Optimizing and Personalising Azacitidine Combination Therapy for Treating Solid Tumours QPOP and CURATE.AI

Optimizing and Personalising Azacitidine Combination Therapy for Treating Solid Tumours QPOP and CURATE.AI

Optimizing and Personalising Azacitidine Combination Therapy for Treating Solid Tumours Using the Quadratic Phenotypic Optimization Platform (QPOP) and an Artificial Intelligence-based Platform (CURATE.AI)

This pilot feasibility study aims to set the foundation to investigate the applicability of QPOP drug selection followed by CURATE.AI-guided dose optimisation of the selected azacitidine combination therapy for solid tumours using CURATE.AI within the current clinical setting. QPOP will identify drug interactions towards optimal efficacy and cytotoxicity from the pre-specified drug pool based on ex vivo experimental data from the individual participant's tissue sample model. With these drug interactions, QPOP will identify the optimal drugs for the specific participant whose biopsy provided the cells for the ex vivo experimentation. Subsequently, CURATE.AI will be used to guide dosing for the selected combination therapy for that participant. Individualised CURATE.AI profiles will be generated based on each participant's response to a set of drug doses. Subsequently, the personalised CURATE.AI profile will be used to recommend the efficacy-driven dose. CURATE.AI will operate only within the safety range for each drug pre-specified for each participant. This pilot feasibility study will inform the investigators on the logistical and scientific feasibility of performing a large-scale randomised controlled trial (RCT) with the selected azacitidine combination therapy regimens and response markers. A secondary objective is to collect toxicity and efficacy data using established and exploratory response markers within and in-between cycles as exploratory outcomes.

Several drug combinations and modulation in drug dosing are given to promote cancer cell elimination in cancer patients. While advances in omics tools have led to greater understanding of the complexity of diseases such as cancer, they have also led to the understanding that large networks of molecular interactions contribute to both disease progression and therapeutic resistance. The rational design of drug combinations is a challenge because complex molecular networks contribute to feedback mechanisms of drug resistance and compensatory oncogenic drivers that limit the efficacy of targeted inhibitors. This challenge is compounded by the vast number of available drugs to identify optimal drug combinations from. In addition to the complexities in identifying optimal drug combinations, optimal dosing remains a challenge as drug synergy is both dose, time- and patient- dependent. The final drug concentration in the body must fall within a narrow range that maximises cancer elimination while minimizing toxic side effects. The complexity of this task increases significantly with the number of drugs given in combination due to increasing parameters and stochastic behaviour of a biological system. Currently, the established approach is to select maximum tolerated doses (MTD) - the highest drug doses that do not cause unacceptable side effects. Treatment efficacy does not guide dose selection. Combined with limited personalisation, this dosing strategy often results in sub-optimal outcomes of the treatment. In this pilot feasibility study, participants will undergo QPOP drug selection, a stage of CURATE.AI profile generation, and a stage of CURATE.AI profile-based, efficacy-driven drug dosing. As there are no prior clinical trial cohorts using CURATE.AI in participants with solid tumours and there are existing data for breast and gastric cancer for input into QPOP, this feasibility pilot study will focus on the practicality and feasibility of using QPOP and CURATE.AI in this clinical context. At the end of the participation of the first 10 patients, an interim analysis will be conducted using the data generated from these participants, which will include formal power and statistical sample size calculations. Based on these outcomes, the investigators will consider cohort expansion or an RCT. Specifically, the interim analysis will aid the decisions on whether to proceed with future RCTs; their design (superiority, equivalence or non-inferiority); logistical and practical aspects of running a large-scale RCT; patient population selection for the RCT; and potential applicability of CURATE.AI in a wider range of systemic therapy regimens, response markers and/or expansion of the current cohort to elicit further data on secondary endpoints and/or new randomized cohorts. Although not standard of care treatment for breast and gastric cancer, azacitidine combination therapy is chosen by the investigators as azacitidine combination therapy as azacitidine is a potent DNA methyltransferase inhibitor (DNMT) that can increase the sensitivity of a range of metastatic or advanced solid tumours, such as breast and gastric cancer, to treatment with docetaxel, paclitaxel, or irinotecan after developing resistance. Studies have also demonstrated the possibility of low dose treatment with chemotherapeutic agents when given together with azacitidine. However, cytotoxicity of azacitidine increases with dose and exposure time, which highlights the need to rapidly identify the optimal azacitidine-containing drug combinations and for personalised dose modulation during treatment. As such, QPOP drug selection and CURATE.AI dose modulation pipeline is in the ideal position to optimise treatment with azacitidine in combination with docetaxel, paclitaxel, or irinotecan via a personalised manner to maximise efficacy while minimising toxicities. Participants will undergo QPOP drugs selection optimisation, and those participants who are identified via QPOP to potentially benefit from azacitidine in combination with docetaxel, paclitaxel, or irinotecan will transition to the CURATE.AI stage of the trial after treatment fails. Participants who have undergone QPOP drug selection (e.g. under QGAIN (2019/00924) or NGAIN trial (2021/00009)) are allowed to enrol for the CURATE.AI modulation period of this study at the approval of the Principal Investigator and Sponsor. CURATE.AI will facilitate personalised treatment to each of the participants by recommending optimal doses in a dynamic fashion. In this phase, only azacitidine dose in the selected azacitidine combination will be modulated by CURATE.AI. Criteria for recruitment allow a high variability in the participant population to reflect a true variability in the cases faced in the clinical practice.

