Abstract: Background: Tumor metabolism is the hallmark of cancer cells. Cancer cells utilize different nutrient sources to drive the metabolic pathways to sustain tumor growth. Glucose (Glu) and Glutamine (Gln) are the primary nutrient sources on which cancer cells thrive. Developing precision diet based on patient’s molecular characteristics can help treat the cancer with dietary modulations along with traditional approaches. Methods: Computational Biology Model (CBM) captures the signaling and metabolic pathways to predict cancer phenotypes and biomarkers. Genomic aberrations (Mutations / Copy Number Variations (CNV)) from a patient’s tumor are input into the CBM to create the patient disease model. CBM is used for diet prediction based on the molecular characteristics of the patient’s disease. CBM is validated using a data set of 54 cancer cell-lines across indications, by assessing nutrient dependency for Glu and Gln. Simulation based prediction of Glu and Gln dependency is based on the expression of transporters and rate limiting enzymes of cellular glucose (SLC2A1-4, HK2) and glutamine (SLC1A5, GLS) uptake. The enzymes regulating the de-novo synthesis of glucose (PCK1/2, FBP1) and glutamine (GLUL) are negative determinants. In the CBM, an index is defined to measure Glu and Gln dependency. Glu Dependency Index = (SLC2A1 + SLC2A2 + SLC2A3 + SLC2A4 + HK2) / (PCK1 + PCK2) Gln Dependency Index = (SLC1A5 + GLS) / (GLUL) Threshold values for Glu and Gln dependency was determined based on the simulation correlation with the cell-line data. The validated CBM was then used for diet prediction for patient genomics. Results: Validation of genomics-based diet prediction by CBM using 54 cancer cell lines had an accuracy, positive predictive value and negative predictive value of 85%, 97% and 44% for Glu dependency and 82%, 94% and 50% for Gln dependency respectively. Using this validated CBM, we present predictions of patient nutrient source dependency based on their tumor genomics: Case Study 1: A Glioblastoma multiforme (GBM) patient case with PTEN EGFRVIII and ALK mutation and high copy number of HIF1A and MIR-145. CBM predicted the patient to be Glu dependent and Gln independent. Case Study 2: A GBM patient case with CTNNB1 mutation and low copy number of PTEN, RB1 and NF1. This patient was predicted to be both Glu and Gln dependent. Case Study 3: A Triple Negative Breast Cancer (TNBC) patient carrying mutations for MYC, BRD4, EP300 and CREBBP. CBM predicted this patient to be Gln dependent and Glu independent. The rationales for the nutrient source dependency predictions based on disease pathway characteristics were determined. Conclusion: Using CBM we could successfully use patient genomic data to predict nutrient dependency of patient’s tumor. This analysis enables creating options for precision diet for a patient to be used as an adjuvant alongside traditional approaches. Citation Format: Shireen Vali, Taher Abbasi, Subrat Mohapatra, Vishwas Joseph, Ashish Kumar Agrawal, Anuj Tyagi, Neelesh Lunkad, Ashokraja Balla. Computational biology model (CBM) predicts nutrient dependency of cancer patients based on Tumor Genomics: Implication of precision diet in cancer therapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1843.

