RNA-Seq analysis with AskOmics Interactive Tool

Overview
Creative Commons License: CC-BY Questions:
  • How to integrate RNA-Seq results with other datasets using AskOmics?

  • How to query these datasets to answer a biological question?

  • How to exploit a distant SPARQL endpoint to retrieve information?

Objectives:
  • Launch an AskOmics Interactive Tool

  • Integrate RNA-Seq and reference datasets into AskOmics

  • Create a complex query to get over-expressed genes, their location on the reference genome, and check if they are included in a known QTL

  • Complete the query to get the human homologs and their location using neXtProt database

Requirements:
Time estimation: 2 hours
Supporting Materials:
Published: Jul 9, 2020
Last modification: Nov 3, 2023
License: Tutorial Content is licensed under Creative Commons Attribution 4.0 International License. The GTN Framework is licensed under MIT
purl PURL: https://gxy.io/GTN:T00297
rating Rating: 5.0 (0 recent ratings, 1 all time)
version Revision: 17

AskOmics is a web application for data integration and querying using the Semantic Web technologies. It helps users to convert multiple data sources (CSV/TSV files, GFF and BED annotation) into “RDF triples” and store them in a specific kind of database: an “RDF triplestore”. Under this form, data can then be queried using a specific language: “SPARQL”. AskOmics hides the complexity of these technologies and allows to perform complex queries using a user-friendly interface.

AskOmics comes in useful for cross-referencing results datasets with various reference data. For example, in RNA-Seq studies, we often need to filter the results on the fold change and the p-value, to get the most significant differentially expressed genes. If you are studying a particular phenotype and already know the position of some QTL associated to this phenotype, you would then want to find the positions of the differentially expressed genes and determine which gene is located within one of those QTL. Finally, you would want to know if these genes have human homologs, and use the neXtProt database to get the location of the proteins coded by the homologs. The whole process involves several tools to parse and manipulate the different data format, and to map datasets on each other. AskOmics offer a solution to 1) automatically convert the multiple formats to RDF, 2) use a user-friendly interface to perform complex SPARQL queries on the RDF datasets to find the genes you are interested in, and 3) connect external SPARQL databases and link external data with your own.

In this tutorial, we will use results from a differential expression analysis. This file is provided for you below. You could also generate the file yourself, by following the RNA-Seq counts to gene tutorial. The file used here was generated from limma-voom but you could use a file from any RNA-seq differential expression tool, such as edgeR or DESeq2, as long as it contains the required columns (see below).

The differentially expressed results will be linked to the official mouse genome annotation, in general feature format (GFF). The file provided is a subset of the mouse annotation (GRCm38.p6) obtained from Ensembl.

We will use a file containing quantitative trait loci (QTL) information, to find if our differentially expressed genes are located inside a known QTL. This file is a subset of a query performed on Mouse Genome Informatics.

A file containing all homologies between mouse and human will be used to get the human homolog genes. This file is provided by MGI.

In the differentially expressed file, and the homologs file, gene are described by a symbol (e.g. Pwgrq10). However, in the annotation file and neXtProt database, gene are represented by Ensembl id (e.g. ENSMUSG00000025969). To link the 2 datasets, we will need a file to map the gene symbol with Ensembl id. This file provided for you was previously generated with an AskOmics query on the mouse annotation file and the homolog file.

To link the human gene with neXtProt database, we will use the RDF abstraction of neXtProt. This file was obtained using the Abstractor tool.

During the integration step, AskOmics builds an RDF description of the data: the abstraction. This abstraction is used to explore the data and build the query. AskOmics can also integrate abstraction of distant endpoint. Abstraction are obtained using abstractor, a python package to generate RDF abstractions from distant endpoints. The query builder interface is used to create a path through the abstraction of each ressources. The path is converted to a SPARQL query that is sent to the multiple SPARQL endpoint.