QPOP, Cancer, Artificial Intelligence, CURATE.AI, Chemotherapy, Azacitidine, docetaxel, paclitaxel, irinotecan

10 participants diagnosed with breast or gastric cancer undergoing or planned for treatment will be enrolled. The participants will undergo QPOP drug selection optimisation, and those participants who are identified via QPOP to potentially benefit from azacitidine in combination with docetaxel, paclitaxel or irinotecan will transition to the CURATE.AI dose modulation phase of the trial. Only azacitidine dose in the selected azacitidine combination will be modulated by CURATE.AI

Participants will undergo two study stages: QPOP drug selection and CURATE.AI-guided dosing modulation. In the QPOP drug selection stage, participants will undergo a baseline biopsy for organoid generation and subsequently receive treatment as per SOC. During this time, QPOP will identify optimal drug combinations for the participant based on ex vivo experiments on the participant's organoid. Participants who are identified via QPOP to potentially benefit from azacitidine in combination therapy (azacitidine + docetaxel, azacitidine + paclitaxel or azacitidine + irinotecan) will move on to the CURATE.AI-guided dosing modulation stage with treatment with azacitidine. Azacitidine treatment will begin once disease progression after SOC treatment is determined based on CT scans. Only azacitidine dose in the selected azacitidine combination will be modulated by CURATE.AI

QPOP is a mechanism-agnostic platform for optimizing drug selection. QPOP uses a quadratic equation to describe the patient-specific drug-drug interaction space based solely on experimentally derived drug-response data on individual patient's tissue sample, from which optimal drug combinations can be identified. Drug selection via QPOP allows for identification of an optimal combination therapy without the need for exhaustive testing of every combination. The first stage of the trial aims to generate a personalised QPOP drugs list for each participant based on experimentally derived data from ex vivo testing in the participant's tissue sample. Optimal drugs from a pre-specified drug pool will be recommended by the QPOP team.

CURATE.AI - a small data, AI-derived technology platform based on this discovery - allows personalised guidance of an individual's dose modulations based only on that individual's data. Additionally, CURATE.AI is mechanism-independent, and dynamically adapts to changes experienced by the participant, providing dynamic dose optimisation throughout the duration of the participant's treatment. The second stage of the trial aims to obtain a personalised CURATE.AI profile for each participant, based on their phenotypic response to a set of drug doses from the drug combinations with azacitidine identified and recommended by the QPOP team. The doses will be recommended by the CURATE.AI team, when relevant to the clinical decision-making process. Once an actionable profile is obtained, dose recommendations are based on the profile and aimed to treat the participant. The maximum period of involvement with this study when azacitidine may be adjusted by CURATE.AI is 18 months.

