Big data and computational biology strategy for personalized prognosis

Area of Science:

• Bioinformatics
• Genomics
• Precision Medicine

Background:

• Big data and precision medicine generate vast amounts of patient gene expression and clinical data.
• Personalized risk and treatment efficacy prediction requires robust patient similarity analysis.

Purpose of the Study:

• To develop a novel methodology for predicting disease/treatment outcomes by analyzing patient similarity.
• To assess the effectiveness of Euclidean distance versus correlation distance for patient similarity measurement.

Main Methods:

• Patients characterized by biological variables (biomarkers/clinical features) represented as prognostic signature vectors (PSVs).
• Euclidean distance used to measure similarity between PSVs.
• Methodology applied to high-grade serous ovarian cancer (HGSC) using a 36-mRNA predictor and patient age.

Main Results:

• Euclidean distance proved effective for unbiased similarity measurement between PSVs.
• Patient age, as a binary variable, positively correlated with disease risk in HGSC.
• Including age in the molecular predictor improved personalized prognosis and identified therapeutic benefits for HGSC patients.

Conclusions:

• The proposed patient similarity methodology accurately predicts personalized outcomes.
• Incorporating clinical variables like age alongside molecular data enhances prognostic accuracy.
• The method is generalizable to other diseases for personalized outcome prediction.