Agenda

In this tutorial, we will cover:

  1. Preparing the inputs
    1. Import data
  2. Upload inputs into AskOmics
    1. Launch AskOmics Interactive Tool
  3. Integrate input files into AskOmics
    1. Integrate GFF files
    2. Integration of tabular (TSV) files
    3. Integration of RDF files
    4. Monitor integration
  4. Query
    1. Query builder overview
    2. Results management
  5. Conclusion

Preparing the inputs

We will use four files for this analysis:

  • Differentially expressed results file (genes in rows, and 4 required columns: identifier (ENTREZID), gene symbol (SYMBOL), log fold change (logFC) and adjusted P values (adj.P.Val))
  • Reference genome annotation file (in GFF format)
  • QTL file (QTL in rows, with 5 required columns: identifier, chromosome, start, end and name)
  • Homolog file (TSV of 13 columns including homolog id, organism name and gene symbol)
  • Correspondence file between gene symbol and Ensembl id (TSV of 3 columns: symbol, the corresponding Ensembl id (mouse and human)
  • neXtProt abstraction (RDF data description of neXtProt database in turtle format)

Import data

Hands-on: Data upload
  1. Create a new history for this RNA-seq exercise e.g. RNA-seq AskOmics

    Click the new-history icon at the top of the history panel:

    UI for creating new history

    1. Click on galaxy-pencil (Edit) next to the history name (which by default is “Unnamed history”)
    2. Type the new name
    3. Click on Save

    If you do not have the galaxy-pencil (Edit) next to the history name:

    1. Click on Unnamed history (or the current name of the history) (Click to rename history) at the top of your history panel
    2. Type the new name
    3. Press Enter

  2. Import the files.

    To import the files, there are two options:

    • Option 1: From a shared data library if available (ask your instructor)
    • Option 2: From Zenodo
    • Copy the link location
    • Click galaxy-upload Upload Data at the top of the tool panel

    • Select galaxy-wf-edit Paste/Fetch Data
    • Paste the link(s) into the text field

    • Press Start

    • Close the window

    As an alternative to uploading the data from a URL or your computer, the files may also have been made available from a shared data library:

    1. Go into Shared data (top panel) then Data libraries
    2. Navigate to the correct folder as indicated by your instructor.
      • On most Galaxies tutorial data will be provided in a folder named GTN - Material –> Topic Name -> Tutorial Name.
    3. Select the desired files
    4. Click on Add to History galaxy-dropdown near the top and select as Datasets from the dropdown menu
    5. In the pop-up window, choose

      • “Select history”: the history you want to import the data to (or create a new one)
    6. Click on Import

    • You can paste the links below into the Paste/Fetch box:

      https://zenodo.org/record/2529117/files/limma-voom_luminalpregnant-luminallactate
      https://zenodo.org/record/3950862/files/Mus_musculus.GRCm38.98.subset.gff3
      https://zenodo.org/record/3950862/files/Symbol.tsv
      https://zenodo.org/record/3950862/files/MGIBatchReport_Qtl_Subset.txt
      https://zenodo.org/record/3950862/files/HOM_MouseHumanSequence.rpt
      https://zenodo.org/record/3950862/files/nextprot_abstraction.ttl
      
  3. Rename the files using the galaxy-pencil (pencil) icon.
    • limma-voom_luminalpregnant-luminallactate to DE results
    • Mus_musculus.GRCm38.98.subset.gff3 to Mus musculus annotation
    • Symbol.tsv to Gene Symbols
    • MGIBatchReport_Qtl_Subset.txt to QTL
    • HOM_MouseHumanSequence.rpt to Homolog groups
    • nextprot_asbtraction.ttl to neXtProt abstraction
  4. Check every datatype.
    • DE results: tabular
    • Mus musculus annotation: gff
    • Gene Symbol: tabular
    • QTL: tabular
    • Homolog groups: tabular
    • neXtprot abstraction: ttl

If the datatypes are wrong, please change it.

  • Click on the galaxy-pencil pencil icon for the dataset to edit its attributes
  • In the central panel, click galaxy-chart-select-data Datatypes tab on the top
  • In the galaxy-chart-select-data Assign Datatype, select your desired datatype from “New type” dropdown
    • Tip: you can start typing the datatype into the field to filter the dropdown menu
  • Click the Save button

Click on the galaxy-eye (eye) icon and take a look at the uploaded files.

Two step are necessary to get our data converted into RDF triples. The first step is to upload the Galaxy datasets into the AskOmics server. The second step is to integrate the uploaded data into the RDF triplestore.

Upload inputs into AskOmics

We will first launch an AskOmics interactive tool, and upload the data into it.

