ScEpiSearch first solves the problem of having common reference repository to compare single cell epigenome profiles. It also provides statistical framework to compare single cell open-chromatin profiles, which is not a trivial problem given the sparsity of the data-sets. A cell has both active and poised regulatory elements. ScEpiSearch tries to provide both views to users. By providing matched single-cell-expression profiles, it helps in detection of active components of the query cells. It also tries to provide notions about poised elements of query cells by comparing it with other published single cell open chromatin profile. Hence it not only helps in identification of cell type but cell state also. Thus if the dynamics of the top matching cell from reference dataset is well studied then one can also predict the behavior of query cell. With current version it can also help in classification of cells based on projection on other cells. With further expansion in reference datasets it has potential to provide information about true heterogeneity and rarity of cell states in the query datasets.
In current version of ScEpiSearch user can submit single cell open chromatin profiles such as scATAC-seq or scDNAse-seq. Since there are very few genome wide profiles for single cell histone modification and single cell DNA methylation, they are not yet part of ScEpiSearch.
You can not submit raw fastq or bam files directly to ScEpiSearch. You should first get tag-count on a peak-list and assimilate data from cells in 2-dimensional matrix where row represent peak location and every column correspond to one cell. For this purpose you can make a union of peak from multiple cells and use other tools such as HT-seq or bedtools to estimate tag-counts on the peaks.
Due to computational expenses, Cross-batch and Cross-species embedding is available in stand-alone version. One can download standalone tool from homepage. Standalone tool is available on request.
Whenever a query is submitted, our webserver creates a customURL where the results would be displayed. User can save this URL and access the results in future also.
It is advisable to download all the results immediately, however user can access their results even after few days of query. However there is no guarantee that their results can be accessed after 2 months of query.
When single cell open chromatin profile is provided to ScEpiSearch, it tries to highlight cell type specific sites (mostly enhancers) for each cell. The list of potential enhancer provides probabilistic landscape of cell type specific gene expression. ScEpiSearch finds and report cells with matching specificity of gene expression of top active genes using it's single cell expression repository. For finding matching single cell epigenome profile, ScEpiSearch uses gene enrichment scores for large number of genes, hence it may also match poised states of cells. Hence match to expression profile provide view of active components in a cell and epigenome match results provides panorama of both poised and active regulatory components.
P-values for the match represents the probability that the result is by chance. We calculate it so that for every reference cell we take care of bias and drop-out rate using null model.
Rank based adjusted P-values for the match represents the rank of a reference cell for a query cell used to calculate new rank based P-value.
The word cloud gives a complete overview of all matching phenotypes occuring in both epigenome and expression matches for all queries. The world cloud is clickable and pop up appears of enriched genes.
The frequency plot shows frequency of genes in all queries combined.
The width of peaks can be according to your choice, however avoid using too narrow like 200bp and too wide peaks like 20kb.
It may be possible that server is overloaded. You can wait for 12 hours and you could get email about completion of result, if you have provided or you could check it yourself using the result URL. If you are not able to get results, then email the developers with the query details and result URL.
There could be multiple reason for bad results
- User has provided data from different genome version
- Low sequencing depth
- Low signal to noise ratio in open-chromatin profiles
- Non zero peaks in open chromatin profiles is less than 10000.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License