Mismatching SARS-CoV-2 Sequences

[TheHarshShow]

Hi there,

I created a PanGraph of 200 SARS-CoV-2 sequences using FASTA sequences as input, and it seems that eleven of them aren't represented incorrectly in the JSON file. I have uploaded the data here. The original FASTA file is denoted by sars_200_orig.fa. The represented sequences (determined by me) are represented by sars_200_pangraph.fa, and the PanGraph JSON file is denoted by sars_200.json. The sequences that we believe aren't matching are England/BRBR-2B7C38D/2021|OV263009.1|2021-11-22, IMS-10178-CVDP-0E892CAB-4101-45AD-A5AB-82C23A77B85B|OX112182.1|2021-10-14, Denmark/DCGC-179132/2021|OW435830.1|2021-10-02, SouthAfrica/NHLS-UCT-GS-AD95/2021|OM739820.1|2021-08-30, IMS-10150-CVDP-7250DCF0-8B47-40DA-89AF-8E56669A8CB5|OU964784.1|2021-10-12, USA/CA-CDC-FG-175698/2021|OL666921.1|2021-11-18, Denmark/DCGC-196557/2021|OW446795.1|2021-10-24, Denmark/DCGC-151767/2021|OV830941.1|2021-08-12, USA/MA-CDCBI-CRSP_4TOCNN2I3HYX32WD/2021|MZ752955.1|2021-08-02, England/LOND-12FD57B/2021|OU391062.1|2021-05-23 and RNA|OX380648.1|2022-10-22.
Can you please look into it?

Best,
Harsh

[mmolari]

Hi Harsh,
thank you once more for your feedback! I'll look into it and report back.
Have a nice start in the year!
Marco

[mmolari]

Hi Harsh,
I did a preliminary test by building the graph with pangraph v.0.7.3 and the command:

pangraph build --test sars_200_orig.fa > graph.json

And I cannot reproduce the error, in the output graph all sequences are correctly reconstructed. When you get the error do you use other specific flags to build the graph?
I also looked into the graph you shared and I agree with you, exactly the sequences that you mentioned are incorrectly reconstructed.

Ps: another thing worth mentioning is that some sequences contain long stretches of N nucleotides, in few cases these cover 10%-30% of the sequence. PanGraph does not work well with that, it might cause excessive fragmentation of the graph or problems with alignment. Just thought I'd mention it so that if they are not central to the analysis maybe you can filter them out.

[TheHarshShow]

Hi Marco,

Thanks a lot for looking into this issue. I will investigate this issue again with my team and let you know the updates on whether we did anything wrong. Also, thanks for pointing out the issue about sequences containing many N's. I believe this might have contributed to some inefficient PanGraphs representing 20k SARS-CoV-2 sequences. I'll investigate this too.

Thanks,
Harsh