Fig. 5

Clustering primary human myoblasts based on transcript-compatibility counts. a The transcript-compatibility counts matrix for 271 primary human myoblasts from [12] is visualized using a diffusion map. Three clusters obtained using affinity propagation are shown along with the distribution of these cells across the four cell-collection timepoints (0, 24, 48, and 72 hours). b The diffusion map obtained using transcript compatibility counts is relabeled using the cells reported by [12]. Clusters 1,2,3 generated by the transcript compatibility based method map to proliferating cells, differentiating myoblasts, and interstitial cells, respectively. According to Trapnell et al.’s labels, the transcript compatibility based method seems to have severely misclassified cell T48_CT_G10 (SRR1033183) as a differentiating myoblast. c Comparing the expressions of 12 differentiating genes in T48_CT_G10 with those of the average proliferating cell and the average differentiating myoblast, 8 out of the 12 genes show expressions similar to what one would expect from a differentiating myoblast. MYOG seems to show an FPKM of 14, which while more than the mean expression of proliferating cells (around 0.28) is much less than the mean expression of differentiating myoblasts (around 61.33). We note that this cell has the highest expression of MYOG among all cells labeled by Trapnell et al. as proliferating cell (and the second highest cell has expression around 5.4). However there are 88 differentiating myoblasts with MYOG expression less than 15 FPKM. Hence it is reasonable to think that this MYOG expression is more typical of differentiating myoblasts than proliferating cells. Only genes CDK1 and CCMB2 show expressions close to what one would expect from a proliferating cell. Even though CDK1 is a highly specific marker for proliferating cells, the above gene profile indicates that classifying cell T48_CT_G10 as a differentiating myoblast seems reasonable