Predicting molecular phenotypes from histopathology images: A transcriptome-wide expression–morphology analysis in breast cancer

Yinxi Wang; Kimmo Kartasalo; Philippe Weitz; Balazs Acs; Masi Valkonen; Christer Larsson; Pekka Ruusuvuori; Johan Hartman; Mattias Rantalainen

doi:10.1158/0008-5472.CAN-21-0482

Predicting molecular phenotypes from histopathology images: A transcriptome-wide expression–morphology analysis in breast cancer

Yinxi Wang
, Kimmo Kartasalo
, Philippe Weitz
, Balazs Acs
, Masi Valkonen
, Christer Larsson
, Pekka Ruusuvuori
, Johan Hartman
, Mattias Rantalainen^*

^*Corresponding author for this work

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

55 Citations (Scopus)

18 Downloads (Pure)

Abstract

Molecular profiling is central in cancer precision medicine but remains costly and is based on tumor average profiles. Morphologic patterns observable in histopathology sections from tumors are determined by the underlying molecular phenotype and therefore have the potential to be exploited for prediction of molecular phenotypes. We report here the first transcriptome-wide expression–morphology (EMO) analysis in breast cancer, where individual deep convolutional neural networks were optimized and validated for prediction of mRNA expression in 17,695 genes from hematoxylin and eosin–stained whole slide images. Predicted expressions in 9,334 (52.75%) genes were significantly associated with RNA sequencing estimates. We also demonstrated successful prediction of an mRNA-based proliferation score with established clinical value. The results were validated in independent internal and external test datasets. Predicted spatial intratumor variabilities in expression were validated through spatial transcriptomics profiling. These results suggest that EMO provides a cost-efficient and scalable approach to predict both tumor average and intratumor spatial expression from histopathology images. Significance: Transcriptome-wide expression morphology deep learning analysis enables prediction of mRNA expression and proliferation markers from routine histopathology whole slide images in breast cancer.

Original language	English
Pages (from-to)	5115-5126
Number of pages	12
Journal	Cancer Research
Volume	81
Issue number	19
DOIs	https://doi.org/10.1158/0008-5472.CAN-21-0482
Publication status	Published - 2021
Publication type	A1 Journal article-refereed

Funding

Y. Wang reports personal fees from Stratipath AB outside the submitted work. B. Ács reports grants from The Swedish Society for Medical Research (Svenska S€allskapet for Medicinsk Forsknings—SSMF) outside the submitted work. P. Ruusuvuori reports grants from Academy of Finland, ERA PerMed JTC2020, and Cancer Foundation Finland and other support from CSC Centre for Scientific Computing during the conduct of the study. J. Hartman reports grants from Swedish Cancer Fund, Medtech Labs, Swedish Breast Cancer Association, and Stockholm Cancer Society during the conduct of the study; personal fees from Roche, Pfizer, Merck, MSD, and Eli Lilly, grants from Cepheid, and grants and personal fees from Novartis outside the submitted work; and is cofounder of and shareholder in Stratipath AB. M. Rantalainen reports grants from Swedish Research Council, Swedish Cancer Society, Swedish e-Science Research Centre (SeRC), ERA PerMed (through Swedish Research Council), Karolinska Institutet (Cancer Research KI; StratCan), and MedTechLabs during the conduct of the study, and is cofounder of and shareholder in Stratipath AB. No disclosures were reported by the other authors. This project was supported by funding from the Swedish Research Council under the frame of ERA PerMed (ERAPERMED2019–224—ABCAP; M. Rantalainen), Swedish Research Council (M. Rantalainen, J. Hartman), Swedish Cancer Society (M. Rantalainen, J. Hartman), Karolinska Institutet (Cancer Research KI; StratCan; M. Rantalainen, J. Hartman), MedTechLabs (M. Rantalainen, J. Hartman), Swedish e-science Research Centre (SeRC)—eCPC (M. Rantalainen), Stockholm Region (J. Hartman), Stockholm Cancer Society (J. Hartman), Swedish Breast Cancer Association (J. Hartman), Academy of Finland (326463, 341967, and 335976; P. Ruusuvuori), Academy of Finland Center of Excellence programme (312043; P. Ruusuvuori), Cancer Foundation Finland (P. Ruusuvuori), ERA PerMed ABCAP (P. Ruusuvuori), CSC—IT Center for Science (Finland; Grand Challenge pilot project AI-EMO, 2001568; P. Ruusuvuori), Tampere University graduate school (K. Karta-salo), and University of Turku Graduate School UTUGS and Turku University Foundation (M. Valkonen). The authors would like to acknowledge the patients, clinicians, and hospital staff participating in the SCAN-B study; the staff at the central SCAN-B laboratory at Division of Oncology, Lund University; the Swedish National Breast Cancer Quality Registry (NKBC); Regional Cancer Center South; and the South Swedish Breast Cancer Group (SSBCG). They also thank Johan Vallon-Christersson (Lund University) for help in preparing these data. The authors thank Duong Nguyen Thuy Tran and other personnel that have been contributing to slide scanning operations in the Rantalainen group/CHIME project at Karolinska Institute. They thank TCGA Research Network, https://www.cancer.gov/tcga, for providing access to part of the data used in this study. The authors thank The Swedish Society for Medical Research (Svenska S€allskapet for Medicinsk Forsknings—SSMF) for a postdoctoral grant and the Hungarian Society of Senology for supporting B. Ács.

