Method for administering a cancer treatment
Provided is a computer-implemented method, including inputting to a trained machine learning classifier genomic information of a non-training subject that includes features from a tumor sample, wherein the trained machine learning classifier trained on features of tumor samples obtained from training subjects and their a responsiveness to checkpoint inhibition treatment and the machine-learning classifier is trained to predict responsiveness to the treatment, and generating a checkpoint inhibition responsiveness classification predictive of the subject's responding to the checkpoint inhibition with the trained machine-learning classifier, and reporting the checkpoint inhibition responsiveness classification using a graphical user interface. Also provided are a computer system for performing the method and a machine learning classifier trained by the method.
1. A method for administering a cancer treatment, comprising:
a) obtaining a responsiveness score for a non-training subject having cancer, said obtaining comprising:
i) inputting total mutational burden, tumor infiltration, and factor expression of the non-training subject to a machine learning classifier;
wherein total mutational burden comprises one or both of all mutations and nonsynonymous mutations, tumor infiltration comprises any one or more of tumor infiltration by cells expressing cluster of differentiation 8 (CD8), tumor infiltration by cells expressing cluster of differentiation 4 (CD4), and tumor infiltration by cells expressing cluster of differentiation 19 (CD19), and factor expression comprises any one or more of beta 2 microglobulin (B2M) expression, proteasome subunit beta 10 (PSMB10) expression, antigen peptide transmitter 1 (TAP1) expression, antigen peptide transporter 2 (TAP2) expression, human leukocyte antigen A (HLA-A) expression, major histocompatibility complex class I B (HLA-B) expression, major histocompatibility complex class I C (HLA-C) expression, major histocompatibility complex class II DQ alpha 1 (HLA-DQA1) expression, HLA class II histocompatibility antigen DRB1 beta chain (HLA-DRB1) expression, HLA class I histocompatibility antigen alpha chain E (HLA-E) expression, natural killer cell granule protein 7 (NKG7) expression, chemokine like receptor 1 (CMKLR1) expression, granzyme A (GZMA) expression, perforin-1 (PRF1) expression, cytotoxic T-lymphocyte-associated protein 4 (CTLA4) expression, programmed cell death protein 1 (PD1) expression, programmed death-ligand 1 (PDL1) expression, programmed cell death 1 ligand 2 (PDL2) expression, lymphocyte-activation gene 3 (LAG3) expression, T cell immunoreceptor with Ig and ITIM domains (TIGIT) expression, cluster of differentiation 276 (CD276) expression, chemokine (C-C motif) ligand 5 (CCL5) expression, cluster of differentiation 27 (CD27) expression, chemokine (C-X-C motif) ligand 9 (CXCL9) expression, C-X-C motif chemokine receptor 6 (CXCR6) expression, indoleamine 2,3-dioxygenase (IDO) expression, signal transducer and activator of transcription 1 (STAT1) expression, 3-fucosyl-N-acetyl-lactosamine (CD15) expression, interleukin-2 receptor alpha chain (CD25) expression, siglec-3 (CD33) expression, cluster of differentiation 39 (CD39) expression, cluster of differentiation (CD118) expression, and forkhead box P3 (FOXP3) expression;
wherein the machine learning classifier is selected from the group consisting of a neural network classifier, a support vector machine, a max entropy classifier, an extreme gradient boosting classifier, a random fern classifier, and a random forest classifier,
and wherein the machine learning classifier was trained on said total mutational burden, tumor infiltration, and factor expression of a plurality of training subjects having cancer and a responsiveness of each of said plurality of training subjects to cancer treatment to predict responsiveness of said non-training subject to said cancer treatment; and
ii) generating, using the machine-learning classifier, a responsiveness score for the non-training subject; and
b) administering, based on said obtaining, the cancer treatment to the non-training subject for responsiveness scores that equal or exceed a predetermined threshold of 5;
wherein the cancer treatment is a checkpoint inhibitor selected from an anti-CTLA4 drug, an anti-PD1 drug, an anti-PDL1 drug, and any combination of two or more of said drugs.
2. The method of claim 1 , wherein, for the non-training subject, for one or more of the plurality of training subjects, or for the non-training subject and one or more of the plurality of training subjects, at least some of the total mutational burden, at least some of the tumor infiltration, at least some of the factor expression, or any combination of two or more of the foregoing, comprise gene sets.
3. The method of claim 2 wherein the gene sets were selected using single sample gene set enrichment analysis.
4. The method of claim 1 , wherein at least one total mutational burden, at least one tumor infiltration, or at least one factor expression, comprises an identifier selected from a valence, an importance, and a weight, and further comprising visually reporting the identifier by displaying a graphical user interface on a display connected to a computer or computer system comprising one or more microprocessor and one or more memory wherein the identifier of the non-training subject is stored, wherein the graphical user interface reports identifiers as aspects of an annulus sector, wherein an angle of the annulus sector reports the importance, an outer radius of the annulus sector reports the weight, and a color of the annulus sector reports the valence.
5. The method of claim 4 wherein the identifier comprises an importance and the importance comprises a Gini index decrease.
6. The method of claim 5 wherein the graphical user interface reports the identifier if and only if the importance is above a threshold.
7. The method of claim 6 wherein the importance is not above the threshold if the square of the importance is not above 0.1.
8. The method of claim 6 , wherein each of the annulus sectors comprises an inner arc and the inner arcs of the annulus sectors are arranged to form a circle.
9. The method of claim 1 , further comprising inputting, or having input, to the trained machine learning classifier a responsiveness of the non-training subject to the cancer treatment and further training the machine learning classifier, wherein further training comprises training, or having trained, the trained machine learning classifier on one or more of total mutational burden, tumor infiltration, and factor expression of a tumor sample obtained from the non-training subject and a responsiveness of the non-training subject to the cancer treatment.
10. The method of claim 1 , wherein the cancer treatment is selected from an anti-CTLA4 antibody, an anti-PD1 antibody, an anti-PDL1 antibody, and any combination of two or more of the foregoing.
11. The method of claim 1 , wherein
total mutational burden comprises all mutations and nonsynonymous mutations.
12. The method of claim 11 , wherein the cancer treatment is selected from an anti-CTLA4 antibody, an anti-PD1 antibody, an anti-PDL1 antibody, and any combination of two or more of the foregoing.
13. The method of claim 1 , wherein
tumor infiltration comprises tumor infiltration by cells expressing cluster of different 8 (CD8), tumor infiltration by cells expressing cluster of differentiation 4 (CD4), and tumor infiltration by cells expressing cluster of differentiation 19 (CD19).
14. The method of claim 13 , wherein the cancer treatment is selected from an anti-CTLA4 antibody, an anti-PD1 antibody, an anti-PDL1 antibody, and any combination of two or more of the foregoing.
15. The method of claim 1 , wherein the machine learning classifier is a random forest classifier.
16. The method of claim 15 , wherein 50,000 trees are used in training the random forest classifier.