Key points are not available for this paper at this time.
e13614 Background: Functional screening methods are increasingly applied to guide patient treatment and could be beneficial in indications with poor survival and response rates, such as pancreatic cancer. However, challenges faced when implementing ex vivo testing include complexities in obtaining sufficient tumor cells for different cancer types and maximizing the number of treatments tested with limited sample material. The latter is especially significant in informing clinical decision making as combination therapies are common with cancer, exponentially increasing the number of regimens to select from. To address this hurdle, we developed Optim.AI, a combinatorial functional precision medicine platform, which utilizes efficiently designed experiments and small data AI to expand to over 530,000 phenotypic response data points and predictively rank all actionable treatments within a 12-drug panel. In this feasibility study, we explored the application of Optim.AI on patient-derived tumor cells in pancreatic cancer. Methods: We established a procedure to run Optim.AI with isolated tumor cells from pancreatic cancer patients. After dissociation, these cells were then incubated to allow for 3D organoid formation within the following week. Upon organoid formation, the cells were treated with an FDA-approved panel of chemotherapy and targeted drugs. Cell viability was quantified post-drug treatment for analysis with Optim.AI software to rank all possible single to 4-drug combinatorial therapies for report generation. Results: The viability of the cells post-processing was high (mean = 97.4% live cells), passing the quality check of having at least 60% live cells before proceeding with Optim.AI testing. The success rate of organoid formation was 100% within 1 week of sample processing. The turnaround time from the receipt of samples to dissemination of the reports to clinicians was approximately a week (mean = 7.67 days). Z’ factor, a measurement of statistical effect size for high throughput screening, was demonstrated to be more than 0.5 for the reports generated, indicative of very good assays. The top ranked drug combinations were sensitive, with their normalized cell viability reported to be under 0.2, while the prior line of treatment was shown to be resistant, 100% concordant with previously observed clinical outcomes. Notably, Gemcitabine with either a HDAC or PI3K inhibitor was ranked high amongst the reports generated. Conclusions: Optim.AI had a high success rate in generating reports within a very short timeframe for pancreatic cancer. Comparing the top ranked therapies against the prior lines of treatment allows the clinicians to better evaluate the alternative treatments. Based on this initial feasibility study, Optim.AI could potentially be viable in guiding clinicians in managing treatment options for pancreatic cancer patients.
Rashid et al. (Wed,) studied this question.