Graduate School of Pharmaceutical Sciences, Ritsumeikan University, Kusatsu 525-8577, Shiga, Japan.
Department of Pharmacy, College of Pharmaceutical Sciences, Ritsumeikan University, Kusatsu 525-8577, Shiga, Japan.
Int J Mol Sci. 2022 Oct 14;23(20):12281. doi: 10.3390/ijms232012281.
Cyclin-dependent kinase-like 5 (CDKL5) is a serine/threonine protein kinase whose pathological mutations cause CDKL5 deficiency disorder. Most missense mutations are concentrated in the catalytic domain. Therefore, anticipating whether mutations in this region affect CDKL5 function is informative for clinical diagnosis. This study comprehensively predicted the pathogenicity of all 5700 missense substitutions in the catalytic domain of CDKL5 using in silico analysis and evaluating their accuracy. Each missense substitution was evaluated as "pathogenic" or "benign". In silico tools PolyPhen-2 HumDiv mode/HumVar mode, PROVEAN, and SIFT were selected individually or in combination with one another to determine their performance using 36 previously reported mutations as a reference. Substitutions predicted as pathogenic were over 88.0% accurate using each of the three tools. The best performance score (accuracy, 97.2%; sensitivity, 100%; specificity, 66.7%; and Matthew's correlation coefficient (MCC), 0.804) was achieved by combining PolyPhen-2 HumDiv, PolyPhen-2 HumVar, and PROVEAN. This provided comprehensive information that could accurately predict the pathogenicity of the disease, which might be used as an aid for clinical diagnosis.
细胞周期蛋白依赖性激酶样 5（CDKL5）是一种丝氨酸/苏氨酸蛋白激酶，其病理性突变导致 CDKL5 缺乏症。大多数错义突变集中在催化结构域。因此，预测该区域的突变是否影响 CDKL5 的功能对于临床诊断具有重要意义。本研究使用计算机分析全面预测了 CDKL5 催化结构域中所有 5700 个错义取代的致病性，并评估了它们的准确性。每个错义取代都被评估为“致病性”或“良性”。分别选择 PolyPhen-2 HumDiv 模式/HumVar 模式、PROVEAN 和 SIFT 等计算工具，或组合使用这些工具，根据 36 个先前报道的突变作为参考，评估它们的性能。使用三种工具中的每一种，预测为致病性的取代都有 88.0%以上的准确率。组合使用 PolyPhen-2 HumDiv、PolyPhen-2 HumVar 和 PROVEAN 可获得最佳性能评分（准确率为 97.2%，灵敏度为 100%，特异性为 66.7%，马修相关系数（MCC）为 0.804）。这提供了全面的信息，可以准确预测疾病的致病性，可作为临床诊断的辅助手段。
