Department of Genetics and Biochemistry, Clemson University, Clemson, SC 29634, USA.
Center for Human Genetics, Clemson University, Greenwood, SC 29646, USA.
Genes (Basel). 2022 Aug 20;13(8):1488. doi: 10.3390/genes13081488.
In the nervous system, synapses are special and pervasive structures between axonal and dendritic terminals, which facilitate electrical and chemical communications among neurons. Extensive studies have been conducted in mice and rats to explore the RNA pool at synapses and investigate RNA transport, local protein synthesis, and synaptic plasticity. However, owing to the experimental difficulties of studying human synaptic transcriptomes, the full pool of human synaptic RNAs remains largely unclear. We developed a new machine learning method, called PredSynRNA, to predict the synaptic localization of human RNAs. Training instances of dendritically localized RNAs were compiled from previous rodent studies, overcoming the shortage of empirical instances of human synaptic RNAs. Using RNA sequence and gene expression data as features, various models with different learning algorithms were constructed and evaluated. Strikingly, the models using the developmental brain gene expression features achieved superior performance for predicting synaptically localized RNAs. We examined the relevant expression features learned by PredSynRNA and used an independent test dataset to further validate the model performance. PredSynRNA models were then applied to the prediction and prioritization of candidate RNAs localized to human synapses, providing valuable targets for experimental investigations into neuronal mechanisms and brain disorders.
在神经系统中，突触是轴突和树突末梢之间特殊而普遍的结构，促进神经元之间的电和化学通讯。已经在小鼠和大鼠中进行了广泛的研究，以探索突触处的 RNA 池，并研究 RNA 转运、局部蛋白质合成和突触可塑性。然而，由于研究人类突触转录组的实验困难，人类突触 RNA 的全部池仍然很大程度上不清楚。我们开发了一种新的机器学习方法，称为 PredSynRNA，用于预测人类 RNA 的突触定位。从以前的啮齿动物研究中编译了树突状定位 RNA 的训练实例，克服了人类突触 RNA 经验实例的不足。使用 RNA 序列和基因表达数据作为特征，构建并评估了具有不同学习算法的各种模型。令人惊讶的是，使用发育中大脑基因表达特征的模型在预测突触定位 RNA 方面表现出优异的性能。我们检查了 PredSynRNA 学习的相关表达特征，并使用独立的测试数据集进一步验证了模型性能。然后将 PredSynRNA 模型应用于预测和优先考虑定位于人类突触的候选 RNA，为神经元机制和大脑疾病的实验研究提供了有价值的目标。
