search

pangenome / pggb

the pangenome graph builder
https://doi.org/10.1101/2023.04.05.535718
MIT License
355 stars 38 forks source link

Problem with the ODGI 1D visualization by path orientation #308

Open KopalliV opened 1 year ago

KopalliV commented 1 year ago

When I run the conda version of PGGB 0.5.3 , the 'ODGI 1D visualization by path orientation' result I get shows very large number of large inversions (which do not exist in my alignment dotplots). The percentage of inversions fluctuates greatly every time I try to re-run it. Can anyone help me with a solution to this problem.?

AndreaGuarracino commented 1 year ago

Can you share more details sych as version of all tools in PGGB (the log files would help), command lines (see log files), and a few ODGI VIZ visualizations?

From: KopalliV @.> Sent: 20 June 2023 18:03 To: pangenome/pggb @.> Cc: Subscribed @.***> Subject: [pangenome/pggb] Problem with the ODGI 1D visualization by path orientation (Issue #308)

When I run the conda version of PGGB 0.5.3 , the 'ODGI 1D visualization by path orientation' result I get shows very large number of large inversions (which do not exist in my alignment dotplots). The percentage of inversions fluctuates greatly every time I try to re-run it. Can anyone help me with a solution to this problem.?

— Reply to this email directly, view it on GitHubhttps://github.com/pangenome/pggb/issues/308, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AO26XHXL27S6ZLJL6YI77WLXMHCUPANCNFSM6AAAAAAZNQFRK4. You are receiving this because you are subscribed to this thread.Message ID: @.***>

KopalliV commented 1 year ago

Command: /vol/volume/miniconda3/envs/pggb-env/bin/pggb -i chr03_combined_SBi_no_outliers.fasta.gz -n 6 -D temp/ -o chr3_pggb/' Versions: wfmash: v0.10.3-3-g8ba3c53 seqwish:v0.7.9-0-gd9e7ab5 smoothxg:v0.6.8-0-ga8a0e9e odgi: v0.8.2-0-g8715c55 gfaffix: v0.1.4 vg: v1.40.0 reporting: v1.14

The following images are the results of two runs with same parameters of the same set of data.

Chr3_S_Bi_no_outliers fasta gz 4725b32 417fcdf 9fb5d42 smooth final og viz_inv_multiqc 2 Chr3_S_Bi_no_outliers fasta gz 4725b32 417fcdf 9fb5d42 smooth final og viz_inv_multiqc

KopalliV commented 1 year ago

Update: I have updated to the latest version of PGGB(0.5.4) on conda and tried to re-run it on the same set of assemblies. I still face the same issue with the visualization of the path orientation. I would be happy to provide the publicly available assembly data I am using to run my analysis for your reference. I kindly request for a solution to this problem.

ekg commented 1 year ago

The orientation plot shows orientation relative to the sense of the graph nodes.

If all your paths are red or black, you have no inversions between them. I agree based on these plots that you don't have any large inversions.

ekg commented 1 year ago

To say this another way, if at the same place you have one path that is red and the other black, this would mean "inversion". If you have both black or both red, there is not an inversion.

subwaystation commented 1 year ago

There is a component in pggb that is non-deterministic, a graph sorting algorithm. As a final step, we also combine this with a so called grooming which resolves spurious inverting links. This combination of tools could lead to your different outcome, though, I am surprised that the two graphs look so differently. Please sort the graphs by path position with odgi sort -i <INPUT_GRAPH> -o <OUTPUT_GRAPH> -Y -t 16 and then visualize both again with odgi viz. This will give us a better feeling of what might be going on.