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Publication forum classification

Publication forum level 2

ASJC Scopus subject areas

Oncology
Cancer Research

Access to Document

10.1158/0008-5472.CAN-21-0482Licence: CC BY-NC-ND

5115.fullFinal published version, 2.12 MBLicence: CC BY-NC-ND

https://urn.fi/URN:NBN:fi:tuni-202111038136Licence: CC BY-NC-ND

Cite this

@article{8e30f98cf3ed45dc999787ba85007490,

title = "Predicting molecular phenotypes from histopathology images: A transcriptome-wide expression{\textendash}morphology analysis in breast cancer",

abstract = "Molecular profiling is central in cancer precision medicine but remains costly and is based on tumor average profiles. Morphologic patterns observable in histopathology sections from tumors are determined by the underlying molecular phenotype and therefore have the potential to be exploited for prediction of molecular phenotypes. We report here the first transcriptome-wide expression{\textendash}morphology (EMO) analysis in breast cancer, where individual deep convolutional neural networks were optimized and validated for prediction of mRNA expression in 17,695 genes from hematoxylin and eosin{\textendash}stained whole slide images. Predicted expressions in 9,334 (52.75\%) genes were significantly associated with RNA sequencing estimates. We also demonstrated successful prediction of an mRNA-based proliferation score with established clinical value. The results were validated in independent internal and external test datasets. Predicted spatial intratumor variabilities in expression were validated through spatial transcriptomics profiling. These results suggest that EMO provides a cost-efficient and scalable approach to predict both tumor average and intratumor spatial expression from histopathology images. Significance: Transcriptome-wide expression morphology deep learning analysis enables prediction of mRNA expression and proliferation markers from routine histopathology whole slide images in breast cancer.",

author = "Yinxi Wang and Kimmo Kartasalo and Philippe Weitz and Balazs Acs and Masi Valkonen and Christer Larsson and Pekka Ruusuvuori and Johan Hartman and Mattias Rantalainen",