Launch AskOmics Interactive Tool

Hands-on: Launch AskOmics IT
  1. AskOmics a visual SPARQL query builder tool to launch the Interactive Tool
    • param-file “Datasets to load into AskOmics”: DE results, Mus musculus annotation, Gene Symbols, QTL, Homolog groups and neXtProt abstraction

Wait a few seconds (or minutes if computing resources are busy) for AskOmics to be ready to use. A view link should appear in the confirmation box just after clicking on the Execute button.

AskOmics is an Interactive tool. It means that when you launch it, it will stay in running state (yellow background) in your History. As long as it stays in this running state, you can access it by looking in the “User” > “Active Interactive Tools” menu (click on its name to view it). When you no longer need it, you can stop it by deleting it from your history, or using the “Stop” button in the “User” > “Active Interactive Tools” page.

Keep in mind that as long as this tool runs, it uses computing resources, so don’t forget to stop it when you no longer have use for it.

Once the AskOmics Interactive Tool is ready, you should see a start page looking like this:

home page. Open image in new tab

Figure 1: AskOmics home page

You can see that there is no data available yet. It’s because data needs to be integrated: it is the next step.

Integrate input files into AskOmics

AskOmics conversion into RDF is called integration.

On the Files page (link at the top of the page), you will see the files you uploaded from Galaxy. We will now integrate all these files.

Hands-on: Integrate data
  1. Got to the Files page
  2. Select all the input files
  3. Click on the Integrate button
files page. Open image in new tab

Figure 2: AskOmics files page

You will land on the Integrate page that shows a preview of the data present in each selected file, depending of its data type.

Integrate GFF files

The GFF preview shows the entities that the file contains. We need to select the entities we want to be integrated.

The Mus musculus annotation file we’re using contains gene and mRNA entities, and we will need both in the rest of the tutorial.

Hands-on: Integrate `Mus musculus annotation`
  1. Search for Mus musculus annotation (preview)
  2. Select gene and mRNA
  3. Click on the Integrate (private dataset) button
mus musculus annotation preview. Open image in new tab

Figure 3: Mus musculus annotation preview

Integration of tabular (TSV) files

The TSV preview shows an HTML table representing the first lines of the TSV file. During integration, AskOmics will convert the file using the header.

The first column of a TSV file will be the entity name. Other columns of the file will be attributes of the entity. Labels of the entity and attributes will be set by the header. Each label can be edited by clicking on it.

Entity and attributes can have special types. The types are defined with the select box below the header. An entity can be a start entity or an entity. A start entity means that the entity may be used to start a query on the AskOmics homepage.

Attributes can take the following types:

  • Numeric: if all the values of the column are numeric
  • Text: if all the values are strings
  • Category: if there is a limited number of repeated values (e.g. ‘green’, ‘yellow’ and ‘red’, each one found in multiple lines)

If the entity describes a locatable element on a genome:

  • Reference: chromosome
  • Strand: strand
  • Start: start position
  • End: end position

A column can also represent a relation between the entity to another. In this case, the header have to be named relationName@TargetedEntity and the type Directed or Symetric relation. A Directed relation is a relation from this entity to the targeted one (e.g. A is B’s father, but B is not A’s father). A Symetric relation is a relation that works in both directions (e.g. A loves B, and B loves A).

Hands-on: Integrate `DE results`
  1. Search for DE results (preview)
  2. Edit attribute names and types:
    • change ENTREZ ID to Differential Expression and set type to start entity
    • change SYMBOL to linkedTo@Gene Symbol and set type to Symetric relation
    • change GENENAME to name and set type to text
    • Keep the other column names and set their types to numeric
  3. Integrate (private dataset)
De results preview. Open image in new tab

Figure 4: DE results preview
Hands-on: Integrate `Gene symbols`
  1. Search for Gene symbols (preview)
  2. Edit attribute names and types:
    • change symbol to Gene Symbol and set type to entity
    • Set to mouse genes@gene type to Symetric relation
  3. Click on the Integrate (private dataset) button
Gene symbols preview. Open image in new tab

Figure 5: Symbol to Ensembl preview
Hands-on: Integrate `QTL`
  1. Search for QTL (preview)
  2. Edit attribute names and types:
    • change Input to QTL and set type to start entity
    • set Chr type to Reference
    • set Start type to Start
    • set End type to End
  3. Click on the Integrate (private dataset) button
QTL preview. Open image in new tab

Figure 6: QTL preview
Hands-on: Integrate `Homolog groups`
  1. Search for Homolog groups (preview)
  2. Edit attribute names and types:
    • change HomoloGene ID to Homolog Group and set type to start entity
    • set Common Organism Name type to category
    • change Symbol to linkedTo@Gene Symbol and set type to Directed relation
    • Keep the other column names and set their types to text
  3. Click on the Integrate (private dataset) button
Homolog groups preview. Open image in new tab

Figure 7: Homolog groups preview

Integration of RDF files

The last dataset we want to integrate is the neXtProt abstraction. This file contains some RDF data that instructs AskOmics how to communicate with a remote RDF database containing neXtProt data.

