Embodiments of the present disclosure generally relate to systems and methods for predicting response to cancer therapy (either in terms of survival rates or in terms of tumor response as measured by standard Response Evaluation Criteria in Solid Tumors (RECIST) criteria) in subjects or populations affected by a disease or disorder, and more specifically for predicting components of genetic interactions, which may be used to predict the likelihood of a subject to respond to a therapy for treatment of the disease or disorder and/or predict improved therapies for treatment of the disease or disorder. The present disclosure also relates to methods of determining the sensitivity of a cancer to an anti-cancer therapy, and methods of treating such cancer.
One aspect of the present disclosure relates to a method for identifying a cancer therapy for a patient. The method may include the operations of accessing one or more databases storing information associated with genetic interactions to obtain a plurality of candidate synthetic lethality (SL) gene partners for a cancer therapy and identifying, based on experimental functional screens, patients' omics and survival data and phylogenetic profile information of each of the plurality of candidate SL gene partners, a subset of the plurality of candidate SL gene partners as potential predictive biomarkers for the cancer therapy. The method may further include the operations of comparing the subset of the plurality of candidate SL gene partners to a cancer-inhibiting drug dataset to filter the subset of the plurality of candidate SL gene partners and identifying, based on a transcriptomics profile of a tumor of the patient and the filtered subset of the plurality of candidate SL gene partners, the cancer therapy for the patient.
Another aspect of the present disclosure relates to a method for identifying a cancer therapy for a patient including accessing one or more databases storing genetic interaction information to obtain a plurality of candidate synthetic rescue (SR) gene partners for a cancer therapy and identifying, based on patients' omics and survival data and phylogenetic profile information, in particular at least one of (1) a product of PD1 and PDL1 activity (e.g., gene expression levels), (2) CTLA4 activity (e.g., protein expression levels), and (3) molecular profiles (including but limited to gene expression levels and somatic copy number alterations (SCNA)) of each of a plurality of candidate SR gene partners, a subset of candidate SR gene partners as potential predictive biomarkers for the cancer therapy. The method may also include ranking the subset of the plurality of candidate SR gene partners based on phylogenetic distance information to filter the subset of the plurality of candidate SL gene partners, filtering, based on an identification of candidate SR gene partners in which a downregulation (or upregulation) of a partner rescuer gene occurs, the ranked subset of the plurality of candidate SR gene partners, and identifying, based on a transcriptomics profile of a tumor of the patient and the filtered subset of the plurality of candidate SR gene partners, a cancer therapy for the patient.
Still another aspect of the present disclosure relates to a system for identifying a cancer therapy for a patient. The system may include a processor and a tangible storage medium storing instructions that are executed by the processor to perform the above operations.
Still another aspect of the present disclosure relates to a method of treating a cancer patient comprising administering a cancer therapy to a patient in need thereof identified according to the methods or operations described above.
Still other aspects of the present disclosure relate to assigning a score to each of the plurality of candidate SL or SR gene partners based on patient response data to each of the plurality of candidate SL or SR gene partners and filtering, based on the assigned scores, the plurality of candidate SL or SR gene partners to identify the subset of the plurality of candidate SL or SR gene partners. Ranking, based on the assigned scores, the plurality of candidate SL or SR gene partners is also contemplated, wherein the subset of the plurality of candidate SL gene partners comprises a subset of the plurality of candidate SL or SR gene partners with the highest assigned scores. Further, the subset of the plurality of candidate SL gene partners may comprise 25 gene partners (for targeted therapies) or the plurality of candidate SR gene partners may comprise 10 gene partners (for checkpoint therapy). These set size parameters and the interaction ranking schemes can be modified and improved as more datasets become available in the future. The transcriptomics profile may also include at least one of a proliferation measurement value, a cytolytic value, or a target gene expression identification level.
There have been significant advances in precision oncology, with an increasing adoption of sequencing tests that identify targetable mutations in cancer driver genes. Aiming to complement these efforts by considering genome-wide tumor alterations at additional “-omics” layers, recent studies have begun to explore the utilization of transcriptomics data to guide cancer patients' treatment. These studies have reported encouraging results, testifying to the potential of such approaches to complement mutation panels and increase the likelihood that patients will benefit from genomics-guided, precision treatments. However, current approaches have a heuristic exploratory nature, raising the need for developing and testing new systematic approaches for utilizing tumor transcriptomics data.
