Computational drug treatment simulations on projections of dysregulated protein networks derived from the myelodysplastic mutanome match clinical response in patients

Computational Biological Modeling Identifies PD-(L)1 Immunotherapy Sensitivity Among Molecular Subgroups of KRAS-Mutated Non-Small-Cell Lung Cancer

PURPOSE

KRAS-mutated (KRAS^MUT) non-small-cell lung cancer (NSCLC) is emerging as a heterogeneous disease defined by comutations, which may confer differential benefit to PD-(L)1 immunotherapy. In this study, we leveraged computational biological modeling (CBM) of tumor genomic data to identify PD-(L)1 immunotherapy sensitivity among KRAS^MUT NSCLC molecular subgroups.

MATERIALS AND METHODS

In this multicohort retrospective analysis, the genotype clustering frequency ranked method was used for molecular clustering of tumor genomic data from 776 patients with KRAS^MUT NSCLC. These genomic data were input into the CBM, in which customized protein networks were characterized for each tumor. The CBM evaluated sensitivity to PD-(L)1 immunotherapy using three metrics: programmed death-ligand 1 expression, dendritic cell infiltration index (nine chemokine markers), and immunosuppressive biomarker expression index (14 markers).

RESULTS

Genotype clustering identified eight molecular subgroups and the CBM characterized their shared cancer pathway characteristics: KRAS^MUT/TP53^MUT, KRAS^MUT/CDKN2A/B/C^MUT, KRAS^MUT/STK11^MUT, KRAS^MUT/KEAP1^MUT, KRAS^MUT/STK11^MUT/KEAP1^MUT, KRAS^MUT/PIK3CA^MUT, KRAS ^MUT/ATM^MUT, and KRAS^MUT without comutation. CBM identified PD-(L)1 immunotherapy sensitivity in the KRAS^MUT/TP53^MUT, KRAS^MUT/PIK3CA^MUT, and KRAS^MUT alone subgroups and resistance in the KEAP1^MUT containing subgroups. There was insufficient genomic information to elucidate PD-(L)1 immunotherapy sensitivity by the CBM in the KRAS^MUT/CDKN2A/B/C^MUT, KRAS^MUT/STK11^MUT, and KRAS^MUT/ATM^MUT subgroups. In an exploratory clinical cohort of 34 patients with advanced KRAS^MUT NSCLC treated with PD-(L)1 immunotherapy, the CBM-assessed overall survival correlated well with actual overall survival (r = 0.80, P < .001).

CONCLUSION

CBM identified distinct PD-(L)1 immunotherapy sensitivity among molecular subgroups of KRAS^MUT NSCLC, in line with previous literature. These data provide proof-of-concept that computational modeling of tumor genomics could be used to expand on hypotheses from clinical observations of patients receiving PD-(L)1 immunotherapy and suggest mechanisms that underlie PD-(L)1 immunotherapy sensitivity.

A genomics-informed computational biology platform prospectively predicts treatment responses in AML and MDS patients

Patients with myelodysplastic syndromes (MDS) or acute myeloid leukemia (AML) are generally older and have more comorbidities. Therefore, identifying personalized treatment options for each patient early and accurately is essential. To address this, we developed a computational biology modeling (CBM) and digital drug simulation platform that relies on somatic gene mutations and gene CNVs found in malignant cells of individual patients. Drug treatment simulations based on unique patient-specific disease networks were used to generate treatment predictions. To evaluate the accuracy of the genomics-informed computational platform, we conducted a pilot prospective clinical study (NCT02435550) enrolling confirmed MDS and AML patients. Blinded to the empirically prescribed treatment regimen for each patient, genomic data from 50 evaluable patients were analyzed by CBM to predict patient-specific treatment responses. CBM accurately predicted treatment responses in 55 of 61 (90%) simulations, with 33 of 61 true positives, 22 of 61 true negatives, 3 of 61 false positives, and 3 of 61 false negatives, resulting in a sensitivity of 94%, a specificity of 88%, and an accuracy of 90%. Laboratory validation further confirmed the accuracy of CBM-predicted activated protein networks in 17 of 19 (89%) samples from 11 patients. Somatic mutations in the TET2, IDH1/2, ASXL1, and EZH2 genes were discovered to be highly informative of MDS response to hypomethylating agents. In sum, analyses of patient cancer genomics using the CBM platform can be used to predict precision treatment responses in MDS and AML patients.

