Machine learning and explainable artificial intelligence reveals the MicroRNAs associated with survival of head and neck squamous cell carcinoma patients.

Dysregulated microRNAs (miRNAs) play a significant role in cancer development and metastasis. In literature, miRNAs have been used for the survival prediction of different types of cancers using AI. Although AI is useful for diagnosis and prognosis prediction of cancer, however, a major criticism of incorporating it into medical fields is that it is essentially a mechanistically uninterpretable opaque "black box", and hence it may not have the required level of accountability, transparency, and reliability in decisions of cancer diagnosis and prognosis for their adoption in clinical settings. Therefore, there is need to develop intelligent models which may explain their prediction so that they may be reliably used by the clinicians. As dysregulated miRNAs are reported to cause cancer metastasis hence, they can play role in survival of patient. Therefore, there is needed to develop ML based techniques which may automatically indicate specific miRNAs involved in survival of patients. In this research, Machine Learning and Explainable AI (XAI) based models have been developed for survival prediction of Head and Neck Squamous Cell Carcinoma (HNSC) patients using miRNA sequences and clinical datasets. miRNAs dataset contains the data of 485 HNSC patients and clinical dataset contains data of 528 patients. The proposed XAI based model explains its prediction by showing the specific miRNA sequences involved in survival of the patients to demonstrate its reliability to be used by clinicians for therapeutic decisions. In this study, it has been shown that explainable ML can provide explicit knowledge of how models make their predictions, which is necessary for increasing the trust and adoption of innovative ML techniques in oncology and healthcare.