Concerning the slow load of genes in Data Exploration
Context in this post by Vicky. tl;dr
Issues:
When we started doing downsampling in the worker for marker heatmap, we effectively stopped preloading marker genes for the gene list. That means that when you click them a new work request needs to go out and it's slow (5-8s).The goal of this ticket is to pre-load some of those genes to to make it more responsive. There are decisions to be made in terms of performance. The amount of marker genes grows with the dataset size (we compute 5 marker genes per sample) while the number of genes per page (of the gene list) does not. Ideally, we would like to preload all the marker genes. However, we don't want to do that if it means that big datasets become too slow.
Goal
Add a mechanism in the UI to preload X genes which should be a single work request with the list of all genes to be preloaded.
Initially try to preload all marker genes.
Explore whether loading all marker genes is slow for very big experiments (large number of sample, not cells)
If the performance is ok (make sure to check with Vicky) finish. Otherwise,
Explore other heuristics like
Setting a max amount of genes to be loaded
Preloading only the first (or more) page of the gene list (for datasets which become slow)
Context
Concerning the slow load of genes in Data Exploration Context in this post by Vicky. tl;dr
Issues:
When we started doing downsampling in the worker for marker heatmap, we effectively stopped preloading marker genes for the gene list. That means that when you click them a new work request needs to go out and it's slow (5-8s).The goal of this ticket is to pre-load some of those genes to to make it more responsive. There are decisions to be made in terms of performance. The amount of marker genes grows with the dataset size (we compute 5 marker genes per sample) while the number of genes per page (of the gene list) does not. Ideally, we would like to preload all the marker genes. However, we don't want to do that if it means that big datasets become too slow.
Goal