One approach aims to utilize the rapidly accumulating data obtained from cancer clinical samples. One of the key objectives in this approach is to systematically map between the genomic and molecular characteristics of tumors and their responses to various drugs. One way by which to tackle this and realize the potential of cancer pharmacogenomics is based on the concept of Synthetic lethal interactions (SLi). SLi describe the relationship between two genes whereby an individual inactivation of either gene results in a viable phenotype, while their combined inactivation is lethal. SLi have been considered as a potential basis for developing selective anticancer drugs. Such drugs are aimed at inhibiting the Synthetic Lethal (SL) partner of a gene that is inactive in the cancer cells. Indeed, as 90% or more of cancer predisposing mutations result in a loss of protein function, by identifying SLi these genomic alterations can be exploited for developing and improving cancer treatments.
It is with these observations in mind, among others, that various aspects of the present disclosure were conceived and developed.
Aspects of the present disclosure involve systems, devices, apparatus, methods, and the like, for a precision oncology framework for synthetic lethality and rescue-mediated precision oncology via the transcriptome (SELECT). The framework is generally aimed at selecting drugs or other treatments for a given patient based on the transcriptome of the patient's tumor, which may be the entire tumor transcriptome. More particularly, the presented approach is based on identifying and utilizing the broader scope of genetic interactions (GIs) of drug targets, which provide biologically testable biomarkers for therapy response prediction. Two types of GIs that are highly relevant to predicting the response to cancer therapies are considered: (1) synthetic lethal (SL) interactions, which describe the relationship between two genes whose concomitant inactivation, but not their individual inactivation, reduces cell viability (e.g., an SL interaction that is widely used in the clinic is of poly (ADP-ribose) polymerase (PARP) inhibitors on the background of disrupted DNA repair); and (2) synthetic rescue (SR) interactions, which denote a type of genetic interactions where a change in the activity of one gene reduces the cell's fitness but an alteration of another gene's activity (termed its SR partner) rescues cell viability (e.g., the rescue of Myc alterations by B-cell lymphoma 2 (BCL2) activation in lymphomas. These are relevant because when a gene is targeted by a small molecule inhibitor or an antibody, the tumor may respond by up or down regulating its rescuer gene(s), conferring resistance to therapies. The inventors have discovered that a patient's response to a cancer therapy can be predicted by analyzing SL interactions, SR interactions, or a combination thereof.
The SELECT framework comprises two basic steps: (A) For each drug whose response is to be predicted, the clinically relevant pan-cancer GIs (the interactions found to be shared across many cancer types) of the drug's target genes is identified and (B) the identified SL/SR partners of the drug emerging from step (A) is used to predict a given patient's response to a given treatment based on her/his tumor's gene expression. The operations from which SELECT differences from previous frameworks may include:
Based on these methodological innovations, the application of SELECT to predict the response of cancer patients to a broad array of targeted and immunotherapy cancer drugs has resulted in the generation of an array of new SL and SR based biomarker signatures, for the first time. From a conceptual perspective, SELECT is shown to be the first systematic transcriptomics-based precision oncology framework that can successfully prioritize effective therapeutic options for cancer patients across many different treatments and cancer types, a much desired outcome that the previously published frameworks have fallen short of.
In one instance, transcriptomic profiles and treatment outcome information of various clinical trials may be obtained from public databases, such as a github repository. In the instances described below, information or data from a repertoire of 45 clinical trials spanning about 4,000 patients from 12 different cancer types was obtained and analyzed. In particular, cancer patient pre-treatment transcriptomics profiles may be collected together with therapy response information from numerous publicly available databases, surveying Gene Expression Omnibus (GEO), ArrayExpress and the literature, and a new unpublished cohort of anti-PD1 treatment in lung adenocarcinoma. Overall, 45 such datasets were found that includes both transcriptomics and clinical response data, spanning 12 chemotherapy, 12 targeted therapy and 21 immunotherapy datasets across 12 different cancer types.
To identify the SL and SR partners of cancer drugs, two computational pipelines may be utilized which identify genetic dependencies that are supported by multiple layers of omics data, including in vitro functional screens, patient tumor DNA and RNA sequencing data, and phylogenetic profile similarity across multiple species. The SELECT framework determines whether genetic dependencies inferred from multi-omics tumor data can be used to determine efficacious therapeutics for individual cancer patients. As such, the SELECT framework is a first of its kind systematic approach for robustly predicting clinical response to chemo, targeted and immune therapies across tens of different treatments and cancer types, offering a new way to complement existing mutation-based approaches.
