RT Journal Article
SR Electronic
T1 Transcriptomic learning for digital pathology
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 760173
DO 10.1101/760173
A1 Benoît Schmauch
A1 Alberto Romagnoni
A1 Elodie Pronier
A1 Charlie Saillard
A1 Pascale Maillé
A1 Julien Calderaro
A1 Meriem Sefta
A1 Sylvain Toldo
A1 Thomas Clozel
A1 Matahi Moarii
A1 Pierre Courtiol
A1 Gilles Wainrib
YR 2019
UL http://biorxiv.org/content/early/2019/09/11/760173.abstract
AB Deep learning methods for digital pathology analysis have proved an effective way to address multiple clinical questions, from diagnosis to prognosis and even to prediction of treatment outcomes. They have also recently been used to predict gene mutations from pathology images, but no comprehensive evaluation of their potential for extracting molecular features from histology slides has yet been performed. We propose a novel approach based on the integration of multiple data modes, and show that our deep learning model, HE2RNA, can be trained to systematically predict RNA-Seq profiles from whole-slide images alone, without the need for expert annotation. HE2RNA is interpretable by design, opening up new opportunities for virtual staining. In fact, it provides virtual spatialization of gene expression, as validated by double-staining on an independent dataset. Moreover, the transcriptomic representation learned by HE2RNA can be transferred to improve predictive performance for other tasks, particularly for small datasets. As an example of a task with direct clinical impact, we studied the prediction of microsatellite instability from hematoxylin & eosin stained images and our results show that better performance can be achieved in this setting.