Prediction of Outcomes with a Computational Biology Model in Newly Diagnosed Glioblastoma Patients Treated with Radiation Therapy and Temozolomide

PURPOSE

Precision medicine has been most successful in targeting single mutations, but personalized medicine using broader genomic tumor profiles for individual patients is less well developed. We evaluate a genomics-informed computational biology model (CBM) to predict outcomes from standard treatments and to suggest novel therapy recommendations in glioblastoma (GBM).

METHODS AND MATERIALS

In this retrospective study, 98 patients with newly diagnosed GBM undergoing surgery followed by radiation therapy and temozolomide at a single institution with available genomic data were identified. Incorporating mutational and copy number aberration data, a CBM was used to simulate the response of GBM tumor cells and generate efficacy predictions for radiation therapy (RTeff) and temozolomide (TMZeff). RTeff and TMZeff were evaluated for association with overall survival and progression-free survival in a Cox regression model. To demonstrate a CBM-based individualized therapy strategy, treatment recommendations were generated for each patient by testing a panel of 45 central nervous system-penetrant US Food and Drug Administration-approved agents.

RESULTS

High RTeff scores were associated with longer survival on univariable analysis (P < .001), which persisted after controlling for age, extent of resection, performance status, MGMT, and IDH status (P = .017). High RTeff patients had a longer overall survival compared with low RTeff patients (median, 27.7 vs 14.6 months). High TMZeff was also associated with longer survival on univariable analysis (P = .007) but did not hold on multivariable analysis, suggesting an interplay with MGMT status. Among predictions of the 3 most efficacious combination therapies for each patient, only 2.4% (7 of 294) of 2-drug recommendations produced by the CBM included TMZ.

CONCLUSIONS

CBM-based predictions of RT and TMZ effectiveness were associated with survival in patients with newly diagnosed GBM treated with those therapies, suggesting a possible predictive utility. Furthermore, the model was able to suggest novel individualized monotherapies and combinations. Prospective evaluation of such a personalized treatment strategy in clinical trials is needed.

Identification of Lenalidomide Sensitivity and Resistance Mechanisms in Non-Del(5q) Myelodysplastic Syndromes

Whereas lenalidomide is an effective therapy for del(5q) MDS patients, a minority of non-del(5q) MDS patients achieve hematologic improvement with lenalidomide. We used computational biology modeling and digital drug simulation to examine genomic data from 56 non-del(5q) MDS patients treated with lenalidomide, and then matched treatment response with molecular pathways. The computer inferred genomic abnormalities associating with lenalidomide treatment response in non-del(5q) MDS to include trisomy 8, del(20q), or RUNX1 loss of function mutations. Genomic abnormalities associating with lenalidomide resistance in non-del(5q) MDS patients included mutations in SF3B1, TET2, WNT3A amplification, MCL1 amplification, and/or PSEN2 amplification. These results may inform protocols for determining appropriateness of lenalidomide in non-del(5q) MDS.

Molecular pathophysiology of the myelodysplastic syndromes: insights for targeted therapy

The clinical heterogeneity of the myelodysplastic syndromes (MDSs) relates to the recently discerned panoply of molecular abnormalities extant within this disease spectrum. Despite increasing recognition of these biologic abnormalities, very limited therapeutic options exist to exploit our increasing understanding of the molecular pathophysiology of MDS, with only 1 therapy (lenalidomide) particularly focused on a specific clinical patient subset (del(5q) cytogenetics) and 2 epigenetic modulators (azacitidine and decitabine) having been approved for treating these patients. This article will review the mutational and biologic landscape of these disorders, as well as the targeted therapeutics currently in clinical trials that are focused on attacking these features. Given the molecular complexity of these disorders and the limited repertoire of effective therapeutic agents, we will also discuss novel approaches attempting to determine potentially effective and personalized treatment options through complementary chemosensitivity and computerized signaling network screening for these disparate MDS patient subsets. Translational use of such resources, combined with the rapidly evolving next-generation molecular technologies, should prove useful in effectuating improved and more selective options for therapy.

