A deep-learning framework to predict cancer treatment response from histopathology images through imputed transcriptomics

Danh Tai Hoang; Gal Dinstag; Eldad D. Shulman; Leandro C. Hermida; Doreen S. Ben-Zvi; Efrat Elis; Katherine Caley; Stephen John Sammut; Sanju Sinha; Neelam Sinha; Christopher H. Dampier; Chani Stossel; Tejas Patil; Arun Rajan; Wiem Lassoued; Julius Strauss; Shania Bailey; Clint Allen; Jason Redman; Tuvik Beker; Peng Jiang; Talia Golan; Scott Wilkinson; Adam G. Sowalsky; Sharon R. Pine; Carlos Caldas; James L. Gulley; Kenneth Aldape; Ranit Aharonov; Eric A. Stone; Eytan Ruppin

doi:10.1038/s43018-024-00793-2

A deep-learning framework to predict cancer treatment response from histopathology images through imputed transcriptomics

Danh Tai Hoang^*, Gal Dinstag, Eldad D. Shulman, Leandro C. Hermida, Doreen S. Ben-Zvi, Efrat Elis, Katherine Caley, Stephen John Sammut, Sanju Sinha, Neelam Sinha, Christopher H. Dampier, Chani Stossel, Tejas Patil, Arun Rajan, Wiem Lassoued, Julius Strauss, Shania Bailey, Clint Allen, Jason Redman, Tuvik BekerPeng Jiang, Talia Golan, Scott Wilkinson, Adam G. Sowalsky, Sharon R. Pine, Carlos Caldas, James L. Gulley, Kenneth Aldape, Ranit Aharonov, Eric A. Stone^*, Eytan Ruppin^*

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

1 Scopus citations

Abstract

Advances in artificial intelligence have paved the way for leveraging hematoxylin and eosin-stained tumor slides for precision oncology. We present ENLIGHT–DeepPT, an indirect two-step approach consisting of (1) DeepPT, a deep-learning framework that predicts genome-wide tumor mRNA expression from slides, and (2) ENLIGHT, which predicts response to targeted and immune therapies from the inferred expression values. We show that DeepPT successfully predicts transcriptomics in all 16 The Cancer Genome Atlas cohorts tested and generalizes well to two independent datasets. ENLIGHT–DeepPT successfully predicts true responders in five independent patient cohorts involving four different treatments spanning six cancer types, with an overall odds ratio of 2.28 and a 39.5% increased response rate among predicted responders versus the baseline rate. Notably, its prediction accuracy, obtained without any training on the treatment data, is comparable to that achieved by directly predicting the response from the images, which requires specific training on the treatment evaluation cohorts.

Original language	English
Journal	Nature Cancer
DOIs	https://doi.org/10.1038/s43018-024-00793-2
State	Accepted/In press - 2024

Funding

Funders	Funder number
Australian Research Council
National Institutes of Health
National Cancer Institute

Access to Document

10.1038/s43018-024-00793-2

Cite this

Hoang, D. T., Dinstag, G., Shulman, E. D., Hermida, L. C., Ben-Zvi, D. S., Elis, E., Caley, K., Sammut, S. J., Sinha, S., Sinha, N., Dampier, C. H., Stossel, C., Patil, T., Rajan, A., Lassoued, W., Strauss, J., Bailey, S., Allen, C., Redman, J., ... Ruppin, E. (Accepted/In press). A deep-learning framework to predict cancer treatment response from histopathology images through imputed transcriptomics. Nature Cancer. https://doi.org/10.1038/s43018-024-00793-2

@article{c19b2e4bb39a4371bed81d70cabb13cf,

title = "A deep-learning framework to predict cancer treatment response from histopathology images through imputed transcriptomics",

abstract = "Advances in artificial intelligence have paved the way for leveraging hematoxylin and eosin-stained tumor slides for precision oncology. We present ENLIGHT–DeepPT, an indirect two-step approach consisting of (1) DeepPT, a deep-learning framework that predicts genome-wide tumor mRNA expression from slides, and (2) ENLIGHT, which predicts response to targeted and immune therapies from the inferred expression values. We show that DeepPT successfully predicts transcriptomics in all 16 The Cancer Genome Atlas cohorts tested and generalizes well to two independent datasets. ENLIGHT–DeepPT successfully predicts true responders in five independent patient cohorts involving four different treatments spanning six cancer types, with an overall odds ratio of 2.28 and a 39.5% increased response rate among predicted responders versus the baseline rate. Notably, its prediction accuracy, obtained without any training on the treatment data, is comparable to that achieved by directly predicting the response from the images, which requires specific training on the treatment evaluation cohorts.",