Several of the operations described above or throughout this disclosure may include information obtained via the systems and methods described in United States Patent Application Publication No. 20190024173, entitled COMPUTER SYSTEM AND METHODS FOR HARNESSING SYNTHETIC RESCUES AND APPLICATIONS THEREOF, United States Patent Application Publication No. 20170154163, entitled CLINICALLY RELEVANT SYNTHETIC LETHALITY BASED METHOD AND SYSTEM FOR CANCER PROGNOSIS AND THERAPY, and/or United States Application Publication No. 20160300010, entitled METHOD AND SYSTEM FOR PREDICTING SELECTIVE CANCER DRUG TARGETS, the entirety of all of which are incorporated by reference herein.
In general, the SL/SR partners are inferred once analyzing DepMap and/or TCGA cohorts and their size set was optimized by training on single clinical trial dataset, prior to their application to a large collection of other test clinical trial datasets. In other words, the transcriptomic profiles and treatment outcome data available are not used in the SL and SR inference. The treatment outcomes of the selected profiles and treatment outcomes may be used to evaluate the resulting post-inference prediction accuracy in operations 112-116. Throughout the analysis, the same fixed sets of parameters in making the predictions for targeted and immunotherapies may be used. Taken together, these procedures markedly reduce the well-known risk of obtaining over-fitted predictors that would fail to predict on datasets other than those on which they were originally built.
To identify clinically relevant SL interactions for targeted therapies, a three-step procedure may be executed, such as that disclosed in Harnessing Synthetic Lethality to Predict the Response to Cancer Treatment, Nat Commun 9, 2546 to Lee, J. S., Das, A., Jerby-Arnon, L., Arafeh, R., Auslander, N., Davidson, M., McGarry, L., James, D., Amzallag, A., Park, S. G., et al. (2018), the entirety of which is hereby incorporated by reference. The procedure may include (1) creating an initial pool of SL pairs identified in cell lines via RNAi/CRISPR-Cas9 (as outlined in Meyers et al., 2017 and Tsherniak et al. (2017) or pharmacological screens (as outlined in An Interactive Resource to Identify Cancer Genetic and Lineage Dependencies Targeted by Small Molecules, Cell 154, 1151-1161 to Basu, A., Bodycombe, N. E., Cheah, J. H., Price, E. V., Liu, K., Schaefer, G. I., Ebright, R. Y., Stewart, M. L., Ito, D., Wang, S., et al. (2013), the contents of which are incorporated herein by reference. For drug target gene T and candidate SL partner gene P, growth reduction induced by knocking out/down gene T or pharmacologically inhibiting gene T is stronger when gene P is inactive is checked, via a Wilcoxon ranksum test. (2) Second, among the candidate gene pairs from the first step, gene pairs are selected whose co-inactivation is associated with better prognosis in patients, using a Cox proportional hazard model, testifying that they may thus hamper tumor progression. (3) SL paired genes with similar phylogenetic profiles across different species may be prioritized. Because of the distinct distribution of P-values for the first two screens, a false discovery correction may be performed with 1% for the in vitro screen (1st step) and 10% for the tumor screen (2nd step).
Through these operations, identification of significant SL partners with these False Discovery Rate (FDR) thresholds for most of the datasets is made. However, for the cases in which significant SL partners with this set of FDR thresholds are not found, the FDR thresholds may be identified in a two step manner; by relaxing the FDR for the in vitro screen to 5% while keeping the FDR for tumor screen at 10%, or further relaxing both FDRs to 20%. If no significant pairs are identified even with 20% FDR, the corresponding drug may be identified as non-predictable by the instant approach.
