Troubleshooting#

This page deals with common issues in running the analysis. For issues with installing or running the software please raise an issue on github.

Most/all of my samples merge when I run a query#

If you see a gigantic merge when running poppunk_assign, for example:

Clusters 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16 have merged into 1_2_3_4_5_6_7_8_9_10_11_12_13_14_15_16

This might be caused by:

  • A single/a few ‘bad’ genome(s) with artificial zero/close-to-zero distances to many other samples, which links much of the network together.

  • A boundary which is actually a bit lax, and when more isolates are added, they have distances just below the boundary between a few clusters, and there are enough of these to link more and more clusters together.

The first issue should be easy to fix with one of the following methods:

  • Use --run-qc (see Data quality control (--qc-db)) with one or more of:

    • --max-pi-dist and/or --max-a-dist to remove outlier distances.

    • --max-zero-dist which sets the maximum proportion of zeros to other samples. Poor quality genomes often end up with zero distances to many samples, linking them together.

    • --max-merge which sets the maximum number of clusters a single query can cause to merge.

    • --betweenness which will print all queries in order which have high stress in the network, and are likely causing merges.

  • Use --serial to run samples through one-by-one. This is a little less efficient than fully batched querying, but much faster than running independent jobs. Note, lineage models and updating the database are not supported with this mode.

The second issue above is potentially more insidiuous, and may require a refit to all the data to obtain a tighter boundary. You can (mostly) keep old cluster IDs via the use of --external-clustering if you do this. Alternatively, you can add the --serial command to type samples one at a time as above.

See issue 194 for more discussion.

Memory/run-time issues#

Here are some tips based on experiences analysing larger datasets:

  • Add --threads – they are used fairly efficiently throughout.

  • Consider the --gpu-sketch and --gpu-dists options is applicable, and a GPU is available.

  • In --refine-model set --pos-shift 0 to avoid creating huge networks with close to \(N^2\) edges. Mixture models normally need to be pruned.

  • In --refine-model you may add the --no-local option to skip that step and decrease run-time, though gains are likely marginal.

  • Use --rapid-nj, if producing microreact output.

Another option for scaling is to run --create-db with a smaller initial set (not using the --full-db command), then use --assign-query to add to this.

Known bugs#

Calculating query-query distances when unnecessary#

A bug in v2.4.0 only, (fixed in v2.5.0 and not in previous versions).

You will always see Found novel query clusters. Calculating distances between them. when running poppunk_assign with more than one input sample. This should only happen when unassigned isolates/novel clusters are found. Our check on this condition became invalid.

Additionally, this may have affected typing when query-query links were present, this appeared as invalid merges in some tests. If you used this, you may wish to re-run with v2.5.0 or higher.

Older HDBSCAN models fail to load#

tl;dr if you see an error ModuleNotFoundError: No module named 'sklearn.neighbors._dist_metrics' you probably need to downgrade sklearn to v0.24.

The change in scikit-learn’s API in v1.0.0 and above mean that HDBSCAN models fitted with `sklearn <=v0.24` will give an error when loaded. If you run into this, the solution is one of: - Downgrade sklearn to v0.24. - Run model refinement to turn your model into a boundary model instead (this will change clusters). - Refit your model in an environment with `sklearn >=v1.0.

If this is a common problem let us know, as we could write a script to ‘upgrade’ HDBSCAN models. See issue [#213](https://github.com/bacpop/PopPUNK/issues/213) for more details.

When I look at my clusters on a tree, they make no sense#

This is a bug caused by alphabetic sorting of labels in PopPUNK >=v2.0.0 with pp-sketchlib <v1.5.1. There are three ways to fix this:

  • Upgrade to PopPUNK >=v2.2 and pp-sketchlib >=v1.5.1 (preferred).

  • Run scripts/poppunk_pickle_fix.py on your .dists.pkl file and re-run model fits.

  • Create the database with poppunk_sketch --sketch and poppunk_sketch --query directly, rather than PopPUNK --create-db.

I have updated PopPUNK, and my clusters still seemed scrambled#

This is possible using query assignment with --update-db, or in some cases with --gpu-dists. Please update to PopPUNK >=v2.4.0 with pp-sketchlib >=v1.7.0

Calculating distances using 0 thread(s)#

This will lead to an error later on in execution. This is due to a version mismatch between PopPUNK and pp-sketchlib. Installation of both packages via conda should keep the versions compatible, but there are ways they can get out of sync.

The solution is as above: upgrade to PopPUNK >=v2.2 and pp-sketchlib >=v1.5.1.

Error/warning messages#

Row name mismatch#

PopPUNK may throw:

RuntimeError: Row name mismatch. Old: 6999_2#17.fa,6259_5#6.fa
New: 6952_7#16.fa,6259_5#6.fa

This is an error where the mash output order does not match the order in stored databases (.pkl). Most likely, the input files are from different runs, possibly due to using --overwrite. Run again, giving each step its own output directory.

Samples are missing from the final network#

When running --assign-query an error such as:

WARNING: Samples 7553_5#54.fa,6999_5#1.fa are missing from the final network

Means that samples present in --distances and or --ref-db are not present in the loaded network. This should be considered an error as it will likely lead to other errors and warnings. Make sure the provided network is the one created by applying the --model-dir to --distances, and that the same output directory has not been used and overwriten by different steps or inputs.

Old cluster split across multiple new clusters#

When running --assign-query, after distances have been calculated and queries are being assigned warnings such as:

WARNING: Old cluster 1 split across multiple new clusters

Mean that a single cluster in the original clustering is now split into more than one cluster. This means something has gone wrong, as the addition of new queries should only be able to merge existing clusters, not cause them to split.

Most likely, the --previous-clustering directory is inconsistent with the --ref-db and/or --model-dir. Make sure the clusters are those created from the network being used to assign new queries.

If you want to change cluster names or assign queries to your own cluster definitions you can use the --external-clustering argument instead.