Azacitidine subcutaneous injection Day 1, 2 and Day 8, 9 + 30 mg/m2 docetaxel intravenously Day 1 and 8 Each chemotherapy cycle will be 21 days.

Azacitidine subcutaneous injection Day 1, 2 and Day 8, 9 + 80 mg/m2 paclitaxel intravenously Day 1 and 8 Each chemotherapy cycle will be 21 days.

Azacitidine intravenously subcutaneous injection Day 1, 2 and Day 8, 9 + 100 mg/m2 irinotecan intravenously Day 1 and 8. Each chemotherapy cycle will be 21 days.

A decision on whether we "successfully apply" the QPOP drug selection requires expert judgement and cannot be made based on a purely numerical process. The expert panel will consider the following factors with careful regard for the individual circumstances of each participant: The goodness-of-fit of the QPOP derived equation is acceptable to allow for a reliable list of effective drug combinations; The drugs list is generated sufficiently early for the participant to potentially benefit

CURATE.AI applicability: percentage of participants in whom the investigators successfully apply CURATE.AI profile.

A decision on whether we "successfully apply" the CURATE.AI profile requires expert judgement and cannot be made based on a purely numerical process. The expert panel will consider the following factors with careful regard for the individual circumstances of each participant: Error/variance (biological/analytical) is sufficiently small to allow accurate predictions; Profile can be generated sufficiently early for the participant to potentially benefit; Dose-dependent relationship is observed; Profile is actionable (i.e. fulfils the co-investigator's pre-specified safety requirements); Systemic changes in the participant which require profile recalibration are rare or readily assimilated into the CURATE.AI algorithm.

Physician adherence: percentage of QPOP recommended drug combinations that were used by the co-investigator.

Patient adherence: percentage of participants who always adhered to the prescribed dose whenever they took their medication, as measured by the standardised pharmacovigilance protocol.

Timely delivery of CURATE.AI recommendations to the clinician: percentage of CURATE.AI recommendations provided in time for the next chemotherapy cycle, across all participants and cycles.

CURATE.AI relevance: percentage of dosing events across all participants and cycles in which CURATE.AI recommendation is considered in the clinical decision-making process

Physician adherence: percentage of CURATE.AI recommended doses that were used by the co-investigator.

percentage of participants in whom the CURATE.AI-guided cumulative dose is substantially (≥10%) different from the projected SOC cumulative dose, which is defined as the maximum dose of the modulated drug*, azacitidine, multiplied by the number of completed chemotherapy cycles. * The maximum dose of the modulated drug azacitidine is 120mg/m2 once daily given on days 1-2 and days 8-9, every 21 days, in combination with weekly docetaxel, paclitaxel or irinotecan for two weeks followed by one week of rest.

Toxicity: percentage of trial participants with clinically relevant toxicities of grades 3-4 based on CTCAE version 4.0.

Data collection and explorative analysis of response marker in higher frequency serial measurements after modulated dosing in relation to standard frequency readings and other efficacy measures, e.g RECIST criteria

Data collection and explorative analysis of ctDNA as: a response marker in serial measurements at given clinical context and after modulated dosing; potential input for CURATE.AI.