The number of SL partners that pass FDR ranges from 50 to 1,000 may depend on the drugs and specific FDR thresholds. Accordingly, SL partners may be filtered to generate a small set that is used to make the drug response predictions. This further filtering has been motivated by the following three reasons: (1) Occam's razor (regularization): predictor with a smaller number of variables are likely to generalize better; (2) biomarker interpretability: small sets of partners are more relevant for clinical use as predictive biomarkers; and (3) patient cohort analysis: when comparing the SL-scores of different drugs to decide which would be a best fit for a given patient, using the same number of top predictors facilitates such an analysis on equal grounds. To determine the top significant SL partners, a limited training on a data set may be conducted. In one example, the data set is the BRAF inhibitor dataset (GSE50509). For example,
To identify GIs for immunotherapy, a general GI inference pipeline may be altered to incorporate the characteristics of immune checkpoint therapy (as disclosed in Lee et al., 2018 and Genome-wide prediction of synthetic rescue mediators of resistance to targeted and immunotherapy, Mol Syst Biol 15, e8323 to Sahu, A. D., J, S. L., Wang, Z., Zhang, G., Iglesias-Bartolome, R., Tian, T., Wei, Z., Miao, B., Nair, N. U., Ponomarova, O., et al. (2019)). In one example, for anti-PD1/PDL1 therapy, where the antibody blocks the physical interaction between PD1 and PDL1, the interaction term (i.e. the product of PD1 and PDL1 gene expression values) may be considered to identify the SR partners of the treatment. For anti-PD1/PDL1 therapy, gene expression may be utilized or analyzed, rather than protein expression, as protein expression of PD1 and PDL1 may not be available for many samples. In another example, for anti-CTLA4 therapy, where the precise mechanism of action involves several ligand/receptors interactions, the CTLA4 itself may be considered, using its protein expression levels (available via reverse phase protein lysate microarray (RPPA) values in TCGA data) as they are likely to better reflect the activity than the mRNA levels. (2) For the GI partner levels, gene expression and somatic copy number alterations (SCNA) data may be utilized as protein expression may be measured only for a small subset of genes. (3) Instead of considering all protein coding genes as candidates for SR partners, the genes that are covered by the NanoString panel may be considered because (i) the gene expression of many of ICI datasets was quantified by NanoString platform and (ii) NanoString panel is enriched with immune system related genes that are highly relevant to the response to immune checkpoint therapy. (4) The first step of the SL/SR inference procedure, which is aimed at identifying candidate genetic interactions from the cell line functional screening data, may be omitted because these interactions are not relevant to immune checkpoint response. In some instances, the genome-wide CRISPR screens in cancer cell/T-cell co-culture may be used, but this data is limited to melanoma and the coverage is not fully genome-wide, where many genes included in the NanoString panel are missing. (5) The mediators of resistance to immune checkpoint therapies using synthetic rescue (SR) interactions, as no statistically significant SL interaction partners may be identified for either PD1 or CTLA4, may also be used. False discovery correction was done with FDR 10%.
To determine the top significant SR partners, a limited training may be conducted on a dataset, as illustrated in the graph
To predict drug response in patients using SL/SR partners, one or more of the identified SL or SR partners for drug response prediction may be analyzed. In particular, an SL-score for chemotherapy and targeted therapy may be defined as the fraction of inactive SL partners in a given sample out of all SL partners of that drug following the notion that an inhibitor would be more effective when a larger number of its drug target genes' SL partners are inactive. The SL score reflects the intuitive notion that inhibiting targeted drug would be more effective when a larger fraction of its SL partners is inactive in the tumor. In each patient drug response dataset, a gene is determined to be inactive if its expression is below bottom tertile across samples in the same dataset. This normalization may be performed (i) to account for the basal expression level of each gene in specific tumor type and (ii) to minimize batch effect occurring when different datasets are combined. Additionally, the SL-score may be multiplied by a target gene factor to obtain the final SL score. This has been motivated by the notion that an inhibitor will be not effective when its target gene is not expressed; thus, the target gene factor may be set to be zero when the target gene is inactive (below bottom 30-percentile in the given sample), and mean expression of the targets genes may be used when the given drug has more than one target gene. SR-scores may be used to predict response to immunotherapy, which quantifies the fraction f, SR partners that are inactive, and 1-f as the SR-score to predict responders. Higher SL- or SR-score is generally predictive of response to therapies.
Using the computed SL/SR-scores, either the classification problem to predict responders may be solved or Kaplan-Meier analysis may be performed to predict patient survival, depending on the availability of the data. For the datasets where the response information is available in the form of RECIST criteria, solving the classification problem may be performed. For the cases where progression-free survival time is available for all patients (with no censoring event), the median progression-free survival from the relevant literature as the cutoff to distinguish the responders from non-responders may be used to solve a classification problem. For the datasets where only overall or progression-free survival with censoring information are available, Kaplan-Meier analysis may be performed.
In still other instances, TCGA anti-PD1 coverage analysis for predicting the cancer type-specific response to checkpoint therapy may be performed. The objective response rates of anti-PD1 therapy in each cancer type in TCGA may thus be predicted via the SR interaction partners of PD1 identified above. The SR scores in each tumor sample in the TCGA compendium may be computed, based on its transcriptomics profiles following the above definition of SR-score, and labeled it as responder or non-responder accordingly using the point of maximal F1-score as threshold across all 9 immune checkpoint datasets, where the SR-score is predictive. Using this fixed cut-off, the fractions of responders for each cancer type may be computed and compared with the actual response rates reported in anti-PD1 clinical trials for 16 cancer types where the data is available using Spearman rank correlation. In each patient drug response datasets, a gene is determined to be inactive if its expression is below bottom tertile across samples in the same dataset following the previous studies.
