A novel machine learning approach reveals latent vascular phenotypes predictive of renal cancer outcome

Nathan Ing, Fangjin Huang, Andrew Conley, Sungyong You, Zhaoxuan Ma, Sergey Klimov, Chisato Ohe, Xiaopu Yuan, Mahul Amin, Robert Figlin, Arkadiusz Gertych, Beatrice S. Knudsen

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

Gene expression signatures are commonly used as predictive biomarkers, but do not capture structural features within the tissue architecture. Here we apply a 2-step machine learning framework for quantitative imaging of tumor vasculature to derive a spatially informed, prognostic gene signature. The trained algorithms classify endothelial cells and generate a vascular area mask (VAM) in H&E micrographs of clear cell renal cell carcinoma (ccRCC) cases from The Cancer Genome Atlas (TCGA). Quantification of VAMs led to the discovery of 9 vascular features (9VF) that predicted disease-free-survival in a discovery cohort (n = 64, HR = 2.3). Correlation analysis and information gain identified a 14 gene expression signature related to the 9VF's. Two generalized linear models with elastic net regularization (14VF and 14GT), based on the 14 genes, separated independent cohorts of up to 301 cases into good and poor disease-free survival groups (14VF HR = 2.4, 14GT HR = 3.33). For the first time, we successfully applied digital image analysis and targeted machine learning to develop prognostic, morphology-based, gene expression signatures from the vascular architecture. This novel morphogenomic approach has the potential to improve previous methods for biomarker development.

Original languageEnglish (US)
Article number13190
JournalScientific reports
Volume7
Issue number1
DOIs
StatePublished - Dec 1 2017

Fingerprint

Kidney Neoplasms
Transcriptome
Blood Vessels
Phenotype
Disease-Free Survival
Biomarkers
Atlases
Masks
Renal Cell Carcinoma
Genes
Linear Models
Neoplasms
Endothelial Cells
Genome
Machine Learning

All Science Journal Classification (ASJC) codes

  • General

Cite this

A novel machine learning approach reveals latent vascular phenotypes predictive of renal cancer outcome. / Ing, Nathan; Huang, Fangjin; Conley, Andrew; You, Sungyong; Ma, Zhaoxuan; Klimov, Sergey; Ohe, Chisato; Yuan, Xiaopu; Amin, Mahul; Figlin, Robert; Gertych, Arkadiusz; Knudsen, Beatrice S.

In: Scientific reports, Vol. 7, No. 1, 13190, 01.12.2017.

Research output: Contribution to journalArticle

Ing, N, Huang, F, Conley, A, You, S, Ma, Z, Klimov, S, Ohe, C, Yuan, X, Amin, M, Figlin, R, Gertych, A & Knudsen, BS 2017, 'A novel machine learning approach reveals latent vascular phenotypes predictive of renal cancer outcome', Scientific reports, vol. 7, no. 1, 13190. https://doi.org/10.1038/s41598-017-13196-4
Ing, Nathan ; Huang, Fangjin ; Conley, Andrew ; You, Sungyong ; Ma, Zhaoxuan ; Klimov, Sergey ; Ohe, Chisato ; Yuan, Xiaopu ; Amin, Mahul ; Figlin, Robert ; Gertych, Arkadiusz ; Knudsen, Beatrice S. / A novel machine learning approach reveals latent vascular phenotypes predictive of renal cancer outcome. In: Scientific reports. 2017 ; Vol. 7, No. 1.
@article{b490d10f211041f2987d02a6a237097b,
title = "A novel machine learning approach reveals latent vascular phenotypes predictive of renal cancer outcome",
abstract = "Gene expression signatures are commonly used as predictive biomarkers, but do not capture structural features within the tissue architecture. Here we apply a 2-step machine learning framework for quantitative imaging of tumor vasculature to derive a spatially informed, prognostic gene signature. The trained algorithms classify endothelial cells and generate a vascular area mask (VAM) in H&E micrographs of clear cell renal cell carcinoma (ccRCC) cases from The Cancer Genome Atlas (TCGA). Quantification of VAMs led to the discovery of 9 vascular features (9VF) that predicted disease-free-survival in a discovery cohort (n = 64, HR = 2.3). Correlation analysis and information gain identified a 14 gene expression signature related to the 9VF's. Two generalized linear models with elastic net regularization (14VF and 14GT), based on the 14 genes, separated independent cohorts of up to 301 cases into good and poor disease-free survival groups (14VF HR = 2.4, 14GT HR = 3.33). For the first time, we successfully applied digital image analysis and targeted machine learning to develop prognostic, morphology-based, gene expression signatures from the vascular architecture. This novel morphogenomic approach has the potential to improve previous methods for biomarker development.",
author = "Nathan Ing and Fangjin Huang and Andrew Conley and Sungyong You and Zhaoxuan Ma and Sergey Klimov and Chisato Ohe and Xiaopu Yuan and Mahul Amin and Robert Figlin and Arkadiusz Gertych and Knudsen, {Beatrice S.}",
year = "2017",
month = "12",
day = "1",
doi = "10.1038/s41598-017-13196-4",
language = "English (US)",
volume = "7",
journal = "Scientific Reports",
issn = "2045-2322",
publisher = "Nature Publishing Group",
number = "1",

