snoHUB2

How to use snoHUB2 and read its results.

Getting started

snoHUB2 predicts biological functions for human small nucleolar RNAs (snoRNAs) from their co-expression with protein-coding genes across ARCHS4 bulk RNA-seq samples. It is free to use, with no login or registration, and there are two ways to use it:

Sample output

These links open real results, so they work the same way as results for your own query.

To try sample input yourself, choose Load example set on the Set analysis page, or select an example snoRNA under the search box on the Home page.

Reading a snoRNA report

Section How to read it
Abstract A short written summary of the evidence. Treat it as a hypothesis for follow-up, not an established finding.
Overview Gene details, location, length, publication count, sequence, and predicted structure.
Expression across tissues, cell types, and cell lines Where the snoRNA has the highest average expression. Many RNA-seq methods capture few snoRNAs, so a group's average also reflects how its studies prepared their libraries.
Similar snoRNAs Nearby points on the UMAP and t-SNE plots have similar co-expression profiles. Closeness does not establish shared function.
Top correlated genes Protein-coding genes positively or negatively correlated with the snoRNA across samples. A strong correlation is not evidence of a direct interaction.
Overall highest correlated terms Gene-set terms with the highest mean Pearson correlation across all libraries. A higher value means the term's genes follow the snoRNA more closely.
Predicted functions by library The same kind of result, one chart per gene-set library. Expand a table to see which genes drive a term.
Upstream transcription factors Transcription factors whose targets overlap the snoRNA's top co-expressed genes, ranked by ChEA3 (a lower score is stronger). The ChEA-KG network, shown on the page or opened in ChEA-KG, shows how the top factors regulate each other. A factor here is a candidate regulator of the co-expression program, not a shown regulator of the snoRNA.
Drug and gene perturbations Perturbations that increase ("up") or decrease ("down") the snoRNA's top co-expressed genes. The snoRNA itself was usually not measured in these experiments. Rows supported by more libraries rank higher, and Conflicts lists libraries with evidence in the other direction.

Using set analysis

Enter 2 to 100 snoRNA symbols or Ensembl gene IDs, one per line or separated by commas. Capitalization does not matter, no other formatting or preprocessing is needed, and the page shows how many names it recognizes as you type.

  • A name that is not recognized is skipped, and the results list it.
  • A symbol shared by several genes includes all of them, so a set can be larger than the number of names you entered.
  • Use Copy link to bookmark or share results. The snoRNAs are stored in the link itself, and snoHUB2 does not store your list. Only the related genes are sent to Perturb-Seqr, ChEA3, and, when you open them, Enrichr and ChEA-KG.

Reading set analysis results

snoHUB2 first finds the protein-coding genes most related to your snoRNAs, then analyzes those genes three ways. The results follow those four steps.

The switch at the top, which stays on screen as you scroll, chooses which related genes are analyzed: genes positively correlated with your snoRNAs (the default), or genes negatively correlated with them. Next to it are links to each section. Below it, a diagram shows the steps of the analysis, from your set to its related genes to the three analyses of those genes. Each step's card lists its top results, filling in as Perturb-Seqr and ChEA3 finish, and opens that step's full results.

1. Protein-coding genes related to your snoRNAs

The 200 genes with the highest Pearson correlation averaged across all snoRNAs in your set, the method ARCHS4 and Geneshot use for gene sets. A gene must correlate with most of the set to rank high, because snoRNAs it does not correlate with pull its average down. In top 500 of shows how many of your snoRNAs have the gene among their own top 500 correlated genes. This gene set is the input to the next three steps. Genes that recur across most snoRNAs, such as olfactory receptor genes, can appear here for any set.

The heatmap shows the top 30 related genes against each of your snoRNAs. A row with the same color across the set is a gene the whole set agrees on. A column that stands out is a snoRNA that differs from the rest.

2. Enrichment analysis

The top 10 terms from each of the 5 gene-set libraries, the sets of genes that overlap the related genes most, tested as in Enrichr. Choose a library to switch between them. Open in Enrichr sends the related genes to Enrichr to test them against all of its libraries.

Result Meaning
Overlap How many of the term's genes are among the related genes.
P-value How likely an overlap this large would be by chance, from a hypergeometric test. Lower is stronger.
Adj. p The p-value adjusted for testing many terms (Benjamini-Hochberg).
Significant The term's adjusted p-value is below 0.05.
Not significant The term is among the library's strongest for this set, but its adjusted p-value is above 0.05.

3. Drug and gene perturbations (Perturb-Seqr)

Drugs and gene perturbations that increase or decrease the related genes, not the snoRNAs themselves. Rows supported by more libraries rank higher. Other direction lists libraries with a significant signature the other way, which can include the same libraries, often from different cell lines or doses. When fewer than 10 are significant, the strongest non-significant ones are added in grey and marked Not significant.

4. Upstream transcription factors (ChEA-KG)

Transcription factors that may regulate the related genes. Score is the factor's mean rank across ChEA3's libraries, so a lower score is stronger; ChEA-KG uses the same ranking. The ChEA-KG network, shown on the page or opened in ChEA-KG, shows the regulation among the top factors.

Genes that correlate with most snoRNAs, such as olfactory receptor genes, can be among the related genes of any set, so terms, perturbations, and factors built on them say little about your set in particular. Sections 3 and 4 run on outside services, so if one is unavailable, its section says so and the rest of the page still works. Very similar snoRNAs, such as copies of one gene, are an easier case in every section, because their copies are hard to tell apart in RNA-seq.

What the results can and cannot show

snoHUB2 results are predictions from co-expression and do not establish causality. They suggest hypotheses for experiments, and they do not show that a snoRNA acts on a gene or pathway. Standard RNA-seq methods capture few snoRNAs, and intronic snoRNAs can follow the expression of their host genes.

Web services

Every service returns JSON and needs no key or registration. Results are under data, and errors are under error.

Service Access URL
Every snoRNA symbol and Ensembl ID that can be queried GET https://snohub2.k8s.maayanlab.cloud/api/snornas
The full report for a snoRNA GET https://snohub2.k8s.maayanlab.cloud/api/snorna/<symbol or Ensembl ID>
Gene annotation for a snoRNA GET https://snohub2.k8s.maayanlab.cloud/api/snorna/<symbol or Ensembl ID>/annotation
Genomic location of a snoRNA GET https://snohub2.k8s.maayanlab.cloud/api/snorna/<symbol or Ensembl ID>/location
The snoRNA similarity map GET https://snohub2.k8s.maayanlab.cloud/api/similarity
Set analysis, with a JSON body such as {"snornas": ["SNORD116-1", "SNORD116-2"]}. Add "direction": "negative" for the negatively correlated genes; the default is "positive". The Perturb-Seqr and ChEA3 services take the same body. POST https://snohub2.k8s.maayanlab.cloud/api/set-analysis
Perturb-Seqr results for a set, with the same JSON body POST https://snohub2.k8s.maayanlab.cloud/api/set-analysis/perturbseqr
ChEA3 results for a set, with the same JSON body POST https://snohub2.k8s.maayanlab.cloud/api/set-analysis/chea3

Download the Python template script. It calls several of these services and parses their output, using only standard Python.

Common questions

My snoRNA is not found. snoHUB2 covers the human snoRNAs that passed its pipeline. Try the Ensembl gene ID, or see /api/snornas for every name it recognizes.

Why do some snoRNAs show only an Ensembl ID? They have no gene symbol, so snoHUB2 labels them by their Ensembl gene ID.

Which browsers work? Use a current version of Chrome, Firefox, Safari, or Edge.

Where can I read more? The Methods page describes each step, and the Downloads page has the full data tables.