Improving the Inference of Gene Expression Regulatory Networks with Data Aggregation Approach

The major issue for the future of bioinformatics is the design of tools to determine the functions and all products of single-cell genes. This requires the integration of different biological disciplines as well as sophisticated mathematical and statistical tools. This study revealed that data mining techniques can be used to develop models for diagnosing high-risk or low-risk lifestyles for colorectal cancer.

In this retrospective study, a dataset consisting of information relevant to 84 patients and 225 healthy individuals with 25 attributes was collected. This information was on patients diagnosed from 2006 to the first quarter of 2014. The most widely used techniques in the medical informatics literature including support vector machine, Naive Bayes, decision tree, and k-nearest neighbor were used to develop the models.

The developed models are able to distinguish peoplechr('39')s lifestyles efficiently. A well-developed non-technical measure can properly determine the true value of individual predictions, whether true or false, at actual costs, and indicate a true measure of the cost savings in the health system by each model. Among the developed models, only two models were able to meet the criteria set for use in the real world.

The developed models should not only be technically evaluated, but should also be examined in terms of metrics accepted for the medical field as well as feasibility for real problem solving.