A deep learning solution to predict tumor mutational status : an evaluation as a first-line diagnostic tool

Abstract

The presence of genomic mutations in cancer can be associated with a response to a targeted therapy. Therefore, it has become a crucial information for giving more efficient treatments to every patient. Detection of mutation is routinely made by DNA-sequencing diagnostic tests. Recent developments showed promising results for tumoral mutational status prediction using new deep learning based methods on histopathological images. However, it is still unknown whether these methods can be useful aside from sequencing methods for efficient population diagnosis. Here, we use a standard prediction pipeline for the detection of clinically relevant genomic alterations in breast, lung and colon cancer. We propose 3 diagnostic strategies using deep learning methods as first-line diagnostic tools. We show that these methods help reduce DNA sequencing by up to 34.6% with a high sensitivity (95%). In a context of limited resources, these methods increase sensitivity up to 75% at a 30% capacity of DNA sequencing tests, up to 85.7% at a 50% capacity, and up to 92.3% at a 70% capacity. These methods can also be used to prioritize patients with a positive predictive value up to 86.7% in the 10% patient most at risk of being mutated.