Artificial intelligence in proteomics: new frontiers from risk prediction to treatment?

Authors:
Publisher website
Google Scholar
Abstract
In cardiology, risk scores are used daily to estimate the cardiovascular disease (CVD) risk of a patient and eventually guide clinical decision making. To create accurate scoring systems, large populations are used, and clinically measurable parameters are included. These scores and calculations represent mostly linear and proportional relationships which correspond to the idea of ‘more A, therefore more B’. This allows for simple interpretability of causal relationships of risk factor and observed events, e.g. smoking or high blood pressure are harmful. Clinical risk models tend to perform well on average but show poor precision for individuals. There are subjects who despite having a high-risk profile will not develop significant CVD or experience events. Those ‘outliers’ may have a different clinical response to external conditions, e.g. heavy smoking, and still not develop any disease manifestation. Some may attribute this to...