Deep learning using tumor HLA peptide mass spectrometry datasets improves neoantigen identification

Brendan Bulik-Sullivan, Jennifer Busby, Christine D. Palmer, Matthew J. Davis, Tyler Murphy, Andrew Clark, Michele Busby, Fujiko Duke, Aaron Yang, Lauren Young, Noelle C. Ojo, Kamilah Caldwell, Jesse Abhyankar, Thomas Boucher, Meghan G. Hart, Vladimir Makarov, Vincent Thomas De Montpreville, Olaf Mercier, Timothy A. Chan, Giorgio ScagliottiPaolo Bironzo, Silvia Novello, Niki Karachaliou, Rafael Rosell, Ian Anderson, Nashat Gabrail, John Hrom, Chainarong Limvarapuss, Karin Choquette, Alexander Spira, Raphael Rousseau, Cynthia Voong, Naiyer A. Rizvi, Elie Fadel, Mark Frattini, Karin Jooss, Mojca Skoberne, Joshua Francis, Roman Yelensky

Research output: Contribution to journalArticle

Abstract

Neoantigens, which are expressed on tumor cells, are one of the main targets of an effective antitumor T-cell response. Cancer immunotherapies to target neoantigens are of growing interest and are in early human trials, but methods to identify neoantigens either require invasive or difficult-to-obtain clinical specimens, require the screening of hundreds to thousands of synthetic peptides or tandem minigenes, or are only relevant to specific human leukocyte antigen (HLA) alleles. We apply deep learning to a large (N = 74 patients) HLA peptide and genomic dataset from various human tumors to create a computational model of antigen presentation for neoantigen prediction. We show that our model, named EDGE, increases the positive predictive value of HLA antigen prediction by up to ninefold. We apply EDGE to enable identification of neoantigens and neoantigen-reactive T cells using routine clinical specimens and small numbers of synthetic peptides for most common HLA alleles. EDGE could enable an improved ability to develop neoantigen-targeted immunotherapies for cancer patients.

Original languageEnglish (US)
Pages (from-to)55-71
Number of pages17
JournalNature Biotechnology
Volume37
Issue number1
DOIs
StatePublished - Jan 1 2019
Externally publishedYes

Fingerprint

Antigens
HLA Antigens
Peptides
Mass spectrometry
Tumors
Mass Spectrometry
Learning
T-cells
Immunotherapy
Neoplasms
Alleles
T-Lymphocytes
Aptitude
Antigen Presentation
Screening
Cells
Datasets
Deep learning

ASJC Scopus subject areas

  • Biotechnology
  • Bioengineering
  • Applied Microbiology and Biotechnology
  • Molecular Medicine
  • Biomedical Engineering

Cite this

Bulik-Sullivan, B., Busby, J., Palmer, C. D., Davis, M. J., Murphy, T., Clark, A., ... Yelensky, R. (2019). Deep learning using tumor HLA peptide mass spectrometry datasets improves neoantigen identification. Nature Biotechnology, 37(1), 55-71. https://doi.org/10.1038/nbt.4313

Deep learning using tumor HLA peptide mass spectrometry datasets improves neoantigen identification. / Bulik-Sullivan, Brendan; Busby, Jennifer; Palmer, Christine D.; Davis, Matthew J.; Murphy, Tyler; Clark, Andrew; Busby, Michele; Duke, Fujiko; Yang, Aaron; Young, Lauren; Ojo, Noelle C.; Caldwell, Kamilah; Abhyankar, Jesse; Boucher, Thomas; Hart, Meghan G.; Makarov, Vladimir; De Montpreville, Vincent Thomas; Mercier, Olaf; Chan, Timothy A.; Scagliotti, Giorgio; Bironzo, Paolo; Novello, Silvia; Karachaliou, Niki; Rosell, Rafael; Anderson, Ian; Gabrail, Nashat; Hrom, John; Limvarapuss, Chainarong; Choquette, Karin; Spira, Alexander; Rousseau, Raphael; Voong, Cynthia; Rizvi, Naiyer A.; Fadel, Elie; Frattini, Mark; Jooss, Karin; Skoberne, Mojca; Francis, Joshua; Yelensky, Roman.

In: Nature Biotechnology, Vol. 37, No. 1, 01.01.2019, p. 55-71.

