R > 0.4, respectively].”

Source: Eytan Ruppin/X