Machine learning analysis of DNA methylation profiles distinguishes primary lung squamous cell carcinomas from head and neck metastases

Philipp Jurmeister, Michael Bockmayr, Philipp Seegerer, Teresa Bockmayr, Denise Treue, Grégoire Montavon, Claudia Vollbrecht, Alexander Arnold, Daniel Teichmann, Keno Bressem, Ulrich Schüller, Maximilian von Laffert, Klaus Muller, David Capper, Frederick Klauschen

Research output: Contribution to journalArticle

Abstract

Head and neck squamous cell carcinoma (HNSC) patients are at risk of suffering from both pulmonary metastases or a second squamous cell carcinoma of the lung (LUSC). Differentiating pulmonary metastases from primary lung cancers is of high clinical importance, but not possible in most cases with current diagnostics. To address this, we performed DNA methylation profiling of primary tumors and trained three different machine learning methods to distinguish metastatic HNSC from primary LUSC. We developed an artificial neural network that correctly classified 96.4% of the cases in a validation cohort of 279 patients with HNSC and LUSC as well as normal lung controls, outperforming support vector machines (95.7%) and random forests (87.8%). Prediction accuracies of more than 99% were achieved for 92.1% (neural network), 90% (support vector machine), and 43% (random forest) of these cases by applying thresholds to the resulting probability scores and excluding samples with low confidence. As independent clinical validation of the approach, we analyzed a series of 51 patients with a history of HNSC and a second lung tumor, demonstrating the correct classifications based on clinicopathological properties. In summary, our approach may facilitate the reliable diagnostic differentiation of pulmonary metastases of HNSC from primary LUSC to guide therapeutic decisions.

Original languageEnglish
JournalScience translational medicine
Volume11
Issue number509
DOIs
Publication statusPublished - 2019 Sep 11

Fingerprint

DNA Methylation
Neoplasm Metastasis
Lung
DNA Fingerprinting
Carcinoma, squamous cell of head and neck
Machine Learning
Squamous Cell Carcinoma
Lung Neoplasms
Neoplasms

ASJC Scopus subject areas

  • Medicine(all)

Cite this

Machine learning analysis of DNA methylation profiles distinguishes primary lung squamous cell carcinomas from head and neck metastases. / Jurmeister, Philipp; Bockmayr, Michael; Seegerer, Philipp; Bockmayr, Teresa; Treue, Denise; Montavon, Grégoire; Vollbrecht, Claudia; Arnold, Alexander; Teichmann, Daniel; Bressem, Keno; Schüller, Ulrich; von Laffert, Maximilian; Muller, Klaus; Capper, David; Klauschen, Frederick.

In: Science translational medicine, Vol. 11, No. 509, 11.09.2019.

Research output: Contribution to journalArticle

Jurmeister, P, Bockmayr, M, Seegerer, P, Bockmayr, T, Treue, D, Montavon, G, Vollbrecht, C, Arnold, A, Teichmann, D, Bressem, K, Schüller, U, von Laffert, M, Muller, K, Capper, D & Klauschen, F 2019, 'Machine learning analysis of DNA methylation profiles distinguishes primary lung squamous cell carcinomas from head and neck metastases', Science translational medicine, vol. 11, no. 509. https://doi.org/10.1126/scitranslmed.aaw8513
Jurmeister, Philipp ; Bockmayr, Michael ; Seegerer, Philipp ; Bockmayr, Teresa ; Treue, Denise ; Montavon, Grégoire ; Vollbrecht, Claudia ; Arnold, Alexander ; Teichmann, Daniel ; Bressem, Keno ; Schüller, Ulrich ; von Laffert, Maximilian ; Muller, Klaus ; Capper, David ; Klauschen, Frederick. / Machine learning analysis of DNA methylation profiles distinguishes primary lung squamous cell carcinomas from head and neck metastases. In: Science translational medicine. 2019 ; Vol. 11, No. 509.
@article{acb90dd706f54d2192269d672af63d34,
title = "Machine learning analysis of DNA methylation profiles distinguishes primary lung squamous cell carcinomas from head and neck metastases",
abstract = "Head and neck squamous cell carcinoma (HNSC) patients are at risk of suffering from both pulmonary metastases or a second squamous cell carcinoma of the lung (LUSC). Differentiating pulmonary metastases from primary lung cancers is of high clinical importance, but not possible in most cases with current diagnostics. To address this, we performed DNA methylation profiling of primary tumors and trained three different machine learning methods to distinguish metastatic HNSC from primary LUSC. We developed an artificial neural network that correctly classified 96.4{\%} of the cases in a validation cohort of 279 patients with HNSC and LUSC as well as normal lung controls, outperforming support vector machines (95.7{\%}) and random forests (87.8{\%}). Prediction accuracies of more than 99{\%} were achieved for 92.1{\%} (neural network), 90{\%} (support vector machine), and 43{\%} (random forest) of these cases by applying thresholds to the resulting probability scores and excluding samples with low confidence. As independent clinical validation of the approach, we analyzed a series of 51 patients with a history of HNSC and a second lung tumor, demonstrating the correct classifications based on clinicopathological properties. In summary, our approach may facilitate the reliable diagnostic differentiation of pulmonary metastases of HNSC from primary LUSC to guide therapeutic decisions.",
author = "Philipp Jurmeister and Michael Bockmayr and Philipp Seegerer and Teresa Bockmayr and Denise Treue and Gr{\'e}goire Montavon and Claudia Vollbrecht and Alexander Arnold and Daniel Teichmann and Keno Bressem and Ulrich Sch{\"u}ller and {von Laffert}, Maximilian and Klaus Muller and David Capper and Frederick Klauschen",
year = "2019",
month = "9",
day = "11",
doi = "10.1126/scitranslmed.aaw8513",
language = "English",
volume = "11",
journal = "Science Translational Medicine",
issn = "1946-6234",
publisher = "American Association for the Advancement of Science",
number = "509",