Abstract: Background: Acute promyelocytic leukemia (APL) is a biologically and clinically distinct subtype of acute myeloid leukemia (AML) with unique molecular pathogenesis, clinical manifestations, and treatment. APL is cytogenetically characterized by a balanced translocation t(15;17) (q24;q21), which involves the retinoic acid receptor alpha (RARA) gene on chromosome 17 and the promyelocytic leukemia (PML) gene on chromosome 15 that results in a PML-RARA fusion gene (PMID: 30575821). The PML-RARA fusion gene is the most critical event involved in the pathogenesis of APL, reported in 99% of APL patients (PMID: 32182684). The fusion confers a selective sensitivity to the targeted drugs, arsenic trioxide (ATO) and all-trans-retinoic acid (ATRA), with response rates over 90% (PMID: 31635329). However, the mechanism of resistance in the minority of non-responders is not well understood. This study used the Cellworks Omics Biology Model (CBM) to predict response to the combination of ATO-ATRA in patients harboring the PML-RARA fusion and identify mechanisms of resistance. Methods: Outcomes of 30 APL patients treated with ATRA or ATRA plus ATO were compared with outcomes predicted by CBM (Table 1). Genomic data from 6 publications (Table 2) derived from whole exome sequencing (WES), targeted next-generation sequencing (NGS), copy number variation (CNV) and/or karyotype data were used. All data was anonymized, de-identified and exempt from IRB review. The available genomic data for each profile was entered into the CBM which generates a patient-specific disease protein network model using PubMed and other online resources. The CBM predicts the patient-specific biomarker and phenotype response of a personalized diseased cell to drug agents, radiation and cell signaling. Disease biomarkers that are unique to each patient were identified within the protein network models. ATO and ATRA were simulated on all 30 patient cases. The treatment impact was assessed by quantitatively measuring the drug's effect on a cell growth score which is a composite of the quantified values for cell proliferation, survival, and apoptosis, along with the simulated impact on each patient-specific disease biomarker score. Each patient-specific model was also digitally screened to identify response to ATO and ATRA. Results: The CBM correctly predicted the response to ATO and ATRA in 28 of 30 cases. The overall prediction accuracy was 93% with a PPV of 100%, NPV of 60%, sensitivity of 93%, and specificity of 100%. In 2 of 30 patients who did not respond to ATO and ATRA, the CBM identified clinically relevant deletions to EZH2, KMT2E, and HIPK2 genes. All three genes are located on chromosome 7 and these non-responders had monosomy 7. Conclusions: The Cellworks Omics Biology Model predicted response to ATO and ATRA in APL patients harboring PML-RARA fusions. Predicting non-response to ATO and ATRA in patients with PML-RARA fusion up-front could prevent ineffective treatment, avoid unnecessary adverse events and reduce treatment costs. Additionally, computational modeling can identify new mechanisms of resistance and suggest alternative regimens for non-responding patients by targeting the patient-specific disease biomarkers unique to each. Disclosures Howard: Servier: Consultancy, Other: Speaker; Boston Scientific: Consultancy; Sanofi: Consultancy, Other: Speaker; EUSA Pharma: Consultancy; Cellworks: Consultancy. Nair:Cellworks Research India Private Limited: Current Employment. Grover:Cellworks Research India Private Limited: Current Employment. Tyagi:Cellworks Research India Private Limited: Current Employment. Kumari:Cellworks Research India Private Limited: Current Employment. Prasad:Cellworks Research India Private Limited: Current Employment. Mitra:Cellworks Research India Private Limited: Current Employment. Lala:Cellworks Research India Private Limited: Current Employment. Azam:Cellworks Research India Private Limited: Current Employment. Gupta:Cellworks Research India Private Limited: Current Employment. Mohapatra:Cellworks Research India Private Limited: Current Employment. G:Cellworks Research India Private Limited: Current Employment. Mundkur:Cellworks Group Inc.: Current Employment. Macpherson:Cellworks Group Inc.: Current Employment. Kapoor:Cellworks Research India Private Limited: Current Employment.

