Fig. 6. Lamian analysis of 184 tuberculosis (TB) samples identifies differential pseudotemporal genes and cell density between male and female while controlling for batch effects.
a UMAP51 showing cells color-coded by pseudotime constructed using NAM PC2. b Study design showing sample batches. c Cell density distribution along pseudotime for each sample (curve) (TCD test p < 2.22 × 10−308, XCD test p = 8.9 × 10−6, both are one-sided). d Heatmaps showing the expression patterns of sex-associated XDE genes (rows) in cells (columns) ordered by pseudotime. White bars are used to separate original expression, model-fitted expression, group trend difference (male minus female), and group mean shift. e Samples are randomly partitioned into two datasets. For each method, the proportion of top N XDE genes that overlap between the two datasets (y-axis) is shown as a function of N (x-axis). f Average overlap in (e) (i.e. area under the curve divided by the x-axis length). g Overlap between XDE genes reported by different methods and sex chromosome genes, along with p-values in permutation test (one-sided, 104 permutations). Methods without outputs within 1 week and 400GB are not shown in (e–g). h Proportion of XDE genes that overlap between two partitions of data, with versus without batch correction. The TB dataset was split into two: one contains one batch, and the other one contains all remaining batches (only use the batches with ≥2 samples in both sexes). For the one with multiple batches, Lamian is run with and without batch correction. The data split was repeated 12 times. Each boxplot shows the distribution (center: median; bounds of box: 1st and 3rd quartiles; bounds of whiskers: data points within 1.5 IQR from the box; minima; maxima) of overlap in the n = 12 splits. Source data are provided as a Source Data file.