Akademik Veri Yönetim Sistemi
XVIII. Tıbbi Biyoloji ve Genetik Kongresi, Ankara, Türkiye, 26 - 29 Ekim 2023, cilt.2, sa.3, ss.102, (Özet Bildiri)
Objective: FMF is complex hereditary autoinflammatory diseases which is commonly observed in Mediterranean region. However, according toThe International Study Group for Systemic Autoinflammatory Diseases (INSAID) consensus criteria half of the MEFV gene variants are uncertain. In this study, we combine multiple tools by implementing Random Forest (RF) algorithm. Therefore, we aim to assign variation of
nknown variants (VOUS) to disease-causing or benign categories.
Methods: We extracted variants of the MEFV gene from the Infevers database, and single nucleotide alterations in coding regions were included; others were excluded from the study. We then determined the optimal number of in-silico tools for our model. On the training dataset, we conducted a RF classifier on known MEFV gene variants. The prediction dataset included 168 VOUS variants.
Results: We included 266 of the 381 MEFV gene variants and four computational tools (Revel, MetaLR, SIFT, and Polyphen-2) in a study. Overall, 98 variants were classified as benign or disease-causing variants. However, the remaining 168 variants were detected as having uncertain significance. According to power analysis, our sample size was sufficient to conduct a RF classifier model. Therefore, we selected 98 known variants as a training dataset, of which 49 were evaluated as disease-causing; however, only 49 were evaluated as benign. After that, the RF algorithm was conducted, and 100% accuracy was obtained. The remaining 168 variants were predicted to be benign or disease-causing..
Conclusion: As many MEFV gene variants categorized into uncertain categories, it is essential to accurately predict MEFV gene variants. Functional investigations play a crucial role in the identification and characterization of VOUS variants. Nevertheless, it is both time-consuming and costly. Employing a RF method to train an optimized number of in silico tools could offer a viable strategy for predicting VOUS variants in routine clinical practice, hence aiding in clinical decision-making.