[0001] This application claims priority to our co-pending U.S. Provisional Application No. with serial number 62/753,687, filed on October 31, 2018, the entire content of which is herein incorporated by reference.

Field of the Invention

[0002] The present disclosure relates to methods for predicting a drug response using omics analysis of individuals with cancer, and more specifically, a response prediction for the treatment of metastatic colon cancer with temozolomide.

Background of the Invention

[0003] The background description includes information that may be useful in understanding the present invention. It is not an admission that any of the information provided herein is prior art or relevant to the presently claimed invention, or that any publication specifically or implicitly referenced is prior art.

[0004] All publications and patent applications herein are incorporated by reference to the same extent as if each individual publication or patent application were specifically and individually indicated to be incorporated by reference. Where a definition or use of a term in an incorporated reference is inconsistent or contrary to the definition of that term provided herein, the definition of that term provided herein applies and the definition of that term in the reference does not apply.

[0005] Temozolomide (TMZ) is a chemotherapeutic agent that is used as a standard chemotherapeutic treatment of glioblastoma and melanoma cancers. See, Hegi et al., 2005, NEJM, 352:997-1003. TMZ has recently shown limited but encouraging activity in patients with metastatic colorectal cancer (mCRC). See, Morano et al., 2018, Ann Oncol., 29:1800- 1806. TMZ is an agent with alkylating/methylating activity at the N-7 or 0-6 positions of guanine residues in DNA, often triggering cell death in sensitive cells. However, in response to DNA alkylation, the O-6-methylguanine-DNA methyltransferase (MGMT) enzyme catalyzes repair of DNA from alkylating agents such as TMZ. Not surprisingly, low MGMT protein expression is associated with TMZ sensitivity. Perplexingly however, some previous melanoma studies demonstrated that epigenetic silencing of MGMT is associated with TMZ resistance and TMZ resistance occurs in about one third of colorectal cancers. The relationship between various MGMT assays and outcomes remains unclear.

[0006] Commonly, MGMT expression status of tumors can be assessed by a digital polymerase chain reaction (PCR) method known as methyl-BEAMing (MB), and a cutoff of >60% CpG methylation of the MGMT promoter predicts a benefit from TMZ. In another approach, mass spectrometry (MS) proteomic analysis can objectively quantify the MGMT protein and other actionable protein biomarkers in formalin fixed, paraffin-embedded (FFPE) tissue sections. Here, a MGMT protein cutoff of 200 amol/ug (predefined based on the assay’s limit of detection) is predictive of benefit in mCRC patients treated with TMZ. MGMT protein quantity may also correlate with MGMT methylation status.

[0007] Among TMZ-treated patients with mCRC, those whose tumors expressed low or undetectable levels of MGMT protein had a longer progression- free survival (PFS) than their counterparts with higher MGMT protein levels. As presented in Abstract # 11601 at the ASCO Annual Meeting, Chicago, Illinois (June 2-6, 2017), a correlation of 80% was observed between MGMT protein expression as quantified by mass spectrometry and MGMT methylation status by MB. Quantitative proteomics objectively measured MGMT protein in FFPE tumor samples and retrospectively identified 9 of 9 responders to TMZ. Digital PCR methylation assay (methyl-BEAMing) retrospectively identified 7 of 8 responders to TMZ. The investigators therefore concluded that quantitative proteomic analysis of MGMT could potentially be used to select mCRC patients for TMZ therapy.

[0008] However, such approach considered proteomic and methylation analysis only in a retrospective manner. Moreover, detection limits of mass spectroscopic analysis and possible shortfalls of methylation detection further potentially reduce accuracy of the responder analysis. Indeed, it was observed that there was only an agreement of about 80% for MGMT status as determined by the mass spectroscopic analysis and methylation analysis. Moreover, the authors did not present any analytic option that would support or hint at a response prediction with clinically useful accuracy of prediction.

[0009] Finally, for glioblastoma patients, some studies have shown that treating MGMT unmethylated patients with TMZ can be detrimental to the GB patient. See, Hegi and Stupp, 2015, Neuro-Oncology, 17:1425-1427. Accordingly, even though various systems and methods for prediction of specific drug response are known in the art, there remains a need for clinically validated systems and methods that allow for simple and robust treatment prediction for a drug with high confidence, and that also allow prediction of the treatment response in a patient-specific manner.

Summary of the Invention

[0010] The inventive subject matter includes a highly accurate method of determining sensitivity or resistance and/or predicting a treatment response to temozolomide (TMZ) in a patient having a tumor that in some tumor environments is known to respond to TMZ therapy. In particular, the method includes obtaining multivariate data of 0-6- methylguanine-DNA methyltransferase (MGMT) from a sample of the tumor. Preferably the multivariate data of MGMT includes obtaining or providing MGMT promoter methylation status, RNA sequencing (RNA-seq) of MGMT transcripts, and MGMT protein

quantification.

