Combinatorial Expression Rules of Ion Channel Genes in Juvenile Rat (Rattus norvegicus) Neocortical Neurons

The electrical diversity of neurons arises from the expression of different combinations of ion channels. The gene expression rules governing these combinations are not known. We examined the expression of twenty-six ion channel genes in a broad range of single neocortical neuron cell types. Using expression data from a subset of twenty-six ion channel genes in ten different neocortical neuronal types, classified according to their electrophysiological properties, morphologies and anatomical positions, we first developed an incremental Support Vector Machine (iSVM) model that prioritizes the predictive value of single and combinations of genes for the rest of the expression pattern. With this approach we could predict the expression patterns for the ten neuronal types with an average 10-fold cross validation accuracy of 87% and for a further fourteen neuronal types not used in building the model, with an average accuracy of 75%. The expression of the genes for HCN4, Kv2.2, Kv3.2 and Caβ3 were found to be particularly strong predictors of ion channel gene combinations, while expression of the Kv1.4 and Kv3.3 genes has no predictive value. Using a logic gate analysis, we then extracted a spectrum of observed combinatorial gene expression rules of twenty ion channels in different neocortical neurons. We also show that when applied to a completely random and independent data, the model could not extract any rules and that it is only possible to extract them if the data has consistent expression patterns. This novel strategy can be used for predictive reverse engineering combinatorial expression rules from single-cell data and could help identify candidate transcription regulatory processes.神经元的电多样性源于离子通道不同组合的表达。目前尚不明确调控此类离子通道组合的基因表达规则。本研究针对多种不同的新皮层神经元单细胞类型，检测了26个离子通道基因的表达情况。我们基于10种经电生理特性、形态学特征及解剖位置分类的新皮层神经元类型的26个离子通道基因表达子集数据，首先构建了增量支持向量机（incremental Support Vector Machine，iSVM）模型，该模型可优先评估单个基因及基因组合对其余表达模式的预测价值。借助该方法，我们可预测这10种神经元类型的表达模式，其10折交叉验证平均准确率达87%；同时可预测另外14种未参与模型构建的神经元类型的表达模式，平均准确率为75%。研究发现，HCN4、Kv2.2、Kv3.2及Caβ3基因的表达可作为离子通道基因组合的极强预测因子，而Kv1.4与Kv3.3基因的表达则无预测价值。随后我们通过逻辑门分析，提取出不同新皮层神经元中20个离子通道的组合式基因表达规则谱。我们还证实，当将模型应用于完全随机的独立数据时，无法提取任何规则；仅当数据存在一致的表达模式时，方能提取此类规则。这一全新策略可用于从单细胞数据中预测性逆向构建组合式表达规则，有助于识别潜在的转录调控过程。