Recognition of altered gene-gene interaction using BiLSTM in different stages of lung adenocarcinoma.

Cancer which is caused by abnormal growth or tumour snatches away millions of lives each year. Depending on the organ affected, we find various cancer types like blood, skin, lung, kidney, breast, prostrate, liver etc. Some of these are curable if diagnosis of the disease can be made early. Others like pancreatic cancer are very difficult to cure. Hence early detection is one of most important aspects of treatment. Cancer affected patient goes through several stages of the disease. There are usually four stages of cancer. The Stage 1 of cancer being the preliminary stage and stage 4 being the last stage of the disease. It has been observed that the deadliness of the disease aggravates with each stage. There is no fixed or well defined time limit for the change in the stages. It may depend on the type of cancer, the physical attributes of the patient and on external factors and life-style habits. It is usually observed that symptoms change with change in stages of the disease. We intend to find out whether there is any change in the overall relationship between the genes amongst the various stages of lung cancer patients. Long Short Term Memory (LSTM), Bidirectional Long Short Term Memory (BiLSTM) and Feed Forward Neural Network (FNN) are used for analyzing the gene dependencies. The comparative analysis of the results obtained is done using various statistical methods of correlation and regression. Analysis of the results reveal significant variance in correlation among the various stages of lung cancer, showing Stage 2 and Stage 3 to be the most strongly correlated and Stage 1 and Stage 2 being the most weakly correlated indicating the inconsistency of genetic relationship between the various stages of lung cancer. [ABSTRACT FROM AUTHOR]