[0011] The inventive method includes calculating the TMZ response prediction using a classification algorithm of the multivariate data of MGMT. For example, the data is analyzed using the K-nearest neighbor (kNN) classifier or alternatively, the Quad Discriminant classifier or Ridge Classifier.

[0012] In specific aspects, the MGMT promoter methylation status is predicted by obtaining whole RNA-seq analysis of MGMT and using Bayesian Ridge regression analysis to predict the MGMT promoter methylation status. Preferably, the MGMT promoter methylation data analysis includes a cutoff value of at least 60% methylation, or more preferably at least 63% methylation.

[0013] Notably, in addition to analyzing the MGMT promoter methylation data with a cutoff value, the method also includes analyzing the RNA-seq data of MGMT in a tumor with the analysis using a cutoff value of about 12 transcripts per million (TPM) or about 3.5 log2 TPM. Additionally, the analysis of the MGMT protein quantification data includes a cutoff value of 200 amol/ug. Preferably, the MGMT protein quantification is obtained using liquid chromatography-mass spectrometry (LC-MS) of a tumor sample.

[0014] The data is obtained from either fresh tumor samples from the patient or tumor tissue form the patient that has been formalin fixed, paraffin-embedded (FFPE). [0015] Advantageously, the method of more accurately predicting the TMZ response for a tumor may be applied to any type of tumor that has been reported to respond to TMZ therapy. For example, the tumor may be a glioma, a neoplasm, a melanoma, a lymphoma, or a colorectal tumor. More specifically, the tumor may be an anaplastic astrocytoma, a gliobastoma, a neuroendocrine neoplasm (NEN), a melanoma, a lymphoma, or a metastatic colorectal tumor.

[0016] Surprisingly, the inventive method results in a response model that is capable of predicting and/or determining sensitivity or resistance of a patient’s tumor to TMZ with an accuracy of at least 80%. As exemplified herein, the accuracy of the predictive method may be at least 83%.

Brief Description of The Drawing

[0017] Fig. 1 depicts exemplary results for predicting temozolomide (TMZ) response based on the combination of MGMT protein quantification, MGMT expression levels, and MGMT promotor methylation each having an Exploratory Cutoff (circle) or Predefined Cutoff (circle with x), as indicated.

[0018] Fig. 2 depicts exemplary results for temozolomide response using the indicated parameters and classification algorithms.

[0019] Fig. 3 is a schematic showing an analysis model in which the methylation is predicted based on regression analysis of whole RNA sequence (RNAseq) expression levels and the predicted methylation is combined with the MGMT proteomic and expression data to obtain a TMZ drug response prediction.

[0020] Fig. 4 is a heat map showing exemplary results for the response predictions on the 1,000 most variable genes across 44 samples using the indicated thresholds.

[0021] Fig. 5 is a listing of exemplary classification algorithms used on selected datasets and combination of datasets.

Detailed Description

[0022] The inventors have surprisingly discovered that a multivariate calculation increases the accuracy of predicting a response to temozolomide (TMZ) as a chemotherapy for treating tumors of some cancers. The inventors contemplate use of more than one type of dataset for a single molecular entity in the construction of a prediction model.

Advantageously, suitable types of datasets include DNA copy number data, DNA mutation data, RNA spice variant data, RNA expression level data, promoter methylation data, epigenetic modification data, protein data, and protein activity data. Preferred aspects of the inventive subject matter include obtaining at least 1) quantifying the MGMT protein levels,

2) quantifying the RNA expression levels of MGMT, and 3) MGMT promoter methylation status of the tumor.

[0023] For any selected combination of datasets, most typically, such data are readily available and/or can be inferred from various pathway models (e.g., PARADIGM). It is also contemplated that where more than one type of dataset is used, at least three different types of datasets will be employed. As will be readily appreciated, the choice of classification algorithm will be at least to some degree a function of the type of dataset, and the skilled person will be able to select an appropriate classification algorithm(s) for a given dataset. Moreover, and as exemplified below, cutoff (e.g., threshold) values may be predetermined, or independently learned using further machine learning. With reference to Fig. 1, as predetermined cutoff values resulted in a more accurate prediction of TMZ response, preferred aspects of the contemplated method include a predetermined cutoff value for at least one of the variables. More preferably, the methylation of MGMT of the tumor is one of the analyzed variables and is analyzed using a predetermined value of at least 60% methylation of the MGTM promoter. In more typical aspects, the predetermined cutoff value of methylation of the MGMT promoter is at least 61%, 62%, or 63%.

[0024] In addition to incorporating a predetermined value for MGMT methylation, preferred multivariate methods also include quantifying the protein levels of MGMT in the tumor, and the more preferred aspects include quantifying the proteins levels of MGMT with a predetermined cutoff value of at least 200 amol/ug (attomoles per microgram). Preferably, the MGMT protein quantification is obtained using liquid chromatography-mass

spectrometry (LC-MS) of a tumor sample.