Inclusion Criteria: Males and females ≥ 21 years of age. Eastern Cooperative Oncology Group (ECOG) Performance Status of 0 to 2. Patients must meet the following clinical laboratory criteria within 21 days of starting treatment: Absolute neutrophil count (ANC) ≥ 1,000/mm3 and platelet ≥ 50,000/mm3 Total bilirubin ≤ 1.5 x the upper limit of the normal range (ULN). Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) ≤ 3 x ULN of ≤ 5 ULN if involvement of the liver. Calculated creatinine clearance ≥ 30 mL/min or creatinine < 1.5 x ULN. Diagnosed with breast or gastric cancer, where docetaxel, paclitaxel or irinotecan is indicated for palliative therapy. Patients who have undergone QPOP drug screen (e.g. under QGAIN (2019/00924) or NGAIN trial (2021/00009) where the drug screen indicated potential benefit of combining azacitidine with taxane or irinotecan. Patients must have raised response marker above upper limit of local laboratory normal (e.g. CEA and/or CA19-9, CA 15-3, CA 125, AFP, and methylation markers such as but not limited to DNMT). Exclusion Criteria: Patients who are lactating or pregnant. Patients with clinically significant hypersensitivity to one or more of the selected regimen's constituent drug(s) (e.g. patients with clinically significant hypersensitivity to irinotecan may not be enrolled on azacitidine + irinotecan, but may be allowed on azacitidine + paclitaxel or azacitidine + docetaxel). Contraindication to any of the required concomitant drugs or supportive treatments. Any clinically significant medical disease or psychiatric condition that, in the co-investigator's opinion, may interfere with protocol adherence or a subject's ability to give informed consent. Major surgery within 28 days prior to start of the treatment. Active congestive heart failure (New York Heart Association [NYHA] Class III or IV), symptomatic ischaemia, or conduction abnormalities uncontrolled by conventional intervention. Myocardial infarction within 4 months prior to informed consent obtained. Patients who previously underwent chemotherapy treatment with either docetaxel, paclitaxel and/or irinotecan may still be able to enrol into treatment with the same drug in combination with azacitidine provided they fulfil all other criteria and approval is sought by PI and Sponsor (e.g. patients previously treated with paclitaxel and are enroling for treatment with paclitaxel + azacitidine). Patients with clinical suspicion or diagnosis of Gilbert's syndrome will not be allowed to enrol with azacitidine + irinotecan, but may be allowed to enrol for treatment with azacitidine + docetaxel or azacitidine + paclitaxel provided they fulfil all other criteria.

Rozenblatt-Rosen O, Deo RC, Padi M, Adelmant G, Calderwood MA, Rolland T, Grace M, Dricot A, Askenazi M, Tavares M, Pevzner SJ, Abderazzaq F, Byrdsong D, Carvunis AR, Chen AA, Cheng J, Correll M, Duarte M, Fan C, Feltkamp MC, Ficarro SB, Franchi R, Garg BK, Gulbahce N, Hao T, Holthaus AM, James R, Korkhin A, Litovchick L, Mar JC, Pak TR, Rabello S, Rubio R, Shen Y, Singh S, Spangle JM, Tasan M, Wanamaker S, Webber JT, Roecklein-Canfield J, Johannsen E, Barabasi AL, Beroukhim R, Kieff E, Cusick ME, Hill DE, Munger K, Marto JA, Quackenbush J, Roth FP, DeCaprio JA, Vidal M. Interpreting cancer genomes using systematic host network perturbations by tumour virus proteins. Nature. 2012 Jul 26;487(7408):491-5. doi: 10.1038/nature11288.

Chandarlapaty S. Negative feedback and adaptive resistance to the targeted therapy of cancer. Cancer Discov. 2012 Apr;2(4):311-9. doi: 10.1158/2159-8290.CD-12-0018. Epub 2012 Mar 22.

Logue JS, Morrison DK. Complexity in the signaling network: insights from the use of targeted inhibitors in cancer therapy. Genes Dev. 2012 Apr 1;26(7):641-50. doi: 10.1101/gad.186965.112.

Norton L, Simon R. The Norton-Simon hypothesis revisited. Cancer Treat Rep. 1986 Jan;70(1):163-9. No abstract available.

Peters WP, Rosner GL, Vredenburgh JJ, Shpall EJ, Crump M, Richardson PG, Schuster MW, Marks LB, Cirrincione C, Norton L, Henderson IC, Schilsky RL, Hurd DD. Prospective, randomized comparison of high-dose chemotherapy with stem-cell support versus intermediate-dose chemotherapy after surgery and adjuvant chemotherapy in women with high-risk primary breast cancer: a report of CALGB 9082, SWOG 9114, and NCIC MA-13. J Clin Oncol. 2005 Apr 1;23(10):2191-200. doi: 10.1200/JCO.2005.10.202. Epub 2005 Mar 14.

Frei E 3rd, Elias A, Wheeler C, Richardson P, Hryniuk W. The relationship between high-dose treatment and combination chemotherapy: the concept of summation dose intensity. Clin Cancer Res. 1998 Sep;4(9):2027-37.

