Prediction And Research The Relationship Between Non-synonymous SNPs And Diseases

Many researches have opened out that non-synonymous SNPs (nsSNP, i.e. SAP) have curtailed linkage to genetic diseases, such as cardiovascular diseases, diabetes, cancer, etc. And nsSNP plays an important role in pharmacology field. Thus prediction of deleterious nsSNPs became a hot point in Genome Wide Association Studies and affords an approach to studying the mechanism of treatment in molecular level.We do a series of bio-information experiment to construct a quick, concise, accurate and faithful model to prediction the deleterious nsSNP based on analysis the relationships between human nsSNP and diseases by machine learning. According to extracted features which contain physicochemical, secondary structure, conservation of protein sequence etc. from disease nsSNPs and polymorphism nsSNPs, we built a mathematics model successfully. In addition it is satisfactory performance after partitioning the dataset according to the nsSNP site Original amino acids type before training the model by Support Vector Machine (SVM). As a control, we also partitioned the training sets into twenty subsets according to the nsSNP site Substitution amino acids type and randomly partitioned. Finally, after the selecting the most proper features by mRMR, we got well-behaved classifiers and a conclusion which is that partition the training set according to a proper way play a positive role in the region of nsSNP prediction research.