}

TY - JOUR

T1 - Machine learning analysis of DNA methylation profiles distinguishes primary lung squamous cell carcinomas from head and neck metastases

AU - Jurmeister, Philipp

AU - Bockmayr, Michael

AU - Seegerer, Philipp

AU - Bockmayr, Teresa

AU - Treue, Denise

AU - Montavon, Grégoire

AU - Vollbrecht, Claudia

AU - Arnold, Alexander

AU - Teichmann, Daniel

AU - Bressem, Keno

AU - Schüller, Ulrich

AU - von Laffert, Maximilian

AU - Muller, Klaus

AU - Capper, David

AU - Klauschen, Frederick

PY - 2019/9/11

Y1 - 2019/9/11

N2 - Head and neck squamous cell carcinoma (HNSC) patients are at risk of suffering from both pulmonary metastases or a second squamous cell carcinoma of the lung (LUSC). Differentiating pulmonary metastases from primary lung cancers is of high clinical importance, but not possible in most cases with current diagnostics. To address this, we performed DNA methylation profiling of primary tumors and trained three different machine learning methods to distinguish metastatic HNSC from primary LUSC. We developed an artificial neural network that correctly classified 96.4% of the cases in a validation cohort of 279 patients with HNSC and LUSC as well as normal lung controls, outperforming support vector machines (95.7%) and random forests (87.8%). Prediction accuracies of more than 99% were achieved for 92.1% (neural network), 90% (support vector machine), and 43% (random forest) of these cases by applying thresholds to the resulting probability scores and excluding samples with low confidence. As independent clinical validation of the approach, we analyzed a series of 51 patients with a history of HNSC and a second lung tumor, demonstrating the correct classifications based on clinicopathological properties. In summary, our approach may facilitate the reliable diagnostic differentiation of pulmonary metastases of HNSC from primary LUSC to guide therapeutic decisions.

AB - Head and neck squamous cell carcinoma (HNSC) patients are at risk of suffering from both pulmonary metastases or a second squamous cell carcinoma of the lung (LUSC). Differentiating pulmonary metastases from primary lung cancers is of high clinical importance, but not possible in most cases with current diagnostics. To address this, we performed DNA methylation profiling of primary tumors and trained three different machine learning methods to distinguish metastatic HNSC from primary LUSC. We developed an artificial neural network that correctly classified 96.4% of the cases in a validation cohort of 279 patients with HNSC and LUSC as well as normal lung controls, outperforming support vector machines (95.7%) and random forests (87.8%). Prediction accuracies of more than 99% were achieved for 92.1% (neural network), 90% (support vector machine), and 43% (random forest) of these cases by applying thresholds to the resulting probability scores and excluding samples with low confidence. As independent clinical validation of the approach, we analyzed a series of 51 patients with a history of HNSC and a second lung tumor, demonstrating the correct classifications based on clinicopathological properties. In summary, our approach may facilitate the reliable diagnostic differentiation of pulmonary metastases of HNSC from primary LUSC to guide therapeutic decisions.

UR - http://www.scopus.com/inward/record.url?scp=85072113453&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85072113453&partnerID=8YFLogxK

U2 - 10.1126/scitranslmed.aaw8513

DO - 10.1126/scitranslmed.aaw8513

M3 - Article

C2 - 31511427

AN - SCOPUS:85072113453

VL - 11

JO - Science Translational Medicine

JF - Science Translational Medicine

SN - 1946-6234

IS - 509

ER -