Hands-on: Integrate `neXtProt abstraction`
  1. Search for neXtProt abstraction (preview)
  2. Check that Distant endpoint is set to https://sparql.nextprot.org/sparql in advanced options
  3. Click on the Integrate (private dataset) button
neXtProt abstraction preview. Open image in new tab

Figure 8: neXtProt abstraction preview

Monitor integration

Integration can take some time depending on the file size. The Datasets page shows the progress.

Hands-on: track integration progress
  1. Go to the Dataset page
  2. Wait for all datasets to be in success state
dataset. Open image in new tab

Figure 9: Datasets table

Query

Once all the data of interest is integrated (converted to RDF), its time to query them. Querying RDF data is done by using the SPARQL language. Fortunately, AskOmics provides a user-friendly interface to build SPARQL queries without having to learn the SPARQL language.

Query builder overview

Simple query

The first step to build a query is to choose a start point for the query.

ask. Open image in new tab

Figure 10: ask page
Hands-on: Start a query
  1. Go to the Ask! page
  2. Select the Differential Expression entity
  3. Start!

Once the start entity is chosen, the query builder is displayed.

The query builder is composed of a graph. Nodes represents entities and links represents relations between entities. The selected entity is surrounded by a red circle. Links and other entities are dotted and lighter because there are not instantiated.

query builder. Open image in new tab

Figure 11: Query builder, Differential Expression is the selected entity, Gene Symbol is a suggested entity

On the right, attributes of the selected entity are displayed as attribute boxes. Each box has an eye icon: an opened eye means the attribute will be displayed on the results.

Hands-on: Ask for all Differential Expression and display some attributes
  1. Display logFC and adj.P.val by clicking on the eye icon
  2. Run & preview
preview results. Open image in new tab

Figure 12: Results preview

Run & preview launch the query with a limit of 30 rows returned. We use this button to get an idea of the results returned.

Filter on attributes

Next query will search for all over-expressed genes. Genes are considered over-expressed if the log fold change is > 2. We are only interested by significant results (Adj P value ≤ 0.05)

Back to the query builder,

Hands-on: Filter attributes to get significant over-expressed genes
  1. Filter logFC with > 2
  2. Filter adj.P.val with ≤ 0.05
  3. Run & preview
preview filtered results. Open image in new tab

Figure 13: Filtered results preview

The preview shows only significantly over-expressed genes.

Filter on relations

Now that we have found our genes of interest, we will link these genes to the reference genome to get information about their location.

To constraint on relation, we have to click on suggested nodes, linked to our entity of interest.

  1. First, hide Label, logFC and adj.P.val of Differential Expression using the eye icon
  2. Instantiate Gene Symbol by clicking on the suggested node, and hide his Label using the eye icon
  3. Instantiate gene by clicking on the gene node
  4. Run & preview
gene query. Open image in new tab

Figure 14: Gene query

Results now show the Ensembl id of our over-expressed genes. We have now access to all the information about the gene entity contained in the GFF file. For example, we can filter on chromosome and display chromosome and strand to get information about gene location.

Hands-on: Filter `gene`
  1. Show reference and strand using the eye icon
  2. Filter reference by selecting X chromosome
  3. Filter strand by selecting + strand
  4. Run & preview
gene filter. Open image in new tab

Figure 15: Gene filtering

Query on the position of elements on the genome.

AskOmics is able to perform special queries between entities that are locatable. These queries are:

  • Entities overlapping another one
  • Entities included in another entity

The FALDO ontology describes sequence feature positions and regions. AskOmics uses FALDO ontology to represent entity positions. GFF are using FALDO, as well as TSV entities with chromosome, strand, start and end.

On the query builder interface, locatable entities are represented with a green circle and relations based on location are represented as green arrow.