Provided herein are methods of determining the susceptibility and/or sensitivity of a cancer to a particular anti-cancer therapy, and applications thereof for treating a cancer in a subject. The subject may be a human patient in need of anti-cancer therapy. The method may comprise determining the SL-score of a subject's cancer sample for the anti-cancer therapy, which may be indicative of the sensitivity of the subject's cancer to the anti-cancer therapy. The method may also comprise administering the anti-cancer therapy to the subject based on the SL-score for the anti-cancer therapy. In one example, a SL-score >0.44 indicates that the subject's cancer is sensitive to the anti-cancer therapy.
An anti-cancer therapy may comprise a drug or drug combination listed in Table 1, and the SL partner genes indicative of the sensitivity of the subject's cancer to the anti-cancer therapy may comprise the group of genes listed in Table 1 that are associated with the anti-cancer therapy. The SL partner genes for an anti-cancer therapy used to determine the SL-score of a subject's cancer may also consist of the SL partner genes listed in Table 1. In one example, the anti-cancer therapy is Vemurafenib, and the SL partner genes comprise FSCN1, ICA1, PMS2, C1GALT1, MMD2, C7orf28B, NT5C3, NDUFA4, RAPGEF5, TMEM106B, ADCYAP1R1, SCIN, NEUROD6, RP9, FAM126A, KLHL7, SKAP2, TRA2A, JAZF1, CBX3, BBS9, SP8, MACC1, GGCT, and TAX1BP1A. In another example, the anti-cancer therapy is Tamoxifen, and the SL partner genes comprise LTBP2, GADL1, CRISP2, SLC13A5, PCDHGA7, NLRP10, AAK1, IL22RA2, RASGRF1, FAM19A3, TPM2, UBR4, LRRFIP1, FOXL1, PCDHGA2, MAMSTR, ABCG4, FBXO32, DSG3, FER, ALPP, PINX1, AVPR1A, LHX6, and PHLPP2. Other examples are provided in Table 1.
Also provided herein are methods of determining the susceptibility and/or sensitivity of a cancer when the anti-cancer therapy is a checkpoint therapy, and applications thereof for treating a cancer in a subject. The subject may be a human patient in need of checkpoint therapy. The method may comprise determining the SR-score of a subject's cancer sample for the checkpoint therapy, which may be indicative of the sensitivity of the subject's cancer to the checkpoint therapy. The method may also comprise administering the checkpoint therapy to the subject based on the SR-score for the checkpoint therapy. In one example, a SR-score ≥0.9 indicates that the subject's cancer is sensitive to the checkpoint therapy.
The checkpoint therapy may be a PD1/PDL1 inhibitor or an anti-CTLA4 therapy. The PD1/PDL1 inhibitor may any such inhibitor known in the art, and may be Pembrolizumab, Nivolumab, Cemiplimab, Atezolizumab, Avelumab, or Durvalumab. For the PD1/PDL1 inhibitor, the SR partner genes used to determine the SR-score may comprise CXCL16, IL15RA, CD27, TNFRSF13C, TNFRSF13B, ICAM4, CD8A, CD4, LTBR, and IFITM2. The anti-CTLA therapy may be any such therapy known in the art, and may be Ipilimumab or tremelimumab. For the anti-CTLA4 therapy, the SR partner genes used to determine the SR-score may comprise CD44, IL22RA2, THBD, BID, F12, CCL13, EWSR1, CD274, IL22RA1, and CDKN1A.
Some of the types of cancers that can be treated by chemo-, targeted- or immuno-checkpoint therapies disclosed herein are known in the art, but a subject's cancer may be sensitive to an anti-cancer or checkpoint therapy even if the therapy has not received regulatory approval for treating the cancer, or has not previously been recognized as being effective against the type of cancer the subject has. Thus, a SL- or SR-score for the subject's cancer may be useful for identifying new types of cancers that are sensitive to the anti-cancer therapy or checkpoint therapy.
