1. Home
  2. VarSeq
  3. VarSeq Known Issues
  4. VarSeq 2.2.3 Known Issues and Work Arounds
  1. Home
  2. VarSeq
  3. VarSeq 2.2.3 Known Issues and Work Arounds

VarSeq 2.2.3 Known Issues and Work Arounds

1. Incorrect ACMG classifier results when using VarSeq 2.2.2 or 2.2.3 to create projects from templates saved in VarSeq 2.2.1

  • Fix: Run a “patch” tool for existing templates to continue using VarSeq 2.2.2 and 2.2.3. See instructions for use below.

Updates were made to the ACMG Classifier for PVS1 and BS2 recommendations in VarSeq 2.2.2. These changes were largely around the handling of population catalogs and variants called in a small number of individuals. Due to a bug in the project upgrade logic, per-criteria rules were not correctly updated for PVS1 and BS2. VarSeq templates created in 2.2.1 that contain a ACMG classifier algorithms will not ever recommend PVS1 or BS2 for any variants when used to create projects in VarSeq 2.2.2 or 2.2.3.

Work around: For VarSeq 2.2.3 users, the ACMG Classifier algorithm can be edited by right-clicking on the algorithm header and re-ran. After re-running, variants are correctly recommended PVS1 criteria.

For VarSeq 2.2.2 users, deleting and re-adding the ACMG Classifier algorithm will result in the correct behavior.

Fix: Use the TemplateUpdater tool to patch existing project template files to be corrected so that existing VarSeq 2.2.2 and 2.2.3 installs produce correct results.

TemplateUpdater Tool

We have created a specialized command line tool to upgrade VarSeq 2.2.1 templates that include the ACMG Classifier algorithm to a version compatible with VarSeq 2.2.2 and later that includes updated logic to correct the PVS1 and BS2 criteria options.

Note that VarSeq 2.2.4 and newer will have the corrected project upgrade logic applied. This tool is unnecessary when VarSeq 2.2.4 or newer is being used for all production workflows.

If you are using VarSeq 2.2.2 or VarSeq 2.2.3 and the ACMG Classifier algorithm to score variants in your spreadsheets, we strongly recommend you use this tool to update your templates to avoid incorrect classifications and filtering of variants in your sample workflows.

Download and unzip the following folder:


Recommended usage:

1. Back up your VarSeq templates folder (copy the “ProjectTemplates” folder to another location). To access your VarSeq template folder, go to Tools–> Open Folder–> User Data Folder–> ProjectTemplates.

2. Open a command shell and navigate to your VarSeq templates folder.

3. Run the tool in dry run mode using the command below to test the results of running the command and see which templates are impacted. The “-d” stands for “dry run”. It will report what it would do, but does not make any changes.

templateupdater.exe -d

4. If the template folder is not found (because they are in a custom path), you can specify the full path to the folder with the project template with -t

Figure 1: Running the templateupdater tool in command shell

Each found project is scanned and updated if the ACMG classifier algorithm in need of updating is found.

5. Once you have backed up your project older and are happy with the results reported from the “dry run”, remove the “-d” flag and the project files will be updated.

6. Confirm that using the these project templates in VarSeq 2.2.2 or later to create new projects now scores these criteria as expected.

2. ShortPathMode

When creating projects on a network drive and in some cases on local drives, there is a 260 character path limit in Windows. When this limit applies, creating projects and files with long names was causing some issues. This often happens half way through the project algorithms running, as our data folder has “informative” names for the files it creates. Therefore we created  “shortPathMode” and it should better handle the use case of projects being created on network drives with a 260 character path limit.

With this mode enabled, files in the data sub-folder of the project with receive names such as 1.tsf, 2.tsf and so forth. It should still be recommended that project names and their starting directories do not total in length of 200 characters, given that some files may continue to use up to 60 characters. For example, under the project folder, the project file will when open have a name of “<project_name>.vsproject-autosave”.

You can see the new file naming by going to Tools > Open Folder > Project Folder > data. This fix will only apply to new projects as it will not rename files that already had the problem.

Fix: Create and include a hosts.json file in the VarSeq Program folder (Tools > Open Folder > Program Folder). Below is an example of the hosts.json file.


2. Web or VSClinical View Blank When Running VarSeq on a Network Drive

  • Workaround: Run VarSeq from local drive or use extra command-line switch

In VarSeq 2.2.3 the web browser built into the software (Chromium) was updated with improved security settings (“sandbox” mode) that prevents it from running on a network drive due to operating system limitation. You will see a blank screen when either trying to start VSClinical workflows or opening a Browser tab.

However, when VarSeq is installed and run from a system local drive, there is no problem running the Chromium browser engine.

Work around: Move the VarSeq installation to local system drive.

Fix: Add –no-sandbox as a command line argument for running VarSeq. You can edit the shortcut to VarSeq on the desktop and add this to the end of the “Target” line for example.

Adding the –no-sandbox option to VarSeq to support running it from a network mapped drive

3. Multi-allelic Insertions/Deletions in Trio Analysis

In trio projects, the inheritance of variants that were in a multi-allelic state in one parent and present in a child with a different variant representation in the child was not being detected. The genotypes of the parents were not accurately filling in for trio projects to identify the variant as inherited in the case of a multi-allelic insertion/deletion in one parent with only the insertion being inherited by the child.

Fix: Update to VarSeq 2.2.3 SP4 or newer.

4. Slow CNV Auto-Classifier Algorithm with VSWarehouse

The CNV ACMG Auto-Classifier was performing lookups for entries in the gene dosage catalog individually instead of in batches. This resulted in significant slowdown when using a VSWarehouse catalog compared to local catalogs, especially for large CNV events.

Work around: Using local gene dosage catalog does not affect the performance of the CNV auto-classifier.

Fix: Update to VarSeq 2.2.3 SP4 or newer.

5. Issue Specific Variants as Biomarkers in VSClinical AMP

Stop-gain and certain slice acceptor variants were not able to be added to the AMP Guideline workflow as biomarkers, only as variants.

Fix: Update to VarSeq 2.2.3 SP4 or newer.

8. VSPipeline Crashing On Shutdown

VSPipeline was crashing with a “segmentation fault” message instead of cleanly shutting down with a non-error code if the VarSeq or VSPipeline instance was configured to VSWarehouse. No project data is lost.

Work around: Disconnect the VarSeq instance from VSWarehouse

Fix: Update to VarSeq 2.2.3 SP4 or newer.

9. External Haploid CNV Records Detected as Deletions

In CNV VCFs created by Dragen, the X Chromosome males has a copy number of 1. However, the VarSeq CNV Caller is calling these events as Het Deletions when they should be diploid/normal when the genotype for the CNV is reference given the haploid chromosome default copy number is 1.

Work around: Contact customer support for more details at support@goldenhelix.com.

10. VSPipeline multi-sample CNV VCF import causes a segmentation fault

When using the update_cnv_import command to import a multi-sample CNV VCF, VSPipeline crashes in the import logic.

Work around: Use multiple single-sample CNV VCF files with VSPipeline

Fix: Update to VarSeq 2.2.3 SP5 or newer.

In VSClinical, there are two specialized questions that provide a PM4 criteria: 1) in-frame variants and 2) stop loss variants.

When filling in an interpretation for a variant saved with one of these PM4 criteria, the internal state for both criteria is filled in with the criteria selection but only one question is visible in the UI, making it appear as if PM4 is applied twice.

Work around: Manually ignore the influence of two PM4 criteria in the classification

Updated on August 24, 2021

Was this article helpful?

Related Articles

Leave a Comment