Data Mining of Pancreatic Cancer Protein Databases

摘要

Data mining of protein databases poses special challenges because many protein databases are non-relational whereas most data mining and machine learning algorithms assume the input data to be a type of relational database that is also representable as an ARFF file. We developed a method to restructure protein databases so that they become amenable for various data mining and machine learning tools. Our restructuring method enabled us to apply both decision tree and support vector machine classifiers to a pancreatic protein database. The SVM classifier that used both GO term and PFAM families to characterize proteins gave us over 73% accuracy in predicting whether a protein is involved in pancreatic cancer.