The SL- or SR-score may be determined according to a method described herein. In one example, for a given anti-cancer therapy, expression levels of the SL or SR partner genes may be provided from a sample of the subject's cancer, and from each of a plurality of reference cancer samples. The number of the SL or SR partner genes that are downregulated in the subject's cancer sample as compared to expression levels in the reference cancer samples may be counted. In one example, a SL or SR partner gene expressed in the subject's cancer sample may be downregulated if its expression levels are in the bottom half, tertile, quartile, or quintile of expression levels of that SL or SR partner gene as measured among the reference cancer samples. In one example, a SL or SR partner gene is downregulated in the subject's cancer sample if the expression level of the SL or SR partner gene is in the bottom tertile of expression levels of the SL or SR partner gene among the reference cancer samples.
To determine a SL-score for an anti-cancer therapy, the number of the SL partner genes that are downregulated in the subject's cancer sample may be divided by the total number of SL partner genes associated with the anti-cancer therapy. In one example, a SL-score >0.44 (for example, where at least 11 of 25 SL partner genes are downregulated in the subject's cancer sample as compared to the reference cancer samples) indicates that the subject's cancer is sensitive to the anti-cancer therapy.
To determine a SR-score for a checkpoint therapy, the number of the SR partner genes that are downregulated in the subject's cancer sample may be divided by the total number of SR partner genes associated with the checkpoint therapy. The result of that calculation may then be subtracted from 1 to determine the SR-score. In one example, a SR-score indicates that the subject's cancer is sensitive to the checkpoint therapy.
A cancer sample referred to herein may be any type of sample known in the art, but may in particular comprise a bulk tumor biopsy. The reference cancer samples may be of the same type of cancer as the subject's cancer. If the subject's cancer type is unknown, then the reference cancer samples may comprise one or more types of cancer that are different from the subject's, and in one example may comprise all cancer samples from a source of SL or SR partner gene expression levels.
The SL or SR partner gene expression levels may be measured from RNA-sequencing (RNAseq) or a microarray data. The gene expression levels may be normalized. In one example, the same normalization method may be used for SL or SR partner gene expression levels of the subject's cancer sample and of the reference cancer samples. The normalization method may be Reads per Kilobase per Million mapped reads (RPKM)/RNAseq by Expectation-Maximization (RSEM), which may be particularly useful when gene expression levels are measured using RNAseq. The SL or SR partner gene expression levels of the reference cancer samples may be provided from any source of data known in the art. The data source may be a database, and may be the Cancer Genome Atlas (TCGA), which is available at www.cancer.gov/tcga (the contents of which are incorporated herein by reference). The cancer may be any cancer known in the art, and may be one described in the TCGA.
Using the above operations and methods using SL-scores, results from several data analyses have been conducted and provided herein. For example,
SL based prediction accuracy levels are overall higher compared to those obtained by several published transcriptomic based predictors, including the proliferation index, IFNg signature, cytolytic score, or the expression of the drug target gene itself (BRAF in this case).
It is noted that patients with higher SL-scores showed significantly better treatment outcome in terms of progression-free survival in one of the datasets analyzed above where this data was available to us, as shown in the Kaplan-Meier curves of
The graph of
In addition to the SL approach, the above operations may also be used for SR-based prediction of response to a therapy or drug. For example, the ability of the SELECT framework to predict clinical response to checkpoint inhibitors is conducted and discussed herein. In particular, to identify the SR interaction partners that are predictive of the response to anti-PD1/PDL1 and anti-CTLA4 therapy, the published pipelines may be modified to take into account the characteristics of immune checkpoint therapy. For anti-PD1/PDL1 therapy, where the antibody blocks the physical interaction between PD1 and PDL1, consideration of the interaction term (i.e. the product of PD1 and PDL1 gene expression values) to identify the SR partners of the treatment may be used. For anti-CTLA4 therapy, where the precise mechanism of action involves several ligand/receptor interactions, focus may be on the CTLA4 itself, using its protein expression levels as they are likely to better reflect the activity than the mRNA levels Using this immune-tailored version of the framework, analysis of the TCGA data to identify the SL and SR partners of PD1/PDL1 and of CTLA4 is performed. In general, SR interactions denote such genetic interactions where inactivation of the target gene is compensated by downregulation (or upregulation) of the partner rescuer gene. Given a drug and tumor transcriptomics data from an individual patient, the fraction f of SR partners that are downregulated (or upregulated) may be quantified. Definition of 1−f as the SR-score may be assumed, where tumors with higher SR scores have less “active” rescuers are hence expected to respond better to the given checkpoint therapy.
