The relationship between tumor immune responses and tumor neoantigens is one of the most fundamental and unsolved questions in tumor immunology, and is the key to understanding the inefficiency of immunotherapy observed in many cancer patients. However, the properties of neoantigens that can elicit immune responses remain unclear. This biological problem can be represented and solved under a multiple instance learning framework, which seeks to model multiple instances (neoantigens) within each bag (patient specimen) with the continuous response (T cell infiltration) observed for each bag. To this end, we develop a Bayesian multiple instance regression method, named BMIR, using a Gaussian distribution to address continuous responses and latent binary variables to model primary instances in bags. By means of such Bayesian modeling, BMIR can learn a function for predicting the bag-level responses and for identifying the primary instances within bags, as well as give access to Bayesian statistical inference, which are elusive in existing works. We demonstrate the superiority of BMIR over previously proposed optimization-based methods for multiple instance regression through simulation and real data analyses. Our method is implemented in R package entitled “BayesianMIR” and is available at https://github.com/inmybrain/BayesianMIR.
CarbonneauMACheplyginaVGrangerE, et al.
Multiple instance learning: a survey of problem characteristics and applications. Pattern Recognition2018;
77: 329–353.
2.
Wang Z, Radosavljevic V, Han B, et al. Aerosol optical depth prediction from satellite observations by multiple instance regression. In: Proceedings of the SIAM International Conference on Data Mining, Atlanta, Georgia, USA, 24--26 April 2008, pp.165--176.
3.
DavisJCostaVSRayS, et al. An integrated approach to feature invention and model construction for drug activity prediction. In: Proceedings of the 24th international conference on machine learning. ICML ‘07, New York, NY, pp.217–224.
Lu L, Bi J, Wolf M, et al. Effective 3d object detection and regression using probabilistic segmentation features in CT images. In: 2011 IEEE Conference on Computer Vision and Pattern Recognition, Colorado Springs, Colorado, USA, 20--25 June 2011, pp. 1049--1056.
6.
Pappas N and Popescu-Belis A. Explaining the stars: Weighted multiple-instance learning for aspect-based sentiment analysis. In: Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing, Doha, Qatar, 25--29 October 2014, pp. 455--466.
FouldsJFrankE.A review of multi-instance learning assumptions. Knowledge Eng Rev2010;
25: 1–25.
9.
CheplyginaVTaxDMLoogM.Multiple instance learning with bag dissimilarities. Pattern Recog2015;
48: 264–275.
10.
WagstaffKLLaneT. Salience assignment for multiple-instance regression. In: Proceedings of the ICML 2007 workshop on constrained optimization and structured output spaces, Corvallis, Oregon, USA, 24 June.
RaySPageD. Multiple instance regression. In: Proceedings of the 18th international conference on machine learning, Williamstown, Massachusetts, USA, June 28--July 1 2001, pp.425--432
13.
DuXZareA.Multiple instance choquet integral classifier fusion and regression for remote sensing applications. IEEE Transact Geosci Remote Sensing2019;
57: 2741–2753.
14.
SubramanianSRanaSGuptaS, et al. Bayesian nonparametric multiple instance regression. In: 2016 23rd International conference on pattern recognition, Cancun, Mexico, 04--08 December 2016, pp.3661--3666.
15.
Gerald PL and Beatriz MC. Neoantigen Vaccines Pass the Immunogenicity Test. Trends in Molecular Medicine 2017; 23: 869--871.
16.
VerdegaalEMEde MirandaNFCCVisserM, et al.
Neoantigen landscape dynamics during human melanoma-t cell interactions.Nature2016;
536: 91.
17.
TranEAhmadzadehMLuYC, et al.
Immunogenicity of somatic mutations in human gastrointestinal cancers.Science2015;
350: 1387–1390.
18.
AkramAInmanRD.Immunodominance: a pivotal principle in host response to viral infections.Clin Immunol2012;
143: 99–115.
19.
RodriguezFHarkinsSSlifkaMK, et al.
Immunodominance in virus-induced cd8+ t-cell responses is dramatically modified by DNA immunization and is regulated by gamma interferon.J Virol2002;
76: 4251–4259.
20.
YewdellJWValMD.Immunodominance in tcd8+ responses to viruses: cell biology, cellular immunology, and mathematical models.Immunity2004;
21: 149–153.
21.
ZellnerA.Applications of bayesian analysis in econometrics.J Royal Stat Soc Ser D (The Statistician)1983;
32: 23–34.
