No abstract

The majority of drugs currently used to treat rheumatoid arthritis (RA) act on a small number of immunomodulatory targets. We applied an integrative biomedical-informatics-based approach and in vivo testing to identify new drug candidates and potential therapeutic targets that could form the basis for future drug development in RA. A computational model of RA was constructed by integrating patient gene expression data, molecular interactions, and clinical drug-disease associations. Drug candidates were scored based on their predicted efficacy across these data types. Ten high-scoring candidates were subsequently screened in a collagen-induced arthritis model of RA. Treatment with exenatide, olopatadine, and TXR-112 significantly improved multiple preclinical endpoints, including animal mobility which was measured using a novel digital platform. These three drug candidates do not act on common RA therapeutic targets; however, links between known candidate pharmacology and pathological processes involved in RA suggest hypothetical mechanisms contributing to the observed efficacy.

Background:Lupus is a heterogeneous, systemic disease that affects millions of patients globally with a high unmet medical need. We present results from our powerful and efficient computational drug discovery platform that identifies hits with first-in-class mechanisms of action that can advance rapidly and successfully through preclinical validation studies. The twoXAR discovery platform uses an artificial-intelligence framework to integrate diverse patient-derived biomedical data sets to build holistic and unbiased models of human disease biology. The utilization of diverse, proprietary algorithms and deep learning principles provides a highly sensitive platform to elucidate complex disease-specific associations between biology and biomedical data that are integrated with a library of existing drug molecules. This enables the identification of novel, high-value drug discovery hits with known pharmacological properties. The twoXAR platform also preserves interpretable data-driven links to disease biology to facilitate efficient validation and optimization studies.Objectives:Apply twoXAR’s computational drug discovery platform for the discovery of first-in-class lupus therapy hits and perform preclinical characterization of selected hits to identify drug discovery lead molecules.Methods:Using clinical SLE patient data, we employed the twoXAR platform to build anin-silicoSLE disease model. Nine molecules with novel mechanisms of action (not previously tested as candidate clinical therapies for lupus) were identified as drug discovery hits and then characterized in preclinical efficacy studies using the MRL mouse model of lupus.Results:In preclinical validation studies with the MRL mouse model, 2 compounds were differentiated by significant efficacy and excellent tolerability. TXR-711 and TXR-712 increased renal function, decreased renal inflammation and decreased inflammation compared to vehicle-treated control mice. In particular, TXR-711 and TXR-712 significantly decreased serum blood urea nitrogen (BUN) levels, decreased proteinuria levels, and significantly improved kidney histology readouts such as glomerulonephritis and tubule basophilia. Additionally, TXR-711 and TXR-712 treatment resulted in significantly decreased inguinal lymph node weight.Conclusion:TXR-711 and TXR-712 were identified as SLE drug discovery leads with novel MOAs for further preclinical development. Ongoing studies with TXR-711 and TXR-712 includes pharmacokinetic, pharmacodynamic, and additional MRL mouse efficacy characterization.Disclosure of Interests:Isaac Hakim Employee of: twoXAR, Inc, Sana Mujahid Employee of: twoXAR, Inc., Aaron C. Daugherty Employee of: twoXAR, Inc., Timothy S. Heuer Employee of: twoXAR, Inc

inadequate response to standard of care (SOC) therapy (NCT04058028). Methods In this adaptive, phase 2, placebo-controlled, doseranging study, subjects (N~300, age 18-75 years) will be randomized to receive placebo or 1 of 3 doses of AMG 570 Q2W for 52 weeks, followed by 16 weeks of safety followup. The primary objective is to evaluate efficacy of AMG 570 compared with placebo at week 24 using the SLE Responder Index (SRI-4). Key secondary endpoints include SRI-4 at week 52 with oral corticosteroid (OCS) reduction (!10 mg/day at baseline to £7.5 mg/day in weeks 44-52) and SRI-4 and Lupus Low Disease Activity State at week 52. Subjects will undergo 2 screening visits to fulfill criteria for active SLE and demonstrate adherence to prior SLE treatment including OCS, immunosuppressants, and/or immunomodulators. Blood screening tests will confirm detectable serum drug levels of baseline SOC medications. RAR aims to allocate more subjects to more efficacious doses while maintaining the placebo allocation constant; the randomization ratio could be adapted after interim analyses based on clinical efficacy. The trial includes interim analyses for futility using the Bayesian approach. Results Study ongoing. Conclusion This study will provide safety and efficacy data for AMG 570 compared with placebo, and its adaptive trial design aims to optimize development of a novel therapy for SLE patients with inadequate response to current SOC. Acknowledgments Amgen Inc. sponsored this study.

Hepatocellular carcinoma (HCC) and pancreatic ductal adenocarcinoma (PDAC) have among the lowest 5-year survival rates of all cancer types at 18% and 9%, respectively. Treatment options for patients with liver or pancreatic cancer are relatively unchanged over the past 10 years. HCC has seen the recent FDA-approval of multi-kinase inhibitor therapies with similar mechanisms of action, including cabozantinib, regorafenib, and lenvatinib, and the immune checkpoint inhibitor nivolumab (conditionally). Despite these advances, the survival rate and median survival time for HCC patients remain poor. The picture for PDAC patients is similar, although with even greater need for new therapies. We present results from a powerful and efficient computational drug discovery platform that produces drug discovery hits with first-in-class mechanisms of action that can advance rapidly and successfully through preclinical validation studies. The twoXAR discovery platform uses an artificial-intelligence framework to integrate diverse patient-derived data sets and build holistic and unbiased models of human disease biology. The utilization of diverse, proprietary algorithms and deep learning principles provides a highly sensitive platform to elucidate detailed disease-specific associations between biology and biomedical data that are integrated with a library of existing drug molecules to deliver novel, high-value drug discovery hits. The twoXAR platform delivers drug discovery hits with known pharmacological properties and preserves the data-driven links to disease biology; this facilitates validation and optimization studies. We employed the twoXAR platform to build in-silico disease models of HCC and PDAC using disease-specific data and generated a set of 10 molecules with predicted efficacy in HCC and a second, independent set of 11 molecules with predicted efficacy in PDAC. These independent sets of disease-specific drug discovery hits represented novel mechanisms of action that had not been tested previously as potential clinical therapies for HCC or PDAC, respectively. TXR-311 and TXR-312, and TXR-405 and TXR-411 were discovered as validated hits for HCC and PDAC, respectively, using in vitro cell proliferation and viability assays with HCC and PDAC tumor cell lines. In these studies, TXR-311 inhibited proliferation and viability of five different HCC tumor cell lines with IC50 values that were 70-fold lower than IC50 values for sorafenib and displayed greater than 500-fold selectivity against primary human hepatocytes. In subsequent in vivo efficacy studies using two HCC patient-derived xenograft (PDX) tumor models, TXR-311 showed excellent tolerability and displayed significant tumor growth inhibition efficacy compared to vehicle-treated controls. TXR-311 presents a first-in-class lead for further development as a potential HCC therapy. Citation Format: Isaac Hakim, Mei-Sze Chua, Wei Wei, Li Ma, Elizabeth Noblin, Samuel So, Aaron C. Daugherty, Timothy S. Heuer. Computational discovery and preclinical validation of therapeutic leads with novel MOAs for hepatocellular carcinoma and pancreatic ductal adenocarcinoma [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr LB-110.