There is currently no discussion in the tutorial about how to select thresholds that determine topic-specific cells. Internally, the Riesenfeld Group tends to use the KL divergence of the top genes in a topic in cells in the topic vs. cells not in the topic at candidate thresholds to estimate a range of reasonable thresholds. This is detailed in topic_threshold_heuristic_plot, but the current version on GitHub has a bug that results in plotting the KL divergence only based on one random topic-specific gene. This needs to be fixed and discussed in the tutorial.
There is currently no discussion in the tutorial about how to select thresholds that determine topic-specific cells. Internally, the Riesenfeld Group tends to use the KL divergence of the top genes in a topic in cells in the topic vs. cells not in the topic at candidate thresholds to estimate a range of reasonable thresholds. This is detailed in
topic_threshold_heuristic_plot
, but the current version on GitHub has a bug that results in plotting the KL divergence only based on one random topic-specific gene. This needs to be fixed and discussed in the tutorial.