This repo contains the copy of "Personalized Medicine" projects' scripts. This project is aimed at developing of risk prediction models for patients with cardiovascular diseases.
Machine learning, cardiovascular deseases, biomarkers
Cardiovascular diseases are the leading cause of death worldwide, and accurately identifying individuals at risk is crucial for prevention and appropriate treatment. However, there is a need to identify specific risk factors tailored to the Russian population, which has been underrepresented in international research. In this study, predictive models were developed using machine learning on a Russian population dataset to predict future complications in patients. Eight key clinical parameters were identified as leading to increased risk. The addition of biomarkers, specifically PCSK9, improved the predictive accuracy, resulting in a Random Forest model with an AUROC of 0.957 in the train subset and 0.859 in the test subset for predicting combined cardiovascular death, myocardial infarction, or stroke within 1-2 years.
We are aiming:
- to analyze the medical data of patients with CVDs from Surgut Cardio Hospital;
- to identify biomarkers correlating to the risk factor increase;
- to compare primary risk biomarkers to previously described risk factors;
- to develop predictive machine learning models;
- to create guidelines for cardio medical community.
