Читать книгу Machine Learning Techniques and Analytics for Cloud Security - Группа авторов - Страница 66

The proposed method PC-LR uses a hybrid approach which is chosen as a ML technique to generate the target output. As our task is to select the driver genes which are related to certain cancers, so it is better to design an algorithm which can act as a binary classifier and identify the relevant genes. It is to be noted that LR always works well for this model. But the gene expression data is of huge volume and so to get rid of the curse of dimensionality is mandatory before start working with LR. This is done using PCA. Our implemented approach has established a group of genes very precisely that are expressed differentially and are correlated to some cancers. The experimentation is executed over two datasets, viz., colon and lung, and has determined a gene set. The creativity and robustness of the system is clearly defined. It is to be noted that mutations of genes might have correlations among themselves and it may or may not vary with different stages of cancer. So, identifying these is also a challenging task. Those genes can effectively be examined by research scientists and biologists for the purpose of laboratory testing by focusing on less number of genes instead of whole genome.

Our work is having scope of extension in future for identifying more genes which might be correlated to mutations. Further identifying interactions among those genes can be very much helpful in prognosis, cancer prevention and treatment. Analyzing interactions of Gene-Gene will be beneficial for finding out more TP genes having key role for mediating cancer. The extension of our study using other omics data might help researchers and biologists to concentrate on cancer study in a targeted way.