Hands-on: Filter `gene`
  1. First, remove the reference filter (unselect X using ctrl+click)
  2. Remove the strand filter (unselect + using ctrl+click)
  3. Hide reference strand using the eye
  4. Instantiate QTL
  5. Click on the link between gene and QTL to edit the relation
  6. Check that the relation is gene included in QTL on the same reference with strict ticked
  7. Run & preview
FALDO query. Open image in new tab

Figure 16: FALDO query

To go further, we can filter on QTL to refine the results.

Hands-on: Filter `gene`
  1. Go back to the QTL node
  2. Show the Name attribute using the eye icon
  3. Filter the name with a regexp with growth
  4. Run & preview

From now, our query is “All Genes that are over-expressed (logFC > 2 and FDR ≤ 0.05) and located on a QTL that is related to growth”. We can save this results with the Run & save button.

Hands-on: Save a result
  1. Run & save

Use neXtProt distant data to refine results

neXtProt is a comprehensive human-centric discovery platform, offering its users a seamless integration of and navigation through protein-related data. It offer a SPARQL endpoint that can be interrogated with AskOmics.

Since we added the neXtProt abstraction into our AskOmics instance, we can link our data to neXtProt.

Hands-on: Find human homolog genes
  1. Go back to the gene node
  2. instantiate Gene Symbol and hide his Label
  3. Instantiate Homolog Group, hide his label and filter his Common Organism Name with human
Organism name filter. Open image in new tab

Figure 17: Organism name filter
  1. From Homolog Group, instantiate another Gene Symbol and hide his Label
  2. Finally, follow the to neXtProt Gene link and instantiate Gene (with a capital G)
  3. Run & preview

The query we’ve just built asks for the human homologs of our over-expressed genes. We use the Gene Symbol to get information from the neXtProt database. AskOmics converts the query into small SPARQL subqueries and send them to the local database and to the remote neXtProt endpoint.

Now we are linked to the neXtProt database, we can obtain information about the proteins encoded by these genes, as well as their location in the cell.

Hands-on: Get the protein and their location
  1. Instantiate Entry
  2. Instantiate Isoform and hide the Label
  3. Many nodes are connected to Isoform. Use the Filter links field to filter nodes linked with a link named location
Filter links. Open image in new tab

Figure 18: Filter links
  1. Instantiate the Subcellular Location node and hide Uri
  2. Use the Filter node field to filter nodes with “location” in their name
  3. Instantiate Uniprot subcellular Location CV (you can use the node filter to clear up the screen)
  4. Run & preview

Finally, our query is “All genes that are over-expressed and located on a QTL that is related to growth, their human homologs and the location of the proteins coded by this genes”. We will save it to the results.

Hands-on: Save a result
  1. Run & save

Results management

The results page displays the saved queries. Queries are sorted by creation date. At the end of the table, action buttons can be used to preview the result, download or send it to Galaxy history.

Hands-on: Edit query name
  1. Go to the Results page
  2. Use the Preview button to check the result
  3. Click on the name to rename the two query with Over-expressed genes on a growth QTL and Over-expressed genes on a growth QTL, their human homologs and protein location (press enter key to validate)
preview table. Open image in new tab

Figure 19: Results table

The Action column contain button to perform certain action:

  • Preview: show a results preview on the bottom of the table
  • Download: Download the results (TSV file)
  • Edit: Edit the query with the query builder
  • SPARQL: edit the query with a SPARQL editor for advanced users
  • Send results to Galaxy: send the results (TSV file) to the most recently used Galaxy history
  • Send query to Galaxy: send the query representation (json file) to the most recently used Galaxy history
Hands-on: Send results to Galaxy
  1. Click on Send results to Galaxy on each query to send them to the last used Galaxy history
  2. Get back to galaxy and wait for the dataset (reload if needed)
history results. Open image in new tab

Figure 20: Galaxy history with the two results

Now that you have used AskOmics to generate this final tabular file, you can continue analysing it with other Galaxy tools. If you are done, don’t forget to close the AskOmics instance by going to the “User” > “Active Interactive Tools” page.

Conclusion

In this tutorial we have seen how to use AskOmics Interactive Tool. We launch the tools with a set of input files, then we have integrated these files into RDF and finally, we built complex queries over this local datasets and neXtProt to answer a biological question.