BICCN Mouse MOp meta-markers

We extracted replicable markers for neuron cell types from a compendium of 7 scRNAseq datasets generated by the BICCN in the mouse primary motor cortex (https://doi.org/10.1101/2020.02.29.970558). The markers were extracted using the MetaMarkers package (https://github.com/gillislab/MetaMarkers) using default parameters and keeping the top 1000 markers for each cell type.<br>We present markers for cell types at each level of the hierarchy defined by the BICCN. - biccn_class_markers.csv: highest level with only 3 cell types (excitatory neurons, inhibitory neurons, non-neurons). - biccn_subclass_markers.csv: intermediate level containing 13 cell types (e.g. PV+ interneurons, L6b excitatory neurons). - biccn_cluster_markers.csv: high-resolution level containing 86 cell types (e.g. Chandelier cells).<br>The number of informative markers varies by cell type. To find the best number of markers, we looked for optimal annotation performance for each cell type. The results are summarized in "optimal_number_markers.csv". For each cell type, performance is reported in column "f1" (0.75 indicates good performance, 1 is perfect performance). To read this table, pick a cell type of interest, find optimal performance, then the range of genes that lead to optimal performance. For example, Chandelier cells (Pvalb Vipr2_2) are perfectly characterized by 50 to 200 markers (F1&gt;0.99).

我们从小鼠初级运动皮层中由脑计划细胞普查网络（Brain Initiative Cell Census Network, BICCN）生成的7个单细胞RNA测序（single-cell RNA sequencing, scRNAseq）数据集汇编中，提取了神经元细胞类型的可重复性分子标记基因（相关数据集链接：https://doi.org/10.1101/2020.02.29.970558）。本次标记基因的提取采用MetaMarkers工具包（https://github.com/gillislab/MetaMarkers），使用默认参数，并保留每种细胞类型的前1000个标记基因。
我们提供了BICCN所定义的层级分类体系中各层级细胞类型的标记基因集合：
- "biccn_class_markers.csv"：对应最高层级，仅包含3种细胞类型（兴奋性神经元、抑制性神经元、非神经元细胞）。
- "biccn_subclass_markers.csv"：对应中间层级，包含13种细胞类型（例如小清蛋白阳性中间神经元（PV+）、6b层兴奋性神经元（L6b））。
- "biccn_cluster_markers.csv"：对应高分辨率层级，包含86种细胞类型（例如钳状细胞（Chandelier cells））。
不同细胞类型的有效标记基因数量存在差异。为确定最优的标记基因数量，我们针对每种细胞类型优化了其注释性能，相关结果汇总于"optimal_number_markers.csv"文件中。每种细胞类型的性能以"f1"列展示（0.75代表性能良好，1代表完美性能）。若要读取该表格，可先选定目标细胞类型，找到其最优性能对应的基因区间，即可获知实现最优性能所需的标记基因范围。例如，钳状细胞（Pvalb Vipr2_2）仅需50至200个标记基因即可实现完美表征（F1值>0.99）。