Discriminating head trauma outcomes using machine learning and genomics.

Area of Science:

• Neurogenetics
• Computational Biology
• Genomic Medicine

Background:

• Persistent post-concussion symptoms (PCS) and severe neurological dysfunction affect individuals after head trauma.
• Genetic factors influencing head trauma response are not fully understood.
• Previous genetic studies on concussion outcomes have yielded inconclusive results.

Purpose of the Study:

• To explore the utility of machine learning (ML) methods combined with whole exome sequencing (WES) data for predicting head trauma outcomes.
• To identify genetic signatures associated with severe neurological responses, persistent PCS, or normal recovery from concussion/mild traumatic brain injury (mTBI).

Main Methods:

• Whole exome sequencing (WES) was performed on 60 individuals across three groups: severe response, persistent PCS, and normal recovery.
• Gradient boosted tree algorithms (XGBoost) were applied to genomic data.
• Variants with CADD scores > 15 were utilized in a 70% training set for model development.

Main Results:

• The ML model achieved an average area under the curve (AUC) of 0.8 (SE=0.019) in distinguishing between the groups.
• High accuracy was supported by acceptable positive/negative prediction values and kappa metrics.
• The study demonstrated strong discrimination trends using exome data from cases and controls.

Conclusions:

• ML methods combined with WES data show potential for predicting severe or prolonged responses to head trauma.
• Non-linear ML approaches, like boosted trees, offer valuable insights in smaller sample sizes compared to linear association analyses.
• Genomic data holds promise for personalized prediction of concussion recovery trajectories.