PRELIMINARY EVALUATION THROUGH THE GRADIENT BOOSTING ALGORITHM (GBM) OF CD146 OVEREXPRESSION IN SPORADIC COLORECTAL CANCER PATIENS

LOHMANN, F.A ; ROSALEZ, P.; CAJAL, A.R.; SANTINO, J.P.; RISK, M.R.; VACCARO C.A.; PIÑERO, T. A.

Mar del Plata

Otro; REUNIÓN CONJUNTA SAIC SAB AAFE AACYTAL 2023; 2023

Sociedad Argentina de Investigación Clínica

Colorectal cancer (CRC) ranks as the third most prevalent malignantneoplasm globally and stands as the third leading oncological causeof death in Argentina. The projected global incidence of CRC for theyear 2030 is expected to reach 2.2 million new cases, accompaniedby an anticipated 1.1 million associated deaths. This underscoresthe critical need to discover novel biomarkers capable of enhancing therapeutic strategies. CD146 has emerged as a biomarker of angiogenesis within the tumor microenvironment across variouscancer types. Accumulated evidence strongly suggests that CD146overexpression correlates with the early development of the primary lesion or progression to metastasis. The aim of this study wasto evaluate the relationship between CD146 overexpression andanatomoclinicopathological, as well as epidemiological variables inCRC. The degree of CD146 overexpression was assessed withinformalin-fixed, paraffin-embedded (FFPE) tissue samples obtainedfrom 41 surgically treated sporadic CRC patients at the Hospital Italiano de Buenos Aires, Argentina. CD146 overexpression was detected in 56% (23/41) of cases and exhibited a correlation with histological grade and Duke stage (χ²=16.64; P=0.05). By employingthe Gradient Boosting Machine (GBM) machine learning algorithm,commonly utilized for statistical classification and regression tasks,a predictive model was constructed, consisting of an ensemble ofweak prediction models, typically represented as decision trees.Through this analysis, variables such as tumor site (38.5%), bodymass index (BMI (21.7%), cancer age (13.7%), sex (8.1%), and thepresence of colorectal polyps during the first 6 and 12 months of follow-up were identified as factors exerting more influence on CD146expression. CD146 shows promise as a consequential prognosticbiomarker in CRC, emphasizing the necessity of jointly assessingwith patient progression as a prospective tool of precision medicine.