[0025] In most preferred aspects, the contemplated method includes more accurately predicting TMZ responsiveness in an individual having a potential TMZ-treatable tumor by obtaining MGMT methylation data with a cutoff value (e.g., at least 60-63%), obtaining MGMT protein levels with a cutoff value (e.g., at least 200 amol/ug), and obtaining MGMT RNA sequencing data with a predetermined cutoff value of about 12 transcripts per million (TPM) or about 3.5 log2 TPM.

[0026] Considering TMZ treatment can be detrimental for patients with an unmethylated MGMT promoter, the presently disclosed multi- variate method for modeling a prediction of TMZ sensitivity can be used to more accurately assess (e.g., predict) the effectiveness of TMZ for any cancer treatable by TMZ. For example, the presently disclosed method may be used to more accurately predict TMZ sensitivity— and therefore, more accurately avoid treating a patient with unmethylated MGMT promoter— for a patient having a glioma (e.g., anaplastic astrocytoma, gliobastoma), neoplasm (e.g., neuroendocrine neoplasm (NEN)), melanoma, lymphoma, or colorectal (e.g., metastatic colorectal) tumor.

[0027] In one preferred aspect of the inventive subject matter, as discussed and exemplified herein, the inventors sought to train a robust predictive model of TMZ response based on 3 preferred quantitative MGMT assays— 1) promoter quantitative methylation, 2) RNA expression (RNAseq), and 3) protein abundance), with validation of this tri-variant calculation in unseen mCRC patients. Viewed from a different perspective, rather than identifying a single type of predictor, the inventors set out to identify multiple predictors in a machine learning setting to integrate various variables in order to obtain a prediction model for predicting a TMZ response with high sensitivity and accuracy. Accordingly, utilizing the multivariate analysis as disclosed herein, a highly accurate method for predicting a response to temozolomide (TMZ) therapy for a tumor may be predicted with an accuracy of at least 75%. As exemplified herein, a response to TMZ may be predicted (e.g., determined) with an accuracy of at least 80%, 81%, or 82%. More typically, a TMZ response is predicted with an accuracy of at least 83%. Preferably, a TMZ response is predicted with an accuracy of at least 84%, 85%, 86%, or 87%.

[0028] Notably, and due to a lack of clinical-grade methylation testing, preferred aspects of the inventive method include models that first predict MGMT methylation based on whole RNA sequence (MGMT RNA-seq) analysis using Bayesian Ridge regression analysis, with this predicted methylation subsequently used to classify the TMZ response. Bayesian Ridge estimates the probabilistic model as p(y\ X, w,oc)= M VIXR\ a), where output y is assumed to be Gaussian distributed around Xw. The prior of the coefficient w is given by a spherical gaussian: [0029] To exemplify the disclosed method, 41 archived tumor samples from 3 TMZ safety trials (INT Study h.20/13; INT Study 20/13 & EudraCT 2012-002766-13) were used to train models. Response to TMZ was defined by RECIST v.l.l criteria for analyzing the progression/regression of solid tumors. MGMT status was assessed by 3 assays— 1) digital PCR/methyl-BEAMing (MB), 2) RNAseq, and 3) liquid chromatography-mass spectrometry (LC-MS). Several multivariate modeling strategies were evaluated using cross-validation (CV) within the training set. The most accurate model in CV was validated in 14 unseen tumor samples from a follow-up study that were similarly assayed. Predefined cutoff values (e.g., thresholds) in each MGMT assay were used as the basis for comparison. Results of these 3 assays used in combination and separately and with or without predetermined cutoff values are shown in Fig. 2. Notably, for accurately measuring MGMT status a combination of these assays using predetermined (e.g., predefined) values provided the highest accuracy, as indicated. Furthermore, as indicated, the k-nearest neighbor (kNN) classification algorithm rendered this preferred accuracy of 87%. Other preferred classification algorithms include Quad Discriminant and Ridge Classifier as shown in Fig. 2.

[0030] As can be seen from the exemplary results in Table 1 below, when multiple variables were used to train and validate, response prediction to TMZ significantly improved as compared to single variables (/. _<? ., methylation or protein or expression used individually). Indeed, based on the integration of multiple variables, TMZ response in refractory mCRC is approximately predictable. Combining predicted methylation, transcript levels, and protein abundance, yields the most accurate and robust method of predicting response. In typical results, a calculated analysis of TMZ responsiveness is predicted with more than 80% accuracy— e.g., 82% to 87% accuracy.