A New Approach for the Diagnosis of Myelodysplastic Syndrome Subtypes Based on Protein Interaction Analysis

Myelodysplastic syndromes (MDS) are a heterogeneous group of hematological malignancies with a high risk of transformation to acute myeloid leukemia (AML). MDS are associated with posttranslational modifications of proteins and variations in the protein expression levels. In this work, we present a novel interactomic diagnostic method based on both protein array and surface plasmon resonance biosensor technology, which enables monitoring of protein-protein interactions in a label-free manner. In contrast to conventional methods based on the detection of individual biomarkers, our presented method relies on measuring interactions between arrays of selected proteins and patient plasma. We apply this method to plasma samples obtained from MDS and AML patients, as well as healthy donors, and demonstrate that even a small protein array comprising six selected proteins allows the method to discriminate among different MDS subtypes and healthy donors.

Sequential azacitidine and lenalidomide for patients with relapsed and refractory acute myeloid leukemia: Clinical results and predictive modeling using computational analysis

BACKGROUND

Patients with relapsed and refractory (R/R) acute myeloid leukemia (AML) have limited treatment options. Genomically-defined personalized therapies are only applicable for a minority of patients. Therapies without identifiable targets can be effective but patient selection is challenging. The sequential combination of azacitidine with high-dose lenalidomide has shown activity; we aimed to determine the efficacy of this genomically-agnostic regimen in patients with R/R AML, with the intention of applying sophisticated methods to predict responders.

METHODS

Thirty-seven R/R AML/myelodysplastic syndrome patients were enrolled in a phase 2 study of azacitidine with lenalidomide. The primary endpoint was complete remission (CR) and CR with incomplete blood count recovery (CRi) rate. A computational biological modeling (CBM) approach was applied retrospectively to predict outcomes based on the understood mechanisms of azacitidine and lenalidomide in the setting of each patients' disease.

FINDINGS

Four of 37 patients (11%) had a CR/CRi; the study failed to meet the alternative hypothesis. Significant toxicity was observed in some cases, with three treatment-related deaths and a 30-day mortality rate of 14%. However, the CBM method predicted responses in 83% of evaluable patients, with a positive and negative predictive value of 80% and 89%, respectively.

INTERPRETATION

Sequential azacitidine and high-dose lenalidomide is effective in a minority of R/R AML patients; it may be possible to predict responders at the time of diagnosis using a CBM approach. More efforts to predict responses in non-targeted therapies should be made, to spare toxicity in patients unlikely to respond and maximize treatments for those with limited options.

Computational modeling of early T-cell precursor acute lymphoblastic leukemia (ETP-ALL) to identify personalized therapy using genomics

Early T-cell precursor acute lymphoblastic leukemia (ETP-ALL) is an aggressive hematological malignancy for which optimal therapeutic approaches are poorly characterized. Using computational biology modeling (CBM) in conjunction with genomic data from cell lines and individual patients, we generated disease-specific protein network maps that were used to identify unique characteristics associated with the mutational profiles of ETP-ALL compared to non-ETP-ALL (T-ALL) cases and simulated cellular responses to a digital library of FDA-approved and investigational agents. Genomics-based classification of ETP-ALL patients using CBM had a prediction sensitivity and specificity of 93% and 87%, respectively. This analysis identified key genomic and pathway characteristics that are distinct in ETP-ALL including deletion of nucleophosmin-1 (NPM1), mutations of which are used to direct therapeutic decisions in acute myeloid leukemia. Computational simulations based on mutational profiles of 62 ETP-ALL patient models identified 87 unique targeted combination therapies in 56 of the 62 patients despite actionable mutations being present in only 37% of ETP-ALL patients. Shortlisted two-drug combinations were predicted to be synergistic in 11 profiles and were validated by in vitro chemosensitivity assays. In conclusion, computational modeling was able to identify unique biomarkers and pathways for ETP-ALL, and identify new drug combinations for potential clinical testing.

