Single-gene query can be a useful way to gauge at gene function, or to understand context through viewing its expression across different tissues, cell lines, in diseased and normal conditions.
When a single-gene query is entered, SEEK by default searches through the collection of 13 or so multi-tissue datasets commonly used in the community. Then by clicking on a row (or a gene) in the expression heatmap, users can see in which conditions is the gene up-regulated.
Other than the 13 multiple-tissue community standard datasets, user can also refine to tissue-specific datasets through the Refine Search function once on the search result page. Note that adding another gene to the query will no longer use the same 13 datasets for search, as it will enter the multi-gene query search mode.
Multi-gene query
When a multi-gene query is entered, SEEK by default will search through the entire collection of ~5200 expression datasets, using the cross-validated, hubbiness corrected algorithm to weight and rank datasets. In this mode, users can subsequently perform Dataset Enrichment (to seek for enriched tissues among up-weighted datasets), Gene Enrichment, and so forth.
For gene function inference, should I search in the whole compendium or in specific tissue-related datasets?
From our experience, whole compendium search gives the most accurate prediction of gene function (sometimes even if the query is single-gene). This draws more statistical power from thousands of datasets in the whole compendium, compared to the limited number of multi-tissue datasets. It borrows coexpression information across tissues, diseases, all datasets to make functional prediction.
If the goal is to reveal biology explaining nuances within a single tissue (such as disease vs control, treatment vs control, or disease subtypes), it would make sense to limit to tissue-specifc datasets.
When to use single-gene and multi-gene?
If the user isn't sure where the gene is expressed, use the single-gene query search mode to activate searching in multi-tissue datasets. If the user wants to prioritize datasets or wishes to assign function to a gene-set, multi-gene query is recommended.
How to improve the result of a single-gene query?
Since a single-gene query may not provide enough context for SEEK to weight the datasets, one way to resolve this issue is by either expanding the query or by limiting the search scope to datasets within a domain (via the Refine Search function). To expand the query, we recommend using functional network such as STRING, IMP, etc. Users can then type the query along with its closest neighbors as a multi-gene query.
