Machine learning-assisted detection of canine mammary tumors using serum autoantibody signatures.

Abstract

Canine mammary tumors (CMTs) are the most common neoplasms in intact female dogs, yet early detection remains challenging due to the lack of clinically validated, noninvasive biomarkers. This study aimed to develop a noninvasive diagnostic model for CMT detection by integrating serum autoantibody biomarkers with machine learning. Serum samples from 154 dogs with mammary tumors (31 benign, 123 malignant) and 39 healthy controls were analyzed using a custom multiplex immunoassay detecting autoantibodies against AGR2, HAPLN1, IGFBP5, and TYMS, normalized to anti-BSA levels. Median fluorescence intensity (MFI), standardized autoantibody ratios, and their combination, together with clinical variables, were used to train random forest classifiers. The model based on standardized autoantibody ratios achieved the best performance, with an AUC of 0.79 (sensitivity 75.3%, specificity 74.4%) for overall CMT detection; 0.78 (92.7%, 61.5%) for malignant CMTs; and 0.77 (82.2%, 71.8%) for early-stagemalignancies. Assuming a CMT prevalence of 0.5 in the hospital-referred population, the positive and negative predictive values ranged from 0.74-0.75 and 0.75-0.91, respectively. This proof-of-concept study demonstrates that a machine learning-assisted multiplex autoantibody assay offers a feasible noninvasive approach for CMT detection. Further validation in larger, independent cohorts is warranted to support clinical translation in veterinary oncology.