Discussion
Our exploratory pathway analysis highlighted oxidative stress responses and neurodevelopmental pathways as key processes perturbed in the DLPFC of suicides. Regarding ML models, KNeighborsClassifier was the best predicting conditions. Here we show that these proteins of the DLPFC may help to identify brain processes associated with suicide and they could be validated as potential biomarkers of this outcome.
Methods
Brain tissue (Brodmann area 9) was sampled from male cases (n=9) and age-matched controls (n=7). We analyzed the proteomic differences between the groups using two-dimensional polyacrylamide gel electrophoresis and mass spectrometry. Bioinformatics tools were used to clarify the biological relevance of the differentially expressed proteins. In addition, this information was analyzed using machine learning (ML) algorithms to propose a model for predicting suicide.
Results
Twelve differentially expressed proteins were also identified (t 14 ≤ 0.5). Using Western blotting, we validated the decrease in expression of peroxiredoxin 2 and alpha-internexin in the suicide cases. ML models were trained using densitometry data from the 2D gel images of each selected protein and the models could differentiate between both groups (control and suicide cases).