Abstract: Background: DNA methyltransferase inhibition (DNMTi) with the hypomethylating agents (HMA) azacitidine (AZA) or decitabine, remains the mainstay of therapy for the majority of high-risk Myelodysplastic Syndromes (MDS) patients. Nevertheless, only 40-50% of MDS patients achieve clinical improvement with DNMTi. There is a need for a predictive clinical approach that can stratify MDS patients according to their chance of benefit from current therapies and that can identify and predict responses to new treatment options. Ideally, patients predicted to be non-responders (NR) could be offered alternative strategies while being spared protracted treatment with HMA alone that has a low likelihood of efficacy. Recently, an intriguing discovery of immune modulation by HMA has emerged. In addition to the benefits of unsilencing differentiation genes and tumor suppressor genes, HMA's reactivate human endogenous retroviral (HERV) genes leading to viral mimicry and upregulation of the immune response as a major mechanism of HMA efficacy. Although the PD-L1/PD1 blockade plus HMA has been recognized as a beneficial combination, there are no established markers to guide decision-making. We report here the utility of immunomic profiling of chromosome 9 copy number status as a significant mechanism of immune evasion and HMA resistance. Methods: 119 patients with known clinical responses to AZA were selected for this study. Publicly available data largely from TCGA and PubMed was utilized for this study. The aberration and copy number variations from individual cases served as input into the Cellworks Computation Omics Biology Model (CBM), a computational biology multi-omic software model, created using artificial intelligence heuristics and literature sourced from PubMed, to generate a patient-specific protein network map. Disease-biomarkers unique to each patient were identified within protein network maps. The Cellworks Biosimulation Platform has the capacity to biosimulate disease phenotypic behavior and was used to create a disease model and then conduct biosimulations to measure the effect of AZA on a cell growth score comprised of a composite of cell proliferation, viability, apoptosis, metastasis, and other cancer hallmarks. Biosimulation of drug response was conducted to identify and predict therapeutic efficacy. Results: Although AZA treatment increased tumor associated antigens and interferon signaling, it also increased PD-L1 expression to inactivate cytotoxic CD8(+) T cells. Copy number alterations of the chromosome 9p region were found to significantly drive PD-L1 expression with multiple genes such as CD274, IFNA1, IFNA2, JAK2, PDCD1LG and KDM4C playing a role in PD-L1 regulation further increasing immune suppression (Figure 1). Among 6 cases of chromosome 9p aberration in this dataset, 9p amp (n=2) were clinical non-responders (NR) while 9p del (n=4) were responders (R) to AZA. In principle, checkpoint immunotherapy could improve outcomes for patients with 9p abnormalities. Additionally, copy number variation loss of key genes located on chromosome 16 involved in antigen processing and presentation such as CIITA, CTCF, IRF8, PSMB10, NLRC5, and SOCS1 were found to negatively impact AZA sensitivity (NR=4; R=0); these patients would also be unlikely to respond to checkpoint immunotherapy. Also, aberrations in melanoma antigen gene (MAGE) family proteins (NR=2; R=O), and STT3A (NR=1; R=5) were found to impact AZA efficacy by decreasing antigen processing on tumor cells. Conclusion: Based on the results from the Cellworks Biosimulation Platform applied to the CBM, copy number variants of chromosome 9p and 16 can be converted into CBM-derived biomarkers for response to checkpoint immunotherapy in combination with HMA. Our results support a future prospective evaluation in larger cohorts of MDS patients. Figure 1 Figure 1. Disclosures Howard: Servier: Consultancy; Cellworks Group Inc.: Consultancy; Sanofi: Consultancy, Other: Speaker fees. Kumar: Cellworks Group Inc.: Current Employment. Castro: Bugworks: Consultancy; Exact sciences Inc.: Consultancy; Guardant Health Inc.: Speakers Bureau; Cellworks Group Inc.: Current Employment; Caris Life Sciences Inc.: Consultancy; Omicure Inc: Consultancy. Grover: Cellworks Group Inc.: Current Employment. Mohapatra: Cellworks Group Inc.: Current Employment. Kapoor: Cellworks Group Inc.: Current Employment. Tyagi: Cellworks Group Inc.: Current Employment. Nair: Cellworks Group Inc.: Current Employment. Suseela: Cellworks Group Inc.: Current Employment. Pampana: Cellworks Group Inc.: Current Employment. Lala: Cellworks Group Inc.: Current Employment. Singh: Cellworks Group Inc.: Current Employment. Shyamasundar: Cellworks Group Inc.: Current Employment. Kulkarni: Cellworks Group Inc.: Current Employment. Narvekar: Cellworks Group Inc.: Current Employment. Sahni: Cellworks Group Inc.: Current Employment. Raman: Cellworks Group Inc.: Current Employment. Balakrishnan: Cellworks Group Inc.: Current Employment. Patil: Cellworks Group Inc.: Current Employment. Palaniyeppa: Cellworks Group Inc.: Current Employment. Balla: Cellworks Group Inc.: Current Employment. Patel: Cellworks Group Inc.: Current Employment. Mundkur: Cellworks Group Inc: Current Employment. Christie: Cellworks Group Inc.: Current Employment. Macpherson: Cellworks Group Inc.: Current Employment. Marcucci: Abbvie: Other: Speaker and advisory scientific board meetings; Novartis: Other: Speaker and advisory scientific board meetings; Agios: Other: Speaker and advisory scientific board meetings.