note = "Funding Information: Y. Wang reports personal fees from Stratipath AB outside the submitted work. B. {\'A}cs reports grants from The Swedish Society for Medical Research (Svenska S€allskapet for Medicinsk Forsknings{\textemdash}SSMF) outside the submitted work. P. Ruusuvuori reports grants from Academy of Finland, ERA PerMed JTC2020, and Cancer Foundation Finland and other support from CSC Centre for Scientific Computing during the conduct of the study. J. Hartman reports grants from Swedish Cancer Fund, Medtech Labs, Swedish Breast Cancer Association, and Stockholm Cancer Society during the conduct of the study; personal fees from Roche, Pfizer, Merck, MSD, and Eli Lilly, grants from Cepheid, and grants and personal fees from Novartis outside the submitted work; and is cofounder of and shareholder in Stratipath AB. M. Rantalainen reports grants from Swedish Research Council, Swedish Cancer Society, Swedish e-Science Research Centre (SeRC), ERA PerMed (through Swedish Research Council), Karolinska Institutet (Cancer Research KI; StratCan), and MedTechLabs during the conduct of the study, and is cofounder of and shareholder in Stratipath AB. No disclosures were reported by the other authors. Funding Information: This project was supported by funding from the Swedish Research Council under the frame of ERA PerMed (ERAPERMED2019{\textendash}224{\textemdash}ABCAP; M. Rantalainen), Swedish Research Council (M. Rantalainen, J. Hartman), Swedish Cancer Society (M. Rantalainen, J. Hartman), Karolinska Institutet (Cancer Research KI; StratCan; M. Rantalainen, J. Hartman), MedTechLabs (M. Rantalainen, J. Hartman), Swedish e-science Research Centre (SeRC){\textemdash}eCPC (M. Rantalainen), Stockholm Region (J. Hartman), Stockholm Cancer Society (J. Hartman), Swedish Breast Cancer Association (J. Hartman), Academy of Finland (326463, 341967, and 335976; P. Ruusuvuori), Academy of Finland Center of Excellence programme (312043; P. Ruusuvuori), Cancer Foundation Finland (P. Ruusuvuori), ERA PerMed ABCAP (P. Ruusuvuori), CSC{\textemdash}IT Center for Science (Finland; Grand Challenge pilot project AI-EMO, 2001568; P. Ruusuvuori), Tampere University graduate school (K. Karta-salo), and University of Turku Graduate School UTUGS and Turku University Foundation (M. Valkonen). The authors would like to acknowledge the patients, clinicians, and hospital staff participating in the SCAN-B study; the staff at the central SCAN-B laboratory at Division of Oncology, Lund University; the Swedish National Breast Cancer Quality Registry (NKBC); Regional Cancer Center South; and the South Swedish Breast Cancer Group (SSBCG). They also thank Johan Vallon-Christersson (Lund University) for help in preparing these data. The authors thank Duong Nguyen Thuy Tran and other personnel that have been contributing to slide scanning operations in the Rantalainen group/CHIME project at Karolinska Institute. They thank TCGA Research Network, https://www.cancer.gov/tcga, for providing access to part of the data used in this study. The authors thank The Swedish Society for Medical Research (Svenska S€allskapet for Medicinsk Forsknings{\textemdash}SSMF) for a postdoctoral grant and the Hungarian Society of Senology for supporting B. {\'A}cs. Publisher Copyright: {\textcopyright} 2021 The Authors; Published by the American Association for Cancer Research",

