ORF9b addition

[zach-hensel]

Great work! I made a small local hack to output fitness for ORF9b for a quick check of XBB.1.16 which has a few ORF9b mutations that might be an interesting addition.

covSpectrum query: https://cov-spectrum.org/explore/World/AllSamples/AllTimes/variants?nucMutations=T12730A%2CT28297C%2CA28447G&

Changes made:

1. Modified gtf file and replaced download in relevant rule with copying the local modified file:

NC_045512v2	ncbiGenes.genePred	transcript	28284	28577	.	+	.	gene_id "ORF9b.1"; transcript_id "ORF9b.1"; 
NC_045512v2	ncbiGenes.genePred	exon	28284	28577	.	+	.	gene_id "ORF9b.1"; transcript_id "ORF9b.1"; exon_number "1"; exon_id "ORF9b.1";
NC_045512v2	ncbiGenes.genePred	CDS	28284	28577	.	+	0	gene_id "ORF9b"; transcript_id "1"; exon_number "1"; exon_id "ORF9b.1";
NC_045512v2	ncbiGenes.genePred	start_codon	28284	28286	.	+	0	gene_id "ORF9b"; transcript_id "ORF9b.1"; exon_number "1"; exon_id "ORF9b.1";
NC_045512v2	ncbiGenes.genePred	stop_codon	28575	28577	.	+	0	gene_id "ORF9b"; transcript_id "ORF9b.1"; exon_number "1"; exon_id "ORF9b.1";

2. Modified first cell in aamut_fitness.py.ipynb to not remove overlapping N and ORF9b mutations.
   .query("not is_overlapping or gene=='N;ORF9b'")

Result: Only examined briefly. ORF9b I5T in XBB.1.16 ranks highly.

[jbloom]

Great, thanks @zach-hensel. Marc Johnson had also been asking about ORF9b.

I am going to make a pull request that adds this into the pipeline. Just adding a few more notes re ORF9b mostly for myself while doing this:

• Tweet by Daniel Focosi re ORF9b: https://twitter.com/dfocosi/status/1628662576027410433?t=nwSRxBWUJ3XLQQdv-ohONQ&s=09
• This paper defines the same coordinates (28284 to 28577) for ORF9b that you give in your manually edited GTF above: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8317007/
• UCSC browser annotation of ORF9b here, again matching the coordinates of 28284 to 28577 you give above.

My pull request will automate the modification of the GTF that you did manually above to place it in workflow of larger pipeline.

[zach-hensel]

Awesome and thank you for double checking. The nsp numbers here also come indirectly from me hastily copying some things and I overlooked the frameshift.

I am not working with this anymore so here are a couple other observations. First, nsp6 L37 would be interesting to unmask. I suppose it's masked because of a combination of occuring early and artifacts in Orf1a sequencing. Second, I made a quick script to rank mutations from the CSV one can export from a cov-spectrum query and looked at XBB.1.16. A reversion of one of the three nucleotides mutated in BA.2 for Orf6 61 popped up with a major fitness increase and you may want to mask this site as it appears to be an artifact. Lastly, it might be interesting to look at predicted fitness of RecCA and particularly for U-to-C mutations in that direction. Orf8 L84S is also interesting in that respect.

[jbloom]

Great, thanks. I will look at the ORF6 site. For the other masked mutations, including L84S, I had to mask all mutations to the Wuhan-Hu-1 reference as nearly all of them have unrealistically high counts that likely indicate some sort of bioinformatics issue such as calling uncovered sites to reference.

[jbloom]

OK, I added estimates and summarized results here: https://twitter.com/jbloom_lab/status/1636470443493449728