MILP_k: Mixed Integer Linear Programming for multiple-biomarker panel identification

• Version 0.0.1
• Last updated: May 6, 2015.

Motivation#

Multi-biomarker panels can capture the nonlinear synergy among biomarkers and they are important to aid in the early diagnosis and ultimately battle complex diseases. However, the multi-biomarker panel identification is challenging from case and control data. For example, the exhaustive search method is computationally infeasible when the data dimension is high. Here, we aim to propose a novel method to identify multi-biomarker panel and apply it to distinguish colorectal cancers (CRC) from benign colorectal tumors based on serum measurement.

Method#

We developed a novel mixed integer programming model for multi-biomarker panel detection. This model directly minimizes the classification error (maximizes the classification accuracy) given the number of biomarkers in the optimal multi-biomarker panel. This mixed integer programming model allows us to go through all the optimal combinations by varying parameter from 1 to n. Moreover, we can check their accuracy and compare the selected combinations. In particular, an optimal multi-biomarker panel can be selected by balancing the parameter k and the classification accuracy.

Software#

CPLEX Optimizer is required in your MATLAB. CPLEX is available on the website: http://www-01.ibm.com/software/commerce/optimization/cplex-optimizer/.

This version of the program is in very preliminary stage and provided just for testing purpose. The program is still under development.

• source.rar

Dataset#

The dataset used in our paper serves as an example to demonstrate the implementation for MILP_k. The raw data(.xlsx) and processed data(.mat) are available as follows.

• data.rar

References#

• Meng Zou, Peng-Jun Zhang, Xin-Yu Wen, Luonan Chen, Ya-Ping Tian, and Yong Wang, A novel mixed integer programming for multi-biomarker panel identification by distinguishing malignant from benign colorectal tumors. Methods, In submission.