author = "Hoang, {Danh Tai} and Gal Dinstag and Shulman, {Eldad D.} and Hermida, {Leandro C.} and Ben-Zvi, {Doreen S.} and Efrat Elis and Katherine Caley and Sammut, {Stephen John} and Sanju Sinha and Neelam Sinha and Dampier, {Christopher H.} and Chani Stossel and Tejas Patil and Arun Rajan and Wiem Lassoued and Julius Strauss and Shania Bailey and Clint Allen and Jason Redman and Tuvik Beker and Peng Jiang and Talia Golan and Scott Wilkinson and Sowalsky, {Adam G.} and Pine, {Sharon R.} and Carlos Caldas and Gulley, {James L.} and Kenneth Aldape and Ranit Aharonov and Stone, {Eric A.} and Eytan Ruppin",

note = "Publisher Copyright: {\textcopyright} This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply 2024.",

year = "2024",

doi = "10.1038/s43018-024-00793-2",

language = "אנגלית",

journal = "Nature Cancer",

issn = "2662-1347",

publisher = "Nature Research",

}

Hoang, DT, Dinstag, G, Shulman, ED, Hermida, LC, Ben-Zvi, DS, Elis, E, Caley, K, Sammut, SJ, Sinha, S, Sinha, N, Dampier, CH, Stossel, C, Patil, T, Rajan, A, Lassoued, W, Strauss, J, Bailey, S, Allen, C, Redman, J, Beker, T, Jiang, P, Golan, T, Wilkinson, S, Sowalsky, AG, Pine, SR, Caldas, C, Gulley, JL, Aldape, K, Aharonov, R, Stone, EA & Ruppin, E 2024, 'A deep-learning framework to predict cancer treatment response from histopathology images through imputed transcriptomics', Nature Cancer. https://doi.org/10.1038/s43018-024-00793-2

TY - JOUR

T1 - A deep-learning framework to predict cancer treatment response from histopathology images through imputed transcriptomics

AU - Hoang, Danh Tai

AU - Dinstag, Gal

AU - Shulman, Eldad D.

AU - Hermida, Leandro C.

AU - Ben-Zvi, Doreen S.

AU - Elis, Efrat

AU - Caley, Katherine

AU - Sammut, Stephen John

AU - Sinha, Sanju

AU - Sinha, Neelam

AU - Dampier, Christopher H.

AU - Stossel, Chani

AU - Patil, Tejas

AU - Rajan, Arun

AU - Lassoued, Wiem

AU - Strauss, Julius

AU - Bailey, Shania

AU - Allen, Clint

AU - Redman, Jason

AU - Beker, Tuvik

AU - Jiang, Peng

AU - Golan, Talia

AU - Wilkinson, Scott

AU - Sowalsky, Adam G.

AU - Pine, Sharon R.

AU - Caldas, Carlos

AU - Gulley, James L.

AU - Aldape, Kenneth

AU - Aharonov, Ranit

AU - Stone, Eric A.

AU - Ruppin, Eytan

PY - 2024

Y1 - 2024

N2 - Advances in artificial intelligence have paved the way for leveraging hematoxylin and eosin-stained tumor slides for precision oncology. We present ENLIGHT–DeepPT, an indirect two-step approach consisting of (1) DeepPT, a deep-learning framework that predicts genome-wide tumor mRNA expression from slides, and (2) ENLIGHT, which predicts response to targeted and immune therapies from the inferred expression values. We show that DeepPT successfully predicts transcriptomics in all 16 The Cancer Genome Atlas cohorts tested and generalizes well to two independent datasets. ENLIGHT–DeepPT successfully predicts true responders in five independent patient cohorts involving four different treatments spanning six cancer types, with an overall odds ratio of 2.28 and a 39.5% increased response rate among predicted responders versus the baseline rate. Notably, its prediction accuracy, obtained without any training on the treatment data, is comparable to that achieved by directly predicting the response from the images, which requires specific training on the treatment evaluation cohorts.

AB - Advances in artificial intelligence have paved the way for leveraging hematoxylin and eosin-stained tumor slides for precision oncology. We present ENLIGHT–DeepPT, an indirect two-step approach consisting of (1) DeepPT, a deep-learning framework that predicts genome-wide tumor mRNA expression from slides, and (2) ENLIGHT, which predicts response to targeted and immune therapies from the inferred expression values. We show that DeepPT successfully predicts transcriptomics in all 16 The Cancer Genome Atlas cohorts tested and generalizes well to two independent datasets. ENLIGHT–DeepPT successfully predicts true responders in five independent patient cohorts involving four different treatments spanning six cancer types, with an overall odds ratio of 2.28 and a 39.5% increased response rate among predicted responders versus the baseline rate. Notably, its prediction accuracy, obtained without any training on the treatment data, is comparable to that achieved by directly predicting the response from the images, which requires specific training on the treatment evaluation cohorts.

UR - http://www.scopus.com/inward/record.url?scp=85197370611&partnerID=8YFLogxK

U2 - 10.1038/s43018-024-00793-2

DO - 10.1038/s43018-024-00793-2

M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???

AN - SCOPUS:85197370611

SN - 2662-1347

JO - Nature Cancer

JF - Nature Cancer

ER -

A deep-learning framework to predict cancer treatment response from histopathology images through imputed transcriptomics

Abstract

Funding

Access to Document

Other files and links

Fingerprint

Cite this