22.
ZellnerA.On assessing prior distributions and Bayesian regression analysis with g-prior distributions, Chapter 15. North-Holland; Elsevier Science Pub. Co, 1986, pp.233–243.
23.
KassREWassermanL.A reference Bayesian test for nested hypotheses and its relationship to the Schwarz criterion. J Am Stat Assoc1995;
90: 928–934.
LiawAWienerM.Classification and regression by randomforest. R News2002;
2: 18–22.
26.
GelmanARubinDB.Inference from iterative simulation using multiple sequences. Stat Sci1992;
7: 457–472.
27.
The Cancer Genome Atlas Research Network.Comprehensive molecular characterization of clear cell renal cell carcinoma. Nature2013;
499: 43–49.
28.
SatoYYoshizatoTShiraishiY, et al.
Integrated molecular analysis of clear-cell renal cell carcinoma.Nat Genet2013;
45: 860–867.
29.
MiaoDMargolisCAGaoW, et al.
Genomic correlates of response to immune checkpoint therapies in clear cell renal cell carcinoma.Science2018;
359: 801–806.
30.
WangTLuRKapurP, et al. An empirical approach leveraging tumorgrafts to dissect the tumor microenvironment in renal cell carcinoma identifies missing link to prognostic inflammatory factors. Cancer Discov2018; 8: 1142--1155.
31.
Lu T, Wang S, Xu L, et al. Tumor neoantigenicity assessment with CSiN score incorporates clonality and immunogenicity to predict immunotherapy outcomes. Science Immunology 2020; 5. doi: 10.1126/sciimmunol.aaz3199
32.
ChowellDKrishnaSBeckerPD, et al.
Tcr contact residue hydrophobicity is a hallmark of immunogenic cd8+ t cell epitopes.Proc Natl Acad Sci2015;
112: E1754–E1762.
33.
LukszaMRiazNMakarovV, et al.
A neoantigen fitness model predicts tumour response to checkpoint blockade immunotherapy.Nature2017;
551: 517–520.
34.
NielsenMLundegaardCBlicherT, et al.
Netmhcpan, a method for quantitative predictions of peptide binding to any hla-a and -b locus protein of known sequence. PloS One2007;
2: 1–10.
35.
CalisJJAMaybenoMGreenbaumJA, et al.
Properties of mhc class i presented peptides that enhance immunogenicity. PloS Computat Biol2013;
9: 1–13.
36.
LarsenMVLundegaardCLamberthK, et al.
Large-scale validation of methods for cytotoxic t-lymphocyte epitope prediction.BMC Bioinform2007;
8: 424.
37.
TurajlicSLitchfieldKXuH, et al.
Insertion-and-deletion-derived tumour-specific neoantigens and the immunogenic phenotype: a pan-cancer analysis. Lancet Oncol2017;
18: 1009–1021.
38.
CarrenoBMMagriniVBecker-HapakM, et al.
A dendritic cell vaccine increases the breadth and diversity of melanoma neoantigen-specific t cells. Science2015;
348: 803–808.
39.
RajasagiMShuklaSAFritschEF, et al.
Systematic identification of personal tumor-specific neoantigens in chronic lymphocytic leukemia.Blood2014;
124: 453–462.
40.
DavisJGoadrichM. The relationship between precision-recall and roc curve. In: Proceedings of the 23rd international conference on machine learning, Pittsburgh, Pennsylvania, USA, 25--29 June 2006, pp.233--240.
41.
Wang S, Yang D, Rong R, et al. Artificial intelligence in lung cancer pathology image analysis. Cancers 2019; 11: 1673. doi: 10.3390/cancers11111673
42.
GeorgeEIMcCullochRE.Variable selection via Gibbs sampling. J Am Stat Assoc1993;
88: 881–889.
43.
MitchellTJBeauchampJJ.Bayesian variable selection in linear regression. J Am Stat Assoc1988;
83: 1023–1032.
44.
CastilloISchmidt-HieberJvan der VaartA.Bayesian linear regression with sparse priors. Annals Stat2015;
43: 1986–2018.
45.
CarvalhoCMPolsonNGScottJG. Handling sparsity via the horseshoe. In: Proceedings of the 12th international conference on artificial intelligence and statistics, Clearwater, Florida USA, 16--19 April 2009, pp.73--80.
46.
PiironenJVehtariA.Sparsity information and regularization in the horseshoe and other shrinkage priors. Electronic J Stat2017;
11: 5018–5051.
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
