A hybrid feature subset selection algorithm for analysis of high correlation proteomic data

Area of Science:

• Biomarker discovery
• Proteomics
• Mass spectrometry

Background:

• Proteomic patterns in biological fluids can indicate organ pathologies.
• Surface-enhanced laser desorption and ionization time-of-flight mass spectrometry (SELDI-TOF MS) generates proteomic profiles.
• Mass spectrometry data often contains redundant and irrelevant features for cancer diagnosis.

Purpose of the Study:

• To develop a hybrid feature subset selection algorithm for identifying ovarian cancer biomarkers.
• To improve the accuracy and reproducibility of biomarker selection from SELDI-TOF MS data.

Main Methods:

• A hybrid algorithm combining maximum-discrimination, minimum-correlation, and peak scoring was proposed.
• The algorithm was applied to two independent SELDI-TOF MS ovarian cancer datasets.
• Linear discriminate analysis was used to identify important biomarkers.

Main Results:

• The algorithm extracted a set of potential protein biomarkers from each dataset.
• Selected biomarkers achieved 100% accuracy, sensitivity, and specificity in diagnosing ovarian cancer.
• The hybrid algorithm demonstrated high discrimination power and reproducibility.

Conclusions:

• The proposed hybrid algorithm is effective for selecting robust protein biomarkers for ovarian cancer detection.
• This approach enhances the reliability of proteomic data analysis for clinical applications.
• Accurate diagnosis of ovarian cancer is achievable with a small set of highly discriminative protein biomarkers.