Thread by @stevenemassey

A new animal coronavirus, possibly from bamboo rat, is present in abundance in sample Q61 from the Huanan Seafood Market

This find, however, stands in stark contrast to the single SARS2 read present in the same sample, which has a high proportion of raccoon dog sequences 🧵

2/ Sample Q61 was generated by Liu et al in their survey of the Huanan Seafood Market

It has generated discussion for having a high level of raccoon dog reads, which has been postulated by some workers as an intermediate host for SARS2
nature.com/articles/s4158…

3/ Previously, I had identified reads related to human OC43 /HKU1 coronaviruses using the coronascan procedure

https://twitter.com/stevenemassey/status/1647286339367481346?s=20

4/ Subsequently, several other investigators and myself determined that these reads were more closely related to rodent coronaviruses using Blast (this allowed matching to partial CoV sequences, and additional whole genome sequences)

https://twitter.com/stevenemassey/status/1649837163033530368?s=20

5/ The next step is to add the complete coronavirus genomes that were hit in the Blast search to the coronascan procedure

Coronascan systematically maps NGS reads against a dataset of complete coronavirus genomes, to find the best match
github.com/semassey/Scann…

6/ The coronascan results show that reads from Q61 map best to the Bamboo rat coronavirus GX/GX19-89/2019 genome (OQ297694)

The genome coverage is 15.3 %

(note I tweaked bowtie2 mapping by adding -a and --non-deteministic options, which allows more multi-mapping of reads)

7/ A level of 15.3 % is comparable to other wildlife associated RNA viruses in sample Q61

https://twitter.com/stevenemassey/status/1644422605720674304?s=20

8/ The coronavirus genomes with the next best % coverage were:

rodent CoV RCoV/GD/2020 genome (MW855473) 8.9 %

Bamboo rat CoV GX-F1-1 genome (OM480511) 8.4 %

Bamboo rat CoV GX/GX18-49/2018 genome (OQ316388) 6.4 %

9/ Given the close relationships of these four CoVs with OC43/HKU1, they are likwise embecoviruses, which are beta-coronaviruses

SARS2 is a sarbecovirus, which is another subgenus of beta-coronavirus
en.wikipedia.org/wiki/Embecovir…

10/ While it is difficult to be certain as to the host animal of the new CoV detected in the Q61 sample, the type of correlational analysis by @jbloom_lab across all NGS samples might reveal if the novel CoV is correlated with bamboo rat reads
biorxiv.org/content/10.110…

11/ Liu et al also did a qualitative spatial correlation analysis which could likewise be informative:

https://twitter.com/stevenemassey/status/1655952641527275527?s=20

12/ Previously, I had noted that the coronascan procedure could be improved by adding more complete and partial coronavirus genomes

However, partial genomes would confuse the ranking using # reads and % coverage. Some form of normalization would be required therefore

13/ Ranking whole coronavirus genomes on the basis of % coverage is useful as % coverage is related to phylogenetic distance (as well as viral titer in the sample)

In addition, masking of low-complexity regions using bbmask.sh should improve accuracy h/t henjin

14/ When I mapped the Q61 reads to the GX/GX19-89/2019 genome only, there was a substantially higher level of mapping: 32.9 % coverage

This is due to absence of other genomes, to which reads might cross-map

Cross-mapping dynamics need to be better understood and controlled for

15/ I took the 331 reads that mapped to GX/GX19-89/2019 only, and tried to assemble them using Megahit

Unfortunately, Megahit did not generate any assembled contigs

(contigs can give precise information regarding SNVs, when compared to the GX/GX19-89/2019 genome)

16/ I then displayed the mapped reads on the GX/GX19-89/2019 genome using Integrative Genome Viewer (IGV) - they are fairly evenly mapped across the genome, although somewhat denser in the second half of the genome

17/ henjin (from Discord metagenomics chat) did a more comprehensive analysis than me, with a larger # of complete CoV genomes and did a distance based clustering tree using ggtree

henjin found the 3 bamboo rat CoVs group between OC43 and HKU1 CoVs

18/ "In the plot ... the x-axis shows the distance to the bamboo rat virus which had the most aligned reads from Env_0576, and the y-axis shows the distance to the HKU1 reference genome"

19/ "Each point is connected with a line to its two neighbors. "Mouse coronavirus PREDICT/PDF-2560" is connected to two of the bamboo rat coronaviruses even though it has over 7,000 nucleotide changes from "Bamboo rat coronavirus isolate GX/GX19-89/2019""

h/t henjin

20/ Given that some of the novel CoV reads are divergent from the GX/GX19-89/2019 genome, possessing SNVs, this means that there are likely additional reads present that do not map to GX/GX19-89/2019 (due to the sequence divergence)

21/ Consequently, the number of reads derived from the new coronavirus is likely higher than 331, strongly outnumbering the single SARS2 read

22/ The value of this novel CoV is that it shows that the # of reads that can be expected from an animal-associated CoV in the market NGS datasets is in the order of 100s

Also, that a market animal CoV has not degraded prior to sampling, being present in significant levels

23/ In the light of these considerations, the presence of only a single SARS2 read, in a sample with a large proportion of raccoon dog reads, appears increasingly inconsistent with the presence of infected raccoon dogs

@threadreaderapp unroll