[0031] Table 1. MGMT Assays and Accuracy of TMZ Prediction

Examples

[0032] In one set of experiments, the inventors investigated the training cohort prediction performance for MGMT protein (as measured by LC-MS), MGMT expression (as measured by RNA sequencing (RNA-seq) in transcripts per million (TPM)), and MGMT promotor methylation (as measured by digital PCR/methyl-BEAMing (MB)). More specifically, to evaluate the ability of the predefined cutoffs to predict response to TMZ, the leave pair out cross validation strategy was used. Predefined and exploratory cutoffs were assessed in unseen samples 330, 308, and 250 times in LC-MS, RNAseq, and MB data respectively. The predefined cutoffs in LC-MS and RNA-seq showed better mean predictive performance (82.1% and 72.2%, respectively) than the MB model (68.0%), and a typical result is depicted in Fig. 1.

[0033] To further investigate the influence of various classification algorithms and training data (/. _<? ., single variable versus multiple variables), the inventors performed several learning approaches in which protein, RNA-Seq, and methylation data were used, alone or in combination, and with or without a predefined threshold. As can be seen from Fig. 2, protein-based models had relatively high prediction accuracy, which was even further superseded by a model that used all three variables. In still further attempts to improve accuracy and simplify clinical or sample requirements, the inventors used previous TMZ studies as training data to build 10 candidate models (+3 predefined cutoffs) and replaced measured methylation with‘predicted methylation’ based on whole RNAseq and using a regression model. Performance was then tested in an unseen testing cohort receiving the combination treatment of TMZ and irinotecan (TEMIRI) exemplified in Fig. 3. Here, the training dataset was a TMZ cohort and included 41 mCRC patients treated with TMZ from 3 phase II studies. Continuous MGMT protein levels by LC-MS, as well as RNA expression data by RNA seq and continuous MGMT methylation percentage data were available for 33 patients. Drug response was noted as binary drug response data. The testing dataset comprised 32 mCRC patients treated with TMZ + irinotecan. Binary drug response data were missing for 3 patients, gene expression values were available for 14 patients, and MGMT protein expression data were available for 21 patients, with gene expression values and protein expression data available in a total of 11 patients. See Table 2. [0034] Table 2.

[0035] Fig. 4 depicts a heat map with exemplary results for the response predictions on the 1,000 most variable genes across 44 samples using preset thresholds as noted, and Fig. 5 is a listing of exemplary classification algorithms used on selected datasets and combination of datasets. As can be seen once more, use of MGMT RNAseq, MGMT protein, and MGMT promotor methylation provided superb training and testing accuracy for response prediction for temozolomide. Likewise, sensitivity, specificity, and Fl score were all substantially increased over other classifiers and individual datasets. Sample level predictions for the best model of Fig. 4 are listed in the Table 3 below.

[0036] Table 3. Sample Level Predictions for Fig. 4.

SAMPLE TRUE LABEL PREDICTED LABEL

G3183 RESISTANT RESISTANT

G3186 RESISTANT RESISTANT

G3187 RESISTANT RESISTANT

G3189 RESISTANT RESISTANT

G3191 RESISTANT SENSITIVE

G3192 RESISTANT RESISTANT

G3196 RESISTANT RESISTANT

G3200 RESISTANT RESISTANT G3414 SENSITIVE SENSITIVE

G3415 RESISTANT RESISTANT

G3198 RESISTANT SENSITIVE

[0037] With reference to Fig. 5, predicting MGMT promoter methylation in a tumor as disclosed herein, is an advantageously accurate method. This methylation prediction uses Bayesian ridge regression module from Scikit-learn library in Python for predicting the methylation values of transcripts in a tumor sample. Bayesian ridge regression is a technique that is used to include 12 regularization parameters by finding the maximum posterior estimation under a gaussian prior over the coefficient w. BayesianRidge estimates the probabilistic model as p(y\ X, w,oc)= M VIXR\ a), where output y is assumed to be gaussian distributed around Xw. The prior of the coefficient w is given by a spherical gaussian:

[0038] It should be apparent to those skilled in the art that many more modifications besides those already described are possible without departing from the inventive concepts herein. The inventive subject matter, therefore, is not to be restricted except in the spirit of the appended claims. Moreover, in interpreting both the specification and the claims, all terms should be interpreted in the broadest possible manner consistent with the context. In particular, the terms“comprises” and“comprising” should be interpreted as referring to elements, components, or steps in a non-exclusive manner, indicating that the referenced elements, components, or steps may be present, or utilized, or combined with other elements, components, or steps that are not expressly referenced. Where the specification claims refers to at least one of something selected from the group consisting of A, B, C..., and N, the text should be interpreted as requiring only one element from the group, not A plus N, or B plus N, etc. Moreover, as used in the description herein and throughout the claims that follow, the meaning of“a,”“an,” and“the” includes plural reference unless the context clearly dictates otherwise. Also, as used in the description herein, the meaning of“in” includes“in” and “on” unless the context clearly dictates otherwise.