}

TY - JOUR

T1 - A novel machine learning approach reveals latent vascular phenotypes predictive of renal cancer outcome

AU - Ing, Nathan

AU - Huang, Fangjin

AU - Conley, Andrew

AU - You, Sungyong

AU - Ma, Zhaoxuan

AU - Klimov, Sergey

AU - Ohe, Chisato

AU - Yuan, Xiaopu

AU - Amin, Mahul

AU - Figlin, Robert

AU - Gertych, Arkadiusz

AU - Knudsen, Beatrice S.

PY - 2017/12/1

Y1 - 2017/12/1

N2 - Gene expression signatures are commonly used as predictive biomarkers, but do not capture structural features within the tissue architecture. Here we apply a 2-step machine learning framework for quantitative imaging of tumor vasculature to derive a spatially informed, prognostic gene signature. The trained algorithms classify endothelial cells and generate a vascular area mask (VAM) in H&E micrographs of clear cell renal cell carcinoma (ccRCC) cases from The Cancer Genome Atlas (TCGA). Quantification of VAMs led to the discovery of 9 vascular features (9VF) that predicted disease-free-survival in a discovery cohort (n = 64, HR = 2.3). Correlation analysis and information gain identified a 14 gene expression signature related to the 9VF's. Two generalized linear models with elastic net regularization (14VF and 14GT), based on the 14 genes, separated independent cohorts of up to 301 cases into good and poor disease-free survival groups (14VF HR = 2.4, 14GT HR = 3.33). For the first time, we successfully applied digital image analysis and targeted machine learning to develop prognostic, morphology-based, gene expression signatures from the vascular architecture. This novel morphogenomic approach has the potential to improve previous methods for biomarker development.

AB - Gene expression signatures are commonly used as predictive biomarkers, but do not capture structural features within the tissue architecture. Here we apply a 2-step machine learning framework for quantitative imaging of tumor vasculature to derive a spatially informed, prognostic gene signature. The trained algorithms classify endothelial cells and generate a vascular area mask (VAM) in H&E micrographs of clear cell renal cell carcinoma (ccRCC) cases from The Cancer Genome Atlas (TCGA). Quantification of VAMs led to the discovery of 9 vascular features (9VF) that predicted disease-free-survival in a discovery cohort (n = 64, HR = 2.3). Correlation analysis and information gain identified a 14 gene expression signature related to the 9VF's. Two generalized linear models with elastic net regularization (14VF and 14GT), based on the 14 genes, separated independent cohorts of up to 301 cases into good and poor disease-free survival groups (14VF HR = 2.4, 14GT HR = 3.33). For the first time, we successfully applied digital image analysis and targeted machine learning to develop prognostic, morphology-based, gene expression signatures from the vascular architecture. This novel morphogenomic approach has the potential to improve previous methods for biomarker development.

UR - http://www.scopus.com/inward/record.url?scp=85031716982&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85031716982&partnerID=8YFLogxK

U2 - 10.1038/s41598-017-13196-4

DO - 10.1038/s41598-017-13196-4

M3 - Article

VL - 7

JO - Scientific Reports

JF - Scientific Reports

SN - 2045-2322

IS - 1

M1 - 13190

ER -