Abstract: Background: The optimal treatment strategy for managing Acute Myeloid Leukemia (AML) and the use of reliable and predictive biomarkers to guide selection of cytotoxic chemotherapy regimens among patients with diverse genomic profiles remain unmet needs in the clinic. The combination of MEC [mitoxantrone (MIT), etoposide (VP16), and cytarabine (ARA-C)] is a commonly used regimen for relapsed or refractory AML patients. Unfortunately, many patients do not respond to MEC, and which of the three drug agents matters most for each individual patient is not known. Predictors of response are needed urgently. Methods: The Computational Omics Biology Model (CBM) is a computational multi-omic biology software model created using artificial intelligence heuristics and literature sourced from PubMed to generate a patient-specific protein network map. The aberration and copy number variations from individual cases served as input into the CBM. Disease-biomarkers unique to each patient were identified within patient-specific protein network maps. Biosimulations were conducted on the Cellworks Biosimulation Platform by measuring the effect of chemotherapy on a cell growth score comprised of a composite of cell proliferation, viability, apoptosis, metastasis, and other cancer hallmarks. Biosimulation of drugs was conducted by mapping the interaction of various drug combinations with the patient's genomic and pathway alterations based on signaling pathway mechanisms and their phenotypic consequences. The Cellworks Biosimulation Platform identified unique chromosomal signatures that permit a stratification of patients that are most likely to respond to MIT, VP16, or ARA-C as well as their combinations. 65 AML patients were selected for this study largely based on genomic data published in TCGA and PubMed: ARA-C [N=12, 7 responders (R) & 5 non-responders (NR)]ARA-C + MIT [N=30, 29 R & 1 NR]ARA-C + MIT + VP16 [N=23, 12 R & 11 NR] Results: Of the12 patients treated with ARA-C alone, 5 were predicted to be NR and 7 were predicted to be R. Of the 5 NR, 4 had 5q del which resulted in loss of APC, CSNK1A1 and SLC22A4 (nucleoside carrier) forming the non-response biomarkers for ARA-C. Notably, the biosimulation predicted lenalidomide to be beneficial for these patients. Out of 7 R, 4 patients also had 5q del, but were predicted to be R because of co-occurring aberrations involving CLSPN del, DHODH del, MSH2 del, EP300 del, CREBBP del, MSH6 Del, and RRM2 del. These genes were exclusively present in ARA-C responders. Of 53 patients who received ARA-C + VP16 + MIT or ARA-C + MIT, 41 patients were predicted to be R and 12 patients were predicted to be NR. The genomic aberrations predicted by biosimulation to be associated with response to this regimen include: NPM1-mut, TET2-mut, IDH1-mut, IDH2-mut, RAD17-del, NRAS-mut (Table 1). Notably, CBM predicted 19 of the 41 R had no genomic biomarkers of response to VP16 or MIT, suggesting these patients might have benefited equally from ARA-C alone with less toxicity and cost. Finally, 11/65 patients were predicted NR to MEC treatment. In the biosimulation, treatment failure was associated with high aberration frequencies of KMT2C-mut/del, FLT3 mut, TWIST1 del, LIMK1 del, SNAI2 amp, FNTA amp, and KAT6A amp. Of note, these genomic markers suggested a likelihood of benefit from other therapies, including vincristine, JQ1 and rigosertib. Conclusions: The Cellworks Biosimulation Platform identified novel polygenic biomarkers of response that can be employed to determine the optimal therapy for relapsed AML patients. Biosimulation permits avoidance of cytotoxic drugs with little chance of efficacy and reveals vulnerabilities in each patient's cancer that can be exploited to improve disease control. In AML, biosimulation promises to improve intensive therapy regimens by tailoring chemotherapy to optimize disease control and minimize toxicity. Figure 1 Figure 1. Disclosures Marcucci: Novartis: Other: Speaker and advisory scientific board meetings; Agios: Other: Speaker and advisory scientific board meetings; Abbvie: Other: Speaker and advisory scientific board meetings. Kumar: Cellworks Group Inc.: Current Employment. Castro: Cellworks Group Inc.: Current Employment; Omicure