Research output: Contribution to journalArticle

Bulik-Sullivan, B, Busby, J, Palmer, CD, Davis, MJ, Murphy, T, Clark, A, Busby, M, Duke, F, Yang, A, Young, L, Ojo, NC, Caldwell, K, Abhyankar, J, Boucher, T, Hart, MG, Makarov, V, De Montpreville, VT, Mercier, O, Chan, TA, Scagliotti, G, Bironzo, P, Novello, S, Karachaliou, N, Rosell, R, Anderson, I, Gabrail, N, Hrom, J, Limvarapuss, C, Choquette, K, Spira, A, Rousseau, R, Voong, C, Rizvi, NA, Fadel, E, Frattini, M, Jooss, K, Skoberne, M, Francis, J & Yelensky, R 2019, 'Deep learning using tumor HLA peptide mass spectrometry datasets improves neoantigen identification', Nature Biotechnology, vol. 37, no. 1, pp. 55-71. https://doi.org/10.1038/nbt.4313
Bulik-Sullivan B, Busby J, Palmer CD, Davis MJ, Murphy T, Clark A et al. Deep learning using tumor HLA peptide mass spectrometry datasets improves neoantigen identification. Nature Biotechnology. 2019 Jan 1;37(1):55-71. https://doi.org/10.1038/nbt.4313
Bulik-Sullivan, Brendan ; Busby, Jennifer ; Palmer, Christine D. ; Davis, Matthew J. ; Murphy, Tyler ; Clark, Andrew ; Busby, Michele ; Duke, Fujiko ; Yang, Aaron ; Young, Lauren ; Ojo, Noelle C. ; Caldwell, Kamilah ; Abhyankar, Jesse ; Boucher, Thomas ; Hart, Meghan G. ; Makarov, Vladimir ; De Montpreville, Vincent Thomas ; Mercier, Olaf ; Chan, Timothy A. ; Scagliotti, Giorgio ; Bironzo, Paolo ; Novello, Silvia ; Karachaliou, Niki ; Rosell, Rafael ; Anderson, Ian ; Gabrail, Nashat ; Hrom, John ; Limvarapuss, Chainarong ; Choquette, Karin ; Spira, Alexander ; Rousseau, Raphael ; Voong, Cynthia ; Rizvi, Naiyer A. ; Fadel, Elie ; Frattini, Mark ; Jooss, Karin ; Skoberne, Mojca ; Francis, Joshua ; Yelensky, Roman. / Deep learning using tumor HLA peptide mass spectrometry datasets improves neoantigen identification. In: Nature Biotechnology. 2019 ; Vol. 37, No. 1. pp. 55-71.
@article{df19023aafdb4690aa18f35d038766b3,
title = "Deep learning using tumor HLA peptide mass spectrometry datasets improves neoantigen identification",
abstract = "Neoantigens, which are expressed on tumor cells, are one of the main targets of an effective antitumor T-cell response. Cancer immunotherapies to target neoantigens are of growing interest and are in early human trials, but methods to identify neoantigens either require invasive or difficult-to-obtain clinical specimens, require the screening of hundreds to thousands of synthetic peptides or tandem minigenes, or are only relevant to specific human leukocyte antigen (HLA) alleles. We apply deep learning to a large (N = 74 patients) HLA peptide and genomic dataset from various human tumors to create a computational model of antigen presentation for neoantigen prediction. We show that our model, named EDGE, increases the positive predictive value of HLA antigen prediction by up to ninefold. We apply EDGE to enable identification of neoantigens and neoantigen-reactive T cells using routine clinical specimens and small numbers of synthetic peptides for most common HLA alleles. EDGE could enable an improved ability to develop neoantigen-targeted immunotherapies for cancer patients.",
author = "Brendan Bulik-Sullivan and Jennifer Busby and Palmer, {Christine D.} and Davis, {Matthew J.} and Tyler Murphy and Andrew Clark and Michele Busby and Fujiko Duke and Aaron Yang and Lauren Young and Ojo, {Noelle C.} and Kamilah Caldwell and Jesse Abhyankar and Thomas Boucher and Hart, {Meghan G.} and Vladimir Makarov and {De Montpreville}, {Vincent Thomas} and Olaf Mercier and Chan, {Timothy A.} and Giorgio Scagliotti and Paolo Bironzo and Silvia Novello and Niki Karachaliou and Rafael Rosell and Ian Anderson and Nashat Gabrail and John Hrom and Chainarong Limvarapuss and Karin Choquette and Alexander Spira and Raphael Rousseau and Cynthia Voong and Rizvi, {Naiyer A.} and Elie Fadel and Mark Frattini and Karin Jooss and Mojca Skoberne and Joshua Francis and Roman Yelensky",
year = "2019",
month = "1",
day = "1",
doi = "10.1038/nbt.4313",
language = "English (US)",
volume = "37",
pages = "55--71",
journal = "Biotechnology",
issn = "1087-0156",
publisher = "Nature Publishing Group",
number = "1",

