FAQ.md

Questions regarding setup and usage:

1. Q: How can I stop a running docker instance of Galaxy-GraphClust?

   A: If you are runnig the container in interactive mode (i.e. docker run -i) use Ctrl+C. To stop ALL docker instances on your computer you can run this command in terminal: sudo docker stop $(sudo docker ps -a -q)

2. Q: After a few experiments and upgrades, lots of disk storage is occupied. How can I clean it up?

Docker takes a conservative approach for cleaning up unnecessary data objects. Below some solutions for cleaning up your hard disk is coming, ordered in the level of conservativeness:

• Using docker system prune command, manual here
• Please make sure no unintended container instance is running on the background. You can get a list of all containers with docker ps -a and remove them if necessary with docker rm ID-or-NAME.
• The above steps do not remove the dangling and not needed images which usually take most of the space. You can use docker images to get a list of them, and docker rmi image-ID to remove individual images.
• To auto-remove a container after exiting, you can use docker run --rm.
• A detailed tutorial about these and further ways can be found here: https://www.tecmint.com/remove-docker-images-containers-and-volumes/

3. Q: I would like to customize the workflow settings but there are so many parameters there. What can I do?

GraphClust2 workflow is collection of more than 15 tools where a majority invoke slightly complex methodologies. We have provided the pre-configurations that we think would be needed by the users, accroding to our own experience and the feedback from the GraphClust2 users and collaborators. Each tool wrapper is supplemented with brief help descriptions for the arguments and/or external links to the tool's documentation. We are extending, the in-Galaxy help descriptions and Galaxy tutorials, your feedback is very appreciated. If you would like to customize the configurations and do not know how to start, we would recommend to start with adapting the first and last steps of the workflow.

GraphClust2 takes a windowing approach for folding and clustering long input sequences. The windows size and overlapping ratio can be adapted to the expectation of the structure features. Starting with shorter window-lengths (50-100nt) would be a good idea, specially if it's not known that the putative structured elements are covering the entire sequence or local elemnts. Using shorter windows, the small elements such as stem-loops can be identified, afterwards you may re-run the pipeline with an increased windows lenght to capture the complete structure.

In the last step cluster_collection_report, GraphClust2 assigns the elements to the best matching cluster and also aligns the best (top) matching entries of each cluster. results_top_num defines how many of the top entries to align, increasing the number would be a good idea to find covariations and identify a reliable conserved element. Usually aligning the top10-30 or higher would help to identify reliable structure conservations and covariations. The other parameter to consider is the covariance model hit criteria (E-value or bitscore). The E-value works very well (and designed for )specially for structured non-coding RNAs with defined boundaries, like sequences in the Rfam database. We have found switching back to the CM-bit score option (option Use CM score for cutoff), to work better for identifying structured elements surrounded (within) a sequence context.

4. Q: The workflow runs forever on my computer. Isn't the liner run-time on of the highlighted remarks?

The apparent practical bottleneck of the workflow specially for local instances is the covariance model calibration. Specifically the cmcalibrate step integrated into the cmbuild Infernal wrapper. This calibration is necessary to compute a reliable E-value for the significance of a CM hit, but time-consuming for generating ~million bases of background sequences.

We would suggest to use the instance on our European Galaxy server, where the cmbuild step is pre-configured to use multi-processors and also the server is supported by thousands of computing nodes. In the docker instance, by default, the calibration is performed on a single core. We would recommend to use ask your galaxy admin to configure the wrapper according to the backend hardware. Alternatively, you can reduce the length of random sequences (-L in cmbuild-cmcalibrate) wrapper. Please refer to the Infernal manual and also take care that the E-values might not be reliable anymore.

5. Q: How can I do more rounds?

   A: To extend an existing workflow of GraphClust and add another round, you should run a workflow called Galaxy-Workflow-single_round_for_extension : GraphClust_two The inputs for this workflow are the files generated by the GraphClust workflow. The names of each input corresponds to the name of the produced file, so you should just choose from a dropdown selection a needed file. Important parameter for this workflow is the round number, which must be specified in NSPDK_cancidateCluster, pgma_graphclust and cluster_collection_report tools. Alternatively it is recommended to use the with-subworkflows flavors.

6. Q: In my Ubuntu host system the container is running but constantly reports error: could not connect to server: Connection refused

   A0: For Ubuntu users we recommend to use 16.04 LTS version which is deleivered with Kernel 4.2 or higher.

   A1: Docker manager is tightly coupled with the host Linux kernel. Under certain Linux kernel the docker storage system might fail. Please proceed with the following commands or contact you administrator ( Warning please be careful of potential data loss with this procedure):

sudo apt update; sudo apt upgrade;
sudo apt-get install linux-image-extra-$(uname -r)  linux-image-extra-virtual
sudo modprobe aufs
sudo service docker stop
sudo rm -rf /var/lib/docker/overlay2
sudo service docker restart

For more information please check Docker documentation: https://docs.docker.com/engine/userguide/storagedriver/aufs-driver/

Login to the docker instance:

To have distinct history and workflows the Galaxy server requires each user to register for first access time. By default anyone with access to the host network can register. No registration confirmation email will be sent to the given email. So you can register with any custom (including non-existent) email address. There exist also a default Admin user described here. To change the default authorization settings please refer to the Galaxy Wiki section Authentication

• To register (first time only):

  • On top right of the panel goto User→Register
  • Provide a custom email address and password, confirm your password and enter a public name

• To login:

  • On top right of the panel goto User→Login
  • Provide your registered email address and password