Inc: Consultancy; Caris Life Sciences Inc.: Consultancy; Exact sciences Inc.: Consultancy; Bugworks: Consultancy; Guardant Health Inc.: Speakers Bureau. Grover: Cellworks Group Inc.: Current Employment. Patil: Cellworks Group Inc.: Current Employment. Alam: Cellworks Group Inc.: Current Employment. Azam: Cellworks Group Inc.: Current Employment. Mohapatra: Cellworks Group Inc.: Current Employment. Tyagi: Cellworks Group Inc.: Current Employment. Kumari: Cellworks Group Inc.: Current Employment. Prasad: Cellworks Group Inc.: Current Employment. Nair: Cellworks Group Inc.: Current Employment. Lunkad: Cellworks Group Inc.: Current Employment. Joseph: Cellworks Group Inc.: Current Employment. G: Cellworks Group Inc.: Current Employment. Chauhan: Cellworks Group Inc.: Current Employment. Basu: Cellworks Group Inc.: Current Employment. Behura: Cellworks Group Inc.: Current Employment. Ghosh: Cellworks Group Inc.: Current Employment. Husain: Cellworks Group Inc.: Current Employment. Mandal: Cellworks Group Inc.: Current Employment. Raman: Cellworks Group Inc.: Current Employment. Patel: Cellworks Group Inc.: Current Employment. Mundkur: Cellworks Group Inc: Current Employment. Christie: Cellworks Group Inc.: Current Employment. Macpherson: Cellworks Group Inc.: Current Employment. Howard: Servier: Consultancy; Cellworks Group Inc.: Consultancy; Sanofi: Consultancy, Other: Speaker fees.

Abstract: 7027 Background: ATRA combined with arsenic trioxide revolutionized the treatment of APL. Based on promising in vitro data, several clinical trials evaluated ATRA combinations in non-APL AML, in which some patients seemed to benefit from the addition. Thus, predicting response a priori is imperative to determine the optimal treatment for each patient. The CBM was used to evaluate the impact of initial therapy with ATRA combined with cytarabine, etoposide, idarubicin (ATRA-CEI) to assess the biomarkers responsible for response in adults with AML. Methods: AML patients participating in clinical trial NCT00151242 had their leukemia sequenced as part of the trial, and genomic profiles were used for computational modeling by the CBM, which uses curated data about genomic aberrations from PubMed as input to generate disease-specific protein network maps and predict drug responses. Disease biomarkers unique to each patient were identified using biosimulation. Digital drug simulations were conducted by measuring the effect of ATRA-CEI on a composite cell growth score of cell proliferation, apoptosis and other hallmarks of cancer. ATRA-CEI was mapped to the patient genome along with a mechanism of action and validated based on the genomic profile and its biological consequences. Results: Of 171 patients treated with ATRA-CEI, 107 (63%) responded (R) and 64 did not (NR). A subset of 18 patients with favorable genomic features were found to be NR and their non-response was correctly predicted by CBM in all 18 cases. Mutations of DNMT3A, EZH2, ASXL, FLT-3, and GART amplification emerged as novel biomarkers of ATRA-CEI failure (only 37 of 107 responders (35%) with these findings, compared to 70 of 107 responders (65%) without these findings (p = 0.0027)). DNMT3A, EZH2, ASXL1 loss of function mutations activate FABP5, a key mechanism of ATRA resistance, and also activate ABCC1 (PgP), which reduces the efficacy of etoposide and idarubicin by upregulating MDR1. In general, monosomy 7 is expected to confer ATRA resistance due to the presence of EZH2 and KMT2E gene deletions. Indeed, 18 of 32 patients with monosomy 7 did not respond. However, the 14 who responded had co-occurrence of deletions involving IGFBP3, PMS2, HUS1, CDK5, XRCC2/4, AKR1B10, and others that overcame ATRA resistance associated with monosomy 7 and were identified by CBM. Use of CBM helps avoid unnecessary use of ATRA in patients unlikely to respond (19% of cases) thus reducing toxicity and cost without changing efficacy, and also identifies those likely to respond, even when they have monosomy 7, where non-response is the norm. Conclusions: ATRA benefits a subset of patients with non-APL AML. CBM predicted response using computational modeling of all genetic alternations, which explains its success versus traditional one-gene-one-drug approaches.