ktmeaton / ncov-recombinant Goto Github PK

So I have a new idea for identifying the parent lineage.

England/MILK-3796834/2022 is an XM recombinant, with regions predicted by sc2rf:

• 44:17410|Omicron/21K
• 21618:29510|Omicron/21L

From nextclade, the mutations by region are as follows:

• 44:17410
  • C241T,C2470T,A2832G,C3037T,T5386G,G8393A,C10029T,C10449A,A11537G,C12513T,T13195C,C14408T,C15240T
• 21618:29510
  • C21618T,G21987A,T22200G,G22578A,C22674T,T22679C,C22686T,A22688G,G22775A,A22786C,G22813T,T22882G,G22992A,C22995A,A23013C,A23040G,A23055G,A23063T,T23075C,A23403G,C23525T,T23599G,C23604A,C23854A,G23948T,A24424T,T24469A,C25000T,C25416T,C25584T,C26060T,C26270T,C26577G,G26709A,C26858T,A27259C,G27382C,A27383T,T27384C,C27807T,A28271T,C28311T,G28487A,G28881A,G28882A,G28883C,A29510C

And if we query these mutations in cov-spectrum with LAPIS...

Parent 1 | 44:17410

Parent 1 is mostly likely BA.1.1.10 (72%, 647/894.)

https://lapis.cov-spectrum.org/open/v1/sample/aggregated?fields=pangoLineage&nucMutations=C241T,C2470T,A2832G,C3037T,T5386G,G8393A,C10029T,C10449A,A11537G,C12513T,T13195C,C14408T,C15240T

{
  "errors":[],
  "info": {
    "apiVersion":1, 
    "dataVersion":1656461191,
    "deprecationDate":null,
    "deprecationInfo":null,
    "acknowledgement":null
  },
  "data":[
    {"pangoLineage":"B.1.1","count":1},
    {"pangoLineage":"BA.1.1.18","count":30},
    {"pangoLineage":"BA.1.1.12","count":3},
    {"pangoLineage":"BA.1.1.10","count":647},
    {"pangoLineage":"BA.1.1","count":186},
    {"pangoLineage":"BA.1","count":27}
  ]
}

Parent 2 | 21618:29510

Parent 2 is mostly likely BA.2 (83%, 38/46). However, there is only one runner up, and it is BA.2.12.1 (17%) which falls within BA.2.

https://lapis.cov-spectrum.org/open/v1/sample/aggregated?fields=pangoLineage&nucMutations=C21618T,G21987A,T22200G,G22578A,C22674T,T22679C,C22686T,A22688G,G22775A,A22786C,G22813T,T22882G,G22992A,C22995A,A23013C,A23040G,A23055G,A23063T,T23075C,A23403G,C23525T,T23599G,C23604A,C23854A,G23948T,A24424T,T24469A,C25000T,C25416T,C25584T,C26060T,C26270T,C26577G,G26709A,C26858T,A27259C,G27382C,A27383T,T27384C,C27807T,A28271T,C28311T,G28487A,G28881A,G28882A,G28883C,A29510C

{
  "errors": [],
  "info":{
    "apiVersion":1,
    "dataVersion":1656461191,
    "deprecationDate":null,
    "deprecationInfo":null,
    "acknowledgement":null
  },
  "data":[
    {"pangoLineage":"BA.2","count":38},
    {"pangoLineage":"BA.2.12.1","count":8}
  ]
}

Resolving

There are a couple of options to resolve the proportions:

• Exclude lineages by a hard cut-off (<1%, <10%, etc.)
• Take the highest proportion lineage.
• Consider lineages in descending order, and report a lineage if it is a sub-lineage of the one with the highest proportion.