Predicting response to BET inhibitors using computational modeling: A BEAT AML project study

Despite advances in understanding the molecular pathogenesis of acute myeloid leukaemia (AML), overall survival rates remain low. The ability to predict treatment response based on individual cancer genomics using computational modeling will aid in the development of novel therapeutics and personalize care. Here, we used a combination of genomics, computational biology modeling (CBM), ex vivo chemosensitivity assay, and clinical data from 100 randomly selected patients in the Beat AML project to characterize AML sensitivity to a bromodomain (BRD) and extra-terminal (BET) inhibitor. Computational biology modeling was used to generate patient-specific protein network maps of activated and inactivated protein pathways translated from each genomic profile. Digital drug simulations of a BET inhibitor (JQ1) were conducted by quantitatively measuring drug effect using a composite AML disease inhibition score. 93% of predicted disease inhibition scores matched the associated ex vivo IC₅₀ value. Sensitivity and specificity of CBM predictions were 97.67%, and 64.29%, respectively. Genomic predictors of response were identified. Patient samples harbouring chromosomal aberrations del(7q) or −7, +8, or del(5q) and somatic mutations causing ERK pathway dysregulation, responded to JQ1 in both in silico and ex vivo assays. This study shows how a combination of genomics, computational modeling and chemosensitivity testing can identify network signatures associating with treatment response and can inform priority populations for future clinical trials of BET inhibitors.

Depletion of ZBTB38 potentiates the effects of DNA demethylating agents in cancer cells via CDKN1C mRNA up-regulation

DNA methyltransferase inhibitor (DNMTi) treatments have been used for patients with myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML), and have shown promising beneficial effects in some other types of cancers. Here, we demonstrate that the transcriptional repressor ZBTB38 is a critical regulator of the cellular response to DNMTi. Treatments with 5-azacytidine, or its derivatives decitabine and zebularine, lead to down-regulation of ZBTB38 protein expression in cancer cells, in parallel with cellular damage. The depletion of ZBTB38 by RNA interference enhances the toxicity of DNMTi in cell lines from leukemia and from various solid tumor types. Further we observed that inactivation of ZBTB38 causes the up-regulation of CDKN1C mRNA, a previously described indirect target of DNMTi. We show that CDKN1C is a key actor of DNMTi toxicity in cells lacking ZBTB38. Finally, in patients with MDS a high level of CDKN1C mRNA expression before treatment correlates with a better clinical response to a drug regimen combining 5-azacytidine and histone deacetylase inhibitors. Collectively, our results suggest that the ZBTB38 protein is a target of DNMTi and that its depletion potentiates the toxicity of DNMT inhibitors in cancer cells, providing new opportunities to enhance the response to DNMT inhibitor therapies in patients with MDS and other cancers.

Computational Modeling and Treatment Identification in the Myelodysplastic Syndromes

Purpose of Review

This review discusses the need for computational modeling in myelodysplastic syndromes (MDS) and early test results.

Recent Findings

As our evolving understanding of MDS reveals a molecularly complicated disease, the need for sophisticated computer analytics is required to keep track of the number and complex interplay among the molecular abnormalities. Computational modeling and digital drug simulations using whole exome sequencing data input have produced early results showing high accuracy in predicting treatment response to standard of care drugs. Furthermore, the computational MDS models serve as clinically relevant MDS cell lines for pre-clinical assays of investigational agents.

Summary

MDS is an ideal disease for computational modeling and digital drug simulations. Current research is focused on establishing the prediction value of computational modeling. Future research will test the clinical advantage of computer-informed therapy in MDS.