Abstract 85P
Background
Osteosarcoma is an aggressive malignancy with a broad profile of drug resistance in patients who relapse or are refractory to first-line chemotherapy. Lacking specific therapeutic targets, alternative approaches are required to optimize antineoplastic therapy. By utilizing extensive datasets for drug sensitivity profiling based on neoplastic cell lines, we applied an artificial intelligence algorithm to explore new therapeutic approaches using transcriptomic data.
Methods
We analyzed the gene expression profile of 12 FFPE samples from pediatric osteosarcoma patients. The algorithm developed at Bio@SNS of the Scuola Normale Superiore leverages existing datasets (GDSC, PRISM) to create predictive models of drug sensitivity. The machine learning model was trained on transcriptomic data from cell lines to predict the growth inhibitory concentration (IC50). Language models were employed to expand descriptions of drug mechanisms of action, highlighting genes/pathways involved in drug response.
Results
We obtained sensitivity scores for 280 antineoplastic drugs, with the model identifying key genes involved in targeted pathways. The 12 samples were grouped based on their drug sensitivity profiles. To validate the model's predictions, representative samples with specific drug sensitivities will be identified, strengthening the model's reliability and clinical translatability.
Conclusions
The model represents a paradigm shift in osteosarcoma research, offering a powerful tool for future explorations. Based on the results and the study of the most common pathways targeted by the most effective drugs, we can compile a list of potential new drugs for in vitro testing on primary osteosarcoma cells and subsequent clinical translation. By utilizing these high-throughput platforms and broader drug sets on diverse patient cohorts, our findings promise advances in osteosarcoma treatment, offering new hope to patients and clinicians.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
Fondazione Pisana per la Scienza.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.