Rashid MBMA, Toh TB, Hooi L, Silva A, Zhang Y, Tan PF, Teh AL, Karnani N, Jha S, Ho CM, Chng WJ, Ho D, Chow EK. Optimizing drug combinations against multiple myeloma using a quadratic phenotypic optimization platform (QPOP). Sci Transl Med. 2018 Aug 8;10(453):eaan0941. doi: 10.1126/scitranslmed.aan0941.

de Lima M, Giralt S, Thall PF, de Padua Silva L, Jones RB, Komanduri K, Braun TM, Nguyen HQ, Champlin R, Garcia-Manero G. Maintenance therapy with low-dose azacitidine after allogeneic hematopoietic stem cell transplantation for recurrent acute myelogenous leukemia or myelodysplastic syndrome: a dose and schedule finding study. Cancer. 2010 Dec 1;116(23):5420-31. doi: 10.1002/cncr.25500. Epub 2010 Jul 29.

Jabbour E, Giralt S, Kantarjian H, Garcia-Manero G, Jagasia M, Kebriaei P, de Padua L, Shpall EJ, Champlin R, de Lima M. Low-dose azacitidine after allogeneic stem cell transplantation for acute leukemia. Cancer. 2009 May 1;115(9):1899-905. doi: 10.1002/cncr.24198.

Singal R, Ramachandran K, Gordian E, Quintero C, Zhao W, Reis IM. Phase I/II study of azacitidine, docetaxel, and prednisone in patients with metastatic castration-resistant prostate cancer previously treated with docetaxel-based therapy. Clin Genitourin Cancer. 2015 Feb;13(1):22-31. doi: 10.1016/j.clgc.2014.07.008. Epub 2014 Aug 1.

Cohen AL, Ray A, Van Brocklin M, Burnett DM, Bowen RC, Dyess DL, Butler TW, Dumlao T, Khong HT. A phase I trial of azacitidine and nanoparticle albumin bound paclitaxel in patients with advanced or metastatic solid tumors. Oncotarget. 2016 Dec 26;8(32):52413-52419. doi: 10.18632/oncotarget.14183. eCollection 2017 Aug 8.

Lee V, Wang J, Zahurak M, Gootjes E, Verheul HM, Parkinson R, Kerner Z, Sharma A, Rosner G, De Jesus-Acosta A, Laheru D, Le DT, Oganesian A, Lilly E, Brown T, Jones P, Baylin S, Ahuja N, Azad N. A Phase I Trial of a Guadecitabine (SGI-110) and Irinotecan in Metastatic Colorectal Cancer Patients Previously Exposed to Irinotecan. Clin Cancer Res. 2018 Dec 15;24(24):6160-6167. doi: 10.1158/1078-0432.CCR-18-0421. Epub 2018 Aug 10.

Sharma A, Vatapalli R, Abdelfatah E, Wyatt McMahon K, Kerner Z, A Guzzetta A, Singh J, Zahnow C, B Baylin S, Yerram S, Hu Y, Azad N, Ahuja N. Hypomethylating agents synergize with irinotecan to improve response to chemotherapy in colorectal cancer cells. PLoS One. 2017 Apr 26;12(4):e0176139. doi: 10.1371/journal.pone.0176139. eCollection 2017. Erratum In: PLoS One. 2020 Nov 30;15(11):e0242750.

Moro H, Hattori N, Nakamura Y, Kimura K, Imai T, Maeda M, Yashiro M, Ushijima T. Epigenetic priming sensitizes gastric cancer cells to irinotecan and cisplatin by restoring multiple pathways. Gastric Cancer. 2020 Jan;23(1):105-115. doi: 10.1007/s10120-019-01010-1. Epub 2019 Sep 25. Erratum In: Gastric Cancer. 2019 Nov 15;:

Li LH, Olin EJ, Buskirk HH, Reineke LM. Cytotoxicity and mode of action of 5-azacytidine on L1210 leukemia. Cancer Res. 1970 Nov;30(11):2760-9. No abstract available.