مشاريع طلاب خريف 2024 - F24
Prediction of the therapeutic response in psoriasis patients using artificial intelligence tools
Abstract
Aim: This study aims to develop a machine learning model based on DNA methylation profiles to predict Anti-TNF-α response in psoriasis patients, distinguishing responders from non-responders.
Materials and Methods: Using Google Colab, five machine learning models Random Forest (RF), Logistic Regression (LR), Decision Tree (DT), Support Vector Machine (SVM), and Multi-Layer Perceptron Classifier (MLPClassifier) were trained on DNA methylation data from 70 psoriasis patients. The cohort was stratified into:
- 49Anti-TNF-α responders (PASI improvement ≥90%)
- 21Anti-TNF-α non-responders (PASI improvement <70%)
The methylation dataset was sourced from the NCBI’s GEO database (Accession:[GSE151278]).
Results: Among the evaluated models, Random Forest (RF) exhibited the highest predictive performance, with a (CV accuracy of 0.750 and test-accuracy: 0.785, precision: 0.835, recall: 0.785, F1: 0.735).Notably, the three most influential variables in our model mapped to genomic loci where differential methylation patterns could potentially regulate the expression of genes encoding proteins directly implicated in psoriasis pathogenesis .
Conclusions: Our machine learning analysis of DNA methylation data identified Random Forest as the optimal predictor of anti-TNF-α response in psoriasis patients (79% accuracy). The top predictive loci were biologically relevant to psoriatic pathways, suggesting clinical potential for treatment stratification. Further validation in larger cohorts could enhance predictive utility.
إعداد: الطالبة نادين الخوري
إشراف: الدكتور لؤي صالح
Prediction of the therapeutic response in psoriasis patients using artificial intelligence tools
Computational design of targeted antibody for TMEM106B filaments in neurodegenerative diseases using bioinformatics approaches
TMEM106B is a lysosomal transmembrane protein expressed in central nervous system neurons and has been implicated in the pathogenesis of neurodegenerative disorders such as FTLD-TDP43. Recent studies have also identified amyloid fibrils of TMEM106B in affected patients. In this work, the three-dimensional structure of TMEM106B was retrieved from the Protein Data Bank, and a potential epitope was predicted using ElliPro. Candidate paratopes were designed through AbodyBuilder2, grafted onto a pre-existing antibody scaffold obtained from the SAbDab database, and subsequently evaluated using molecular docking in HADDOCK. Among the tested complexes, the aducanumab antibody (PDB ID: 6CO3), a clinically approved therapeutic for Alzheimer’s disease, demonstrated the most favorable docking performance against the selected TMEM106B epitope. These findings highlight TMEM106B as a promising therapeutic target, while underscoring the need for further studies on safety, efficacy, neuronal delivery strategies, and the design of multi-target antibodies capable of addressing the complexity of neurodegenerative pathologies.
إعداد: الطالب يزن شرف
إشراف: الدكتور عبد القادر عبادي