• Samples are being excluded because they have no labelled private mutations.
• Experiment with keeping samples that also have privateNucMutations.unlabeledSubstitutions or privateNucMutations.reversionSubstitutions

https://github.com/aineniamh/snipit

This is the list of implemented positive controls:

• XA | Alpha recombinant (poor lineage accuracy)
• XB | Conflicting designation issue
• XC | Alpha recombinant (poor lineage accuracy)
• XD
• XE
• XG
• XH
• XJ
• XK | No public genomes
• XL
• XM
• XN
• XP
• XQ
• XR
• XS
• XT | No public genomes | Restricted to South Africa
• XU | No public genomes | Restricted to India, Japan, Australia
• XW | ...
• XY | ...
• XZ | ...
• XAA | ...
• XAB | ...
• XAC | ...
• XAD | ...
• XAE | ...
• XAF | ...
• XAG | ...
• XAH | ...
• XAJ | ...
• XAK | ...
• XAL | ...
• XAM | ...
• XAN | ...
• XAP | ...
• XAQ | ...
• XAR | ...
• XAS | ...
• XAT | ...
• XAU | ...
• XAV | ...
• XAW | ...
• XAY | ...
• XAZ | ‼ Priority | Large Lineage

• There are challenges with certain security issues (certificate validation) in downloading the nextclade dataset.
• There is a patch regarding a similar issue in nextstrain: nextstrain/ncov#875

I'd like to give RIPPLES another try, now that I'm more proficient with the matUtils commands.

Steps:

• 1. Create a text file of cluster ids. This will be the first observed sequence for each recombinant lineage.
• 2. Run ripples on those cluster ids.

Applies to rule nextclade_recombinants

• Consider removing steps to extract subtrees

Ideally, make all columns filters.

ex. lineage_usher

This became apparent once I hit the 400+ sample range.

The following rules currently produce empty log files. If checked, log content has been added:

• nextclade
• sc2rf_recombinants
• faToVcf
• usher_metadata
• summary
• plot?

One option is to have alt rules for plot and report.

• plot_historical
• report_historical

This is reported in the column cov-spectrum_query in the linelists. The subs listed must be shared by all samples within the cluster.

Recombinant lineages change extremely rapidly. Include program versions for:

• ncov-recombinant
• nextclade
• sc2rf
• usher

Include dataset versions for:

• nextclade-data
• public-latest

Example:

The reason is to simplify maintenance, as I need to update a large number of profiles each time I want to update the base profobuf.

Fix by specifying plt.rcParams["svg.fonttype"] = "none"

There was improper error handling for how to plot (or not plot) breakpoints when none were identified. This has been fixed in scripts/plot_breakpoints.py as validated with the controls-negative dataset.

gisaid_epi_isl shouldn't be a default metadata column, I need to remove that from the extra_cols param.

Something has changed in the implementation of UShER subtree extraction because an initial upgrade from v0.5.3 to v0.5.6 crashed the pipeline. But v0.5.6 has some interesting new options for subtrees, so I'll look into it!

• I'd like the option to detect duplicates strains based on a matching column (ex. genbank_accession or gisaid_epi_isl).
• And then label or remove from the subtree JSON.

There is too much content in a changelog to fit in the small text box.

XAK has been flagged as a VUM by the ECDC:

https://www.ecdc.europa.eu/en/covid-19/variants-concern

I think it looks cleaner

The tutorial strains have high levels of ambiguity, maybe I should replace these with different sequences? But on the other hand, these are good examples of how the pipeline handles this ambiguity.

• XM-example-2: Ns around breakpoints
• proposed467-example-2: Ns around breakpoints
• miscBA1BA2Post17k-example-1: Ns around breakpoints and IUPAC ambiguity
• 19955,20055 are common sites to be ambiguous.

Datasets:

• BA.2* and BA.5.1* | proposed771
• BA.5.2.1 and (BA.4* or BA.2*) | proposed820

This help in interpreting lineages that have been split into multiple clusters (ex. XM)

This help in interpreting lineages with the same breakpoint and parents (but different subs). For example, XQ and XR differ by one sub (around 17500).

The script create_profile.sh will error out in some cases when there is no empty line between fasta sequences.

We can parse out the gisaid accessions (EPI_ISL_*) from the strain names into a new column.