year = "2021",

doi = "10.1158/0008-5472.CAN-21-0482",

language = "English",

volume = "81",

pages = "5115--5126",

journal = "Cancer Research",

issn = "0008-5472",

publisher = "American Association for Cancer Research Inc.",

number = "19",

}

TY - JOUR

T1 - Predicting molecular phenotypes from histopathology images

T2 - A transcriptome-wide expression–morphology analysis in breast cancer

AU - Wang, Yinxi

AU - Kartasalo, Kimmo

AU - Weitz, Philippe

AU - Acs, Balazs

AU - Valkonen, Masi

AU - Larsson, Christer

AU - Ruusuvuori, Pekka

AU - Hartman, Johan

AU - Rantalainen, Mattias

N1 - Funding Information: Y. Wang reports personal fees from Stratipath AB outside the submitted work. B. Ács reports grants from The Swedish Society for Medical Research (Svenska S€allskapet for Medicinsk Forsknings—SSMF) outside the submitted work. P. Ruusuvuori reports grants from Academy of Finland, ERA PerMed JTC2020, and Cancer Foundation Finland and other support from CSC Centre for Scientific Computing during the conduct of the study. J. Hartman reports grants from Swedish Cancer Fund, Medtech Labs, Swedish Breast Cancer Association, and Stockholm Cancer Society during the conduct of the study; personal fees from Roche, Pfizer, Merck, MSD, and Eli Lilly, grants from Cepheid, and grants and personal fees from Novartis outside the submitted work; and is cofounder of and shareholder in Stratipath AB. M. Rantalainen reports grants from Swedish Research Council, Swedish Cancer Society, Swedish e-Science Research Centre (SeRC), ERA PerMed (through Swedish Research Council), Karolinska Institutet (Cancer Research KI; StratCan), and MedTechLabs during the conduct of the study, and is cofounder of and shareholder in Stratipath AB. No disclosures were reported by the other authors. Funding Information: This project was supported by funding from the Swedish Research Council under the frame of ERA PerMed (ERAPERMED2019–224—ABCAP; M. Rantalainen), Swedish Research Council (M. Rantalainen, J. Hartman), Swedish Cancer Society (M. Rantalainen, J. Hartman), Karolinska Institutet (Cancer Research KI; StratCan; M. Rantalainen, J. Hartman), MedTechLabs (M. Rantalainen, J. Hartman), Swedish e-science Research Centre (SeRC)—eCPC (M. Rantalainen), Stockholm Region (J. Hartman), Stockholm Cancer Society (J. Hartman), Swedish Breast Cancer Association (J. Hartman), Academy of Finland (326463, 341967, and 335976; P. Ruusuvuori), Academy of Finland Center of Excellence programme (312043; P. Ruusuvuori), Cancer Foundation Finland (P. Ruusuvuori), ERA PerMed ABCAP (P. Ruusuvuori), CSC—IT Center for Science (Finland; Grand Challenge pilot project AI-EMO, 2001568; P. Ruusuvuori), Tampere University graduate school (K. Karta-salo), and University of Turku Graduate School UTUGS and Turku University Foundation (M. Valkonen). The authors would like to acknowledge the patients, clinicians, and hospital staff participating in the SCAN-B study; the staff at the central SCAN-B laboratory at Division of Oncology, Lund University; the Swedish National Breast Cancer Quality Registry (NKBC); Regional Cancer Center South; and the South Swedish Breast Cancer Group (SSBCG). They also thank Johan Vallon-Christersson (Lund University) for help in preparing these data. The authors thank Duong Nguyen Thuy Tran and other personnel that have been contributing to slide scanning operations in the Rantalainen group/CHIME project at Karolinska Institute. They thank TCGA Research Network, https://www.cancer.gov/tcga, for providing access to part of the data used in this study. The authors thank The Swedish Society for Medical Research (Svenska S€allskapet for Medicinsk Forsknings—SSMF) for a postdoctoral grant and the Hungarian Society of Senology for supporting B. Ács. Publisher Copyright: © 2021 The Authors; Published by the American Association for Cancer Research

PY - 2021

Y1 - 2021

N2 - Molecular profiling is central in cancer precision medicine but remains costly and is based on tumor average profiles. Morphologic patterns observable in histopathology sections from tumors are determined by the underlying molecular phenotype and therefore have the potential to be exploited for prediction of molecular phenotypes. We report here the first transcriptome-wide expression–morphology (EMO) analysis in breast cancer, where individual deep convolutional neural networks were optimized and validated for prediction of mRNA expression in 17,695 genes from hematoxylin and eosin–stained whole slide images. Predicted expressions in 9,334 (52.75%) genes were significantly associated with RNA sequencing estimates. We also demonstrated successful prediction of an mRNA-based proliferation score with established clinical value. The results were validated in independent internal and external test datasets. Predicted spatial intratumor variabilities in expression were validated through spatial transcriptomics profiling. These results suggest that EMO provides a cost-efficient and scalable approach to predict both tumor average and intratumor spatial expression from histopathology images. Significance: Transcriptome-wide expression morphology deep learning analysis enables prediction of mRNA expression and proliferation markers from routine histopathology whole slide images in breast cancer.

AB - Molecular profiling is central in cancer precision medicine but remains costly and is based on tumor average profiles. Morphologic patterns observable in histopathology sections from tumors are determined by the underlying molecular phenotype and therefore have the potential to be exploited for prediction of molecular phenotypes. We report here the first transcriptome-wide expression–morphology (EMO) analysis in breast cancer, where individual deep convolutional neural networks were optimized and validated for prediction of mRNA expression in 17,695 genes from hematoxylin and eosin–stained whole slide images. Predicted expressions in 9,334 (52.75%) genes were significantly associated with RNA sequencing estimates. We also demonstrated successful prediction of an mRNA-based proliferation score with established clinical value. The results were validated in independent internal and external test datasets. Predicted spatial intratumor variabilities in expression were validated through spatial transcriptomics profiling. These results suggest that EMO provides a cost-efficient and scalable approach to predict both tumor average and intratumor spatial expression from histopathology images. Significance: Transcriptome-wide expression morphology deep learning analysis enables prediction of mRNA expression and proliferation markers from routine histopathology whole slide images in breast cancer.

U2 - 10.1158/0008-5472.CAN-21-0482

DO - 10.1158/0008-5472.CAN-21-0482

M3 - Article

C2 - 34341074

AN - SCOPUS:85117044610

SN - 0008-5472

VL - 81

SP - 5115

EP - 5126

JO - Cancer Research

JF - Cancer Research

IS - 19

ER -

Predicting molecular phenotypes from histopathology images: A transcriptome-wide expression–morphology analysis in breast cancer

Abstract

Funding

UN SDGs

Publication forum classification

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this