}

TY - JOUR

T1 - Deep learning using tumor HLA peptide mass spectrometry datasets improves neoantigen identification

AU - Bulik-Sullivan, Brendan

AU - Busby, Jennifer

AU - Palmer, Christine D.

AU - Davis, Matthew J.

AU - Murphy, Tyler

AU - Clark, Andrew

AU - Busby, Michele

AU - Duke, Fujiko

AU - Yang, Aaron

AU - Young, Lauren

AU - Ojo, Noelle C.

AU - Caldwell, Kamilah

AU - Abhyankar, Jesse

AU - Boucher, Thomas

AU - Hart, Meghan G.

AU - Makarov, Vladimir

AU - De Montpreville, Vincent Thomas

AU - Mercier, Olaf

AU - Chan, Timothy A.

AU - Scagliotti, Giorgio

AU - Bironzo, Paolo

AU - Novello, Silvia

AU - Karachaliou, Niki

AU - Rosell, Rafael

AU - Anderson, Ian

AU - Gabrail, Nashat

AU - Hrom, John

AU - Limvarapuss, Chainarong

AU - Choquette, Karin

AU - Spira, Alexander

AU - Rousseau, Raphael

AU - Voong, Cynthia

AU - Rizvi, Naiyer A.

AU - Fadel, Elie

AU - Frattini, Mark

AU - Jooss, Karin

AU - Skoberne, Mojca

AU - Francis, Joshua

AU - Yelensky, Roman

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Neoantigens, which are expressed on tumor cells, are one of the main targets of an effective antitumor T-cell response. Cancer immunotherapies to target neoantigens are of growing interest and are in early human trials, but methods to identify neoantigens either require invasive or difficult-to-obtain clinical specimens, require the screening of hundreds to thousands of synthetic peptides or tandem minigenes, or are only relevant to specific human leukocyte antigen (HLA) alleles. We apply deep learning to a large (N = 74 patients) HLA peptide and genomic dataset from various human tumors to create a computational model of antigen presentation for neoantigen prediction. We show that our model, named EDGE, increases the positive predictive value of HLA antigen prediction by up to ninefold. We apply EDGE to enable identification of neoantigens and neoantigen-reactive T cells using routine clinical specimens and small numbers of synthetic peptides for most common HLA alleles. EDGE could enable an improved ability to develop neoantigen-targeted immunotherapies for cancer patients.

AB - Neoantigens, which are expressed on tumor cells, are one of the main targets of an effective antitumor T-cell response. Cancer immunotherapies to target neoantigens are of growing interest and are in early human trials, but methods to identify neoantigens either require invasive or difficult-to-obtain clinical specimens, require the screening of hundreds to thousands of synthetic peptides or tandem minigenes, or are only relevant to specific human leukocyte antigen (HLA) alleles. We apply deep learning to a large (N = 74 patients) HLA peptide and genomic dataset from various human tumors to create a computational model of antigen presentation for neoantigen prediction. We show that our model, named EDGE, increases the positive predictive value of HLA antigen prediction by up to ninefold. We apply EDGE to enable identification of neoantigens and neoantigen-reactive T cells using routine clinical specimens and small numbers of synthetic peptides for most common HLA alleles. EDGE could enable an improved ability to develop neoantigen-targeted immunotherapies for cancer patients.

UR - http://www.scopus.com/inward/record.url?scp=85059650557&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85059650557&partnerID=8YFLogxK

U2 - 10.1038/nbt.4313

DO - 10.1038/nbt.4313

M3 - Article

C2 - 30556813

AN - SCOPUS:85059650557

VL - 37

SP - 55

EP - 71

JO - Biotechnology

JF - Biotechnology

SN - 1087-0156

IS - 1

ER -