Isewon Itunuoluwa, Binaansim Stephen, Adegoke Faith, Emmanuel Jerry, Oyelade Jelili
Department of Computer and Information Sciences, Covenant University, Ota, Ogun State, Nigeria.
Covenant Bioinformatics Research (CUBRe), Covenant University, Ota, Nigeria.
PLoS One. 2024 Dec 23;19(12):e0315530. doi: 10.1371/journal.pone.0315530. eCollection 2024.
Essential genes are those whose presence is vital for a cell's survival and growth. Detecting these genes in disease-causing organisms is critical for various biological studies, including understanding microbe metabolism, engineering genetically modified microorganisms, and identifying targets for treatment. When essential genes are expressed, they give rise to essential proteins. Identifying these genes, especially in complex organisms like Plasmodium falciparum, which causes malaria, is challenging due to the cost and time associated with experimental methods. Thus, computational approaches have emerged. Early research in this area prioritised the study of less intricate organisms, inadvertently neglecting the complexities of metabolite transport in metabolic networks. To overcome this, a Network-based Machine Learning framework was proposed. It assessed various network properties in Plasmodium falciparum, using a Genome-Scale Metabolic Model (iAM_Pf480) from the BiGG database and essentiality data from the Ogee database. The proposed approach substantially improved gene essentiality predictions as it considered the weighted and directed nature of metabolic networks and utilised network-based features, achieving a high accuracy rate of 0.85 and an AuROC of 0.7. Furthermore, this study enhanced the understanding of metabolic networks and their role in determining gene essentiality in Plasmodium falciparum. Notably, our model identified 9 genes previously considered non-essential in the Ogee database but now predicted to be essential, with some of them potentially serving as drug targets for malaria treatment, thereby opening exciting research avenues.
必需基因是指那些其存在对细胞的存活和生长至关重要的基因。在致病生物体中检测这些基因对于各种生物学研究至关重要，包括了解微生物代谢、工程改造转基因微生物以及确定治疗靶点。当必需基因表达时，它们会产生必需蛋白质。识别这些基因，尤其是在像导致疟疾的恶性疟原虫这样的复杂生物体中，由于与实验方法相关的成本和时间，具有挑战性。因此，计算方法应运而生。该领域的早期研究优先关注较简单生物体的研究，无意中忽略了代谢网络中代谢物运输的复杂性。为了克服这一点，提出了一种基于网络的机器学习框架。它使用来自BiGG数据库的基因组规模代谢模型（iAM_Pf480）和来自Ogee数据库的必需性数据，评估了恶性疟原虫的各种网络特性。所提出的方法显著改进了基因必需性预测，因为它考虑了代谢网络的加权和有向性质，并利用了基于网络的特征，实现了0.85的高精度率和0.7的曲线下面积（AuROC）。此外，这项研究增进了对代谢网络及其在确定恶性疟原虫基因必需性方面作用的理解。值得注意的是，我们的模型识别出9个在Ogee数据库中先前被认为是非必需的基因，但现在预测是必需的，其中一些可能作为疟疾治疗的药物靶点，从而开辟了令人兴奋的研究途径。
