The Spanish company specialized in artificial intelligence (AI), Topazium, in collaboration with Biomakers, has presented at a meeting of the prestigious American Society of Clinical Oncology how AI can accelerate the identification of new therapeutic targets to improve the effectiveness of immunotherapy in the treatment of non-small cell lung cancer, thereby optimizing treatment based on the molecular characteristics of the tumor.
In the field of AI applications in medicine, Topazium has partnered with Biomakers, a precision oncology company with a comprehensive genomic platform that aims to transform current healthcare. This collaboration has focused on the study of PD-1/PD-L1, with the goal of optimizing treatment protocols and identifying which patients are most likely to benefit from this therapy. For this purpose, whole sequencing was performed on 46 biopsy samples from patients with NSCLC (squamous cell carcinoma and adenocarcinoma subtypes). The data were analyzed using the GFPrint software, a crucial component of the study. This machine learning framework can virtually represent tumor exomes in a latent space; in other words, it expresses the entirety of the information obtained in a simplified manner without altering its content, facilitating the analysis of large volumes of genetic data while preserving relevant details such as tumor mutations. Once the data were represented, the AI identified groups of patients based on the genetic similarity of their tumors. Biostatistical analyses identified two patient groups (C0 and C1). These groups differed mainly by tumor histological characteristics and PD-L1 expression. The C0 group consisted of significantly fewer patients with adenocarcinoma, but with significantly higher PD-L1 positivity in both carcinomas and adenocarcinomas. Moreover, PD-L1 expression was three times higher in C0 patients (6.3±9.7%) compared with C1 patients (2.3±7.5%). A total of 1,357 genes were found to be exclusively mutated in the C0 group, belonging to three signaling pathways: HIF-1, Ca2+ signaling, and actin cytoskeleton regulation. Additionally, 10 genes involved in HIF-1 signaling were identified that appear to play a role in PD-L1 expression according to previous evidence. These genes could represent targets for pharmacological intervention concomitant with PD-1/PD-L1 blockade, as well as potential markers for selecting patients most likely to respond to immunotherapy in NSCLC treatment.
This study not only identified potential markers and targets for NSCLC treatment, but also highlighted the capability of artificial intelligence in precision medicine, particularly in optimizing, analyzing, and interpreting large volumes of data. This will enable more precise and efficient decision-making in precision medicine in the future.
Read at: https://drive.google.com/file/d/1m9McScC0U5eYg4Dta2ySr5-CQ_JH4NE0/view
