Artificial intelligence based models for predicting head and neck cancer from genomics data of single cells

Abstract

Head and Neck Squamous Cell Carcinoma (HNSC) or Head and Neck Cancer is the sixth most highly prevalent cancer type worldwide. Early detection of HNSC is one of the important challenges in managing the treatment of cancer patients. Existing techniques for detecting HNSC are costly, expansive, and invasive in nature. In this study, an attempt has been made to develop classification models using machine and deep learning techniques to discriminate HNSC and normal samples. In addition. models have been developed to predict HPV associated HNSC samples. All models in this study have been developed on two datasets (GSE181919 and GSE139324) of single-cell genomics obtained from RNA-seq technology. These models were trained on the training dataset and validated on internal and external datasets. Our deep learning models outperform machine learning models in the prediction of HNSC samples on bother datasets. We further classified these HNSC samples into HPV associated and HPV non-associated HNSC samples with high precision. In summary, this is a pilot study to understand the application of single-cell genomics in predicting HNSC and its type.