To evaluate the accuracy of SR-based predictions, a collected set of 21 immune checkpoint therapy datasets, comprising 1050 patients, may be gathered that includes both pre-treatment transcriptomics data and therapy response information (either by RECIST or Progression-Free Survival (PFS)). Tumor types represented in these datasets include melanoma, non-small cell lung cancer, renal cell carcinoma, metastatic gastric cancer, and urothelial carcinoma cohorts treated with anti-PD1/PDL1 or anti-CTLA4-, or their combination.
The graph of
To study if tumor-specific SR scores can explain the variability observed in the objective response rates (ORR) of different tumor types to immune checkpoint therapy, the SR-scores for anti-PD1 therapy for each tumor sample in the TCGA may be computed. Based on the latter and the threshold for determining responders, the fraction of predicted responders in each cancer type in the TCGA cohort may be computed. A comparison of these predicted fractions to the actual ORR may be collected from anti-PD1 clinical trials of 16 cancer types. Notably, these two measures significantly correlate, demonstrating that SR-scores are effective predictors of ORR to checkpoint therapy in aggregate across different cancer types.
Summed up the three classes of the drugs studied, the genetic interaction-based approach achieves an AUC greater than 0.7 predictive performance levels in 24 out of 35 datasets containing RECIST response information, spanning 3 out of 12 non-targeted cytotoxic agents, 6 out of 8 targeted therapies and 15 out of 18 immunotherapy cohorts (including our new SMC dataset). More particularly,
Still another evaluation of the SELECT approach may be conducted utilizing a multi-arm basket clinical trial setting that incorporates transcriptomics data for cancer therapy in adult patients with advanced solid tumors. This multi-center study may include an arm recommending treatment based on actionable mutations in a panel of cancer driver genes and another based on the patients' transcriptomics data. In the evaluation performed, consideration of gene expression data of 71 patients with 50 different targeted treatments (single or combinations) for which significant SL partners were identified may be processed. Of the patient data, one patient had a complete response, 7 had a partial response and 11 were reported to have stable disease (labeled as responders), while 52 had progressive disease (labeled as non-responders).
Applying the SELECT approach discussed above, SL partners for each of the drugs prescribed in the study may be first identified. Confirmation that the resulting SL-scores of the therapies used in the trial as significantly higher in responders than non-responders is illustrated in the graph of
To illustrate the potential future application of SELECT for patient stratification, we describe here two individual cases arising in the trial data analysis. The first involves an 82-year-old male neuroendocrine cancer patient who was treated with everolimus because of an PIK3CA overexpression, and the patient indeed responded to the therapy. SELECT also recommends the treatment of everolimus, as shown in
Samples that display a strong SL vulnerability to one drug tend to have SL-mediated vulnerabilities to many other targeted agents, indicating that SL-based treatment opportunities may actually increase in advanced tumors. Reassuringly, an SL-based drug coverage analysis in another independent transcriptomics-based trials dataset from the Tempus cohort, focusing on the same cancer types and drugs as those studied in the trial, shows a similar pattern of top recommended drugs (as shown in
In addition to the trial cohorts, we analyzed the recently released POG570 cohort, where the post-treatment transcriptomics data together with treatment history is available for advanced or metastatic tumors of 570 patients. We first confirmed that the samples SL-scores are associated with longer treatment duration, which served as a proxy for therapeutic response in the original publication (shown in
Finally, we asked whether SELECT can successfully estimate the objective response rates (ORR) observed across different drug treatments in different clinical trials for a given cancer type. As these trials did measure and report the patients' tumor transcriptomics, we estimated, for each drug, its coverage (the patients who are predicted to respond based on their SL scores being larger than the 0.44 response threshold) in the TCGA cohort of the relevant cancer type (Methods). We collected ORR data from multiple clinical trials in melanoma and non-small cell lung cancer (a total of 3,246 patients from 18 trials). Reassuringly, we find that the resulting estimated coverage is significantly correlated with the observed ORR in both cancer types.
Beginning in operation 502, the method 500 may obtain, from one or more databases storing genetic interaction information, a plurality of candidate synthetic lethality (SL) gene partners for a cancer therapy. The one or more databases may store any number of SL gene partners for different cancer therapies and may be accessible by the computing device via a network or may be directly connected to the computing device. In another embodiment, the one or more databases may also or separately store candidate synthetic rescuer (SR) gene partners for a cancer therapy and such SR gene partners may also or separately be obtained. In operation 504, the method 500 may identify a subset of the obtained candidate SL gene partners based on patient response data and phylogenetic profile information of each of the plurality of candidate SL gene partners. The patient response data and/or phylogenetic profile information may be obtained from a separate database, the same database, or may be calculated by a computing device executing the method 500. In an alternate implementation, a subset of candidate SR gene partners may be identified as potential predictive biomarkers for the cancer therapy. The identification of the biomarkers may be based on at least one of (1) a product of PD1 and PDL1 gene expression levels, (2) CTLA4 protein expression levels, and (3) gene expression levels and somatic copy number alterations (SCNA) of each of a plurality of candidate SR gene partners.
