Though their findings will need to be clinically validated, researchers using machine learning tools have identified 12 new candidate biomarkers for malignant pleural mesothelioma. The entirely computational research, published in the October 2026 issue of Cancer Reports, provides novel genetic clues that may potentially lead to more effective treatment for the rare, asbestos-related disease.
Biomarkers are Vital Targets for Mesothelioma Treatment
Mesothelioma is an aggressive asbestos-related cancer of the lining of the lungs. It is notoriously difficult to catch early, and median survival after diagnosis ranges from roughly 9 to 17 months, largely because the disease doesn’t become symptomatic until decades after exposure and because the disease has proven resistant to traditional treatments. Researchers have identified markers for the disease, including fibulin-3 and osteopontin, but they have not proven reliable enough to detect the disease early or distinguish it from other conditions.
The research that identified the previously unknown mesothelioma biomarkers was conducted by M. L. Rohman and S. Biswas, from the Department of Statistics and Data Science of Islamic University in Kushtia, Bangladesh. Using mesothelioma tissue data from a public gene expression database, the researchers identified more than 12,600 genes that behaved differently in cancerous tissue, then sorted them into networks of genes that rise and fall together. From those networks, they selected 12 “hub” genes tied to cell division, cell adhesion, and immune signaling, all of which are processes that cancer cells commonly hijack.
Mesothelioma Detection Model Shows Early Promise
Mesothelioma survival data from The Cancer Genome Atlas showed that 11 of the 12 genes were associated with how long patients lived. One gene of particular interest, PVRIG, acts as an immune checkpoint, and blocking it has strengthened immune responses in mouse studies. This strongly suggests that it could be a target for future immunotherapy research.
Mesothelioma tumors were distinguished from normal tissue with high accuracy when the 12-gene panel was tested in five machine learning models. The top performer, a method called extreme gradient boosting, scored 97.5 percent on a standard measure of diagnostic performance and correctly classified about 90 percent of samples. Still, these results were the result of strictly computational research and should be viewed as a starting point for further study rather than as a new test. The models were trained and evaluated on the same small dataset; the comparison tissue was normal lung rather than normal pleura, and details such as asbestos exposure history and tumor stage were unavailable. The authors are recommending larger independent studies and laboratory experiments to confirm the genes’ roles.
If you or someone you love has been diagnosed with mesothelioma, the Patient Advocates at Mesothelioma.net are here to help. Contact us today at 1-800-692-8608 to learn more.