In operation 506, the subset of the plurality of candidate SL gene partners may be filtered via a comparison to a BRAF inhibitor dataset, also obtained by a computing device from the database or a separate database. The comparison may provide an SL-score for each of the subset of the plurality of candidate SL gene partners such that the subset may be ranked based on the SL-score. In the alternate embodiment in which SR gene partners are considered, the method may rank the subset of the plurality of candidate SR gene partners based on phylogenetic distance information to filter the subset of the plurality of candidate SL gene partners and filter the ranked subset of the plurality of candidate SR gene partners based on an identification of candidate SR gene partners in which a downregulation (or upregulation) of a partner rescuer gene occurs. Finally, in operation 508, a cancer therapy for a patient may be identified, based at least on a transcriptomics profile of a tumor of the patient and the filtered subset of the plurality of candidate SL gene partners or SR gene partners, the cancer therapy for the patient. As such, through the method 500 of
I/O device 640 may also include an input device (not shown), such as an alphanumeric input device, including alphanumeric and other keys for communicating information and/or command selections to the processors 602-606. Another type of user input device includes cursor control, such as a mouse, a trackball, or cursor direction keys for communicating direction information and command selections to the processors 602-606 and for controlling cursor movement on the display device.
System 600 may include a dynamic storage device, referred to as main memory 616, or a random access memory (RAM) or other computer-readable devices coupled to the processor bus 612 for storing information and instructions to be executed by the processors 602-606. Main memory 616 also may be used for storing temporary variables or other intermediate information during execution of instructions by the processors 602-606. System 600 may include a read only memory (ROM) and/or other static storage device coupled to the processor bus 612 for storing static information and instructions for the processors 602-606. The system set forth in
According to one embodiment, the above techniques may be performed by computer system 600 in response to processor 604 executing one or more sequences of one or more instructions contained in main memory 616. These instructions may be read into main memory 616 from another machine-readable medium, such as a storage device. Execution of the sequences of instructions contained in main memory 616 may cause processors 602-606 to perform the process steps described herein. In alternative embodiments, circuitry may be used in place of or in combination with the software instructions. Thus, embodiments of the present disclosure may include both hardware and software components.
A machine readable medium includes any mechanism for storing or transmitting information in a form (e.g., software, processing application) readable by a machine (e.g., a computer). Such media may take the form of, but is not limited to, non-volatile media and volatile media. Non-volatile media includes optical or magnetic disks. Volatile media includes dynamic memory, such as main memory 616. Common forms of machine-readable medium may include, but is not limited to, magnetic storage medium (e.g., floppy diskette); optical storage medium (e.g., CD-ROM); magneto-optical storage medium; read only memory (ROM); random access memory (RAM); erasable programmable memory (e.g., EPROM and EEPROM); flash memory; or other types of medium suitable for storing electronic instructions.
Embodiments of the present disclosure include various steps, which are described in this specification. The steps may be performed by hardware components or may be embodied in machine-executable instructions, which may be used to cause a general-purpose or special-purpose processor programmed with the instructions to perform the steps. Alternatively, the steps may be performed by a combination of hardware, software and/or firmware.
Various modifications and additions can be made to the exemplary embodiments discussed without departing from the scope of the present invention. For example, while the embodiments described above refer to particular features, the scope of this invention also includes embodiments having different combinations of features and embodiments that do not include all of the described features. Accordingly, the scope of the present invention is intended to embrace all such alternatives, modifications, and variations together with all equivalents thereof.
This application claims priority to U.S. Provisional Application No. 63/107,737 entitled “SYNTHETIC LETHALITY-MEDIATED PRECISION ONCOLOGY VIA TURMOR TRANSCRPTOME” filed on Oct. 30, 2020, the entirety of which is incorporated herein by reference.
This invention was made with Government support under project number ZIA BC 011803 by the National Institutes of Health, National Cancer Institute. The United States Government has certain rights in the invention.
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/US21/57229 | 10/29/2021 | WO |
Number | Date | Country | |
---|---|---|---|
63107737 | Oct 2020 | US |