Paralogs.md

Paralogs assessement

Author: Léo-Paul Dagallier
Last update: 2023-10-10

---

Paralogs extraction

HybPiper can extract all the copies assembled for each genes when multiple copies per gene per sample where assembled.

hybpiper paralog_retriever namelist.txt -t_dna targetfile.fasta --heatmap_filetype pdf --heatmap_dpi 300

It will create 2 output directories: paralogs_all and paralogs_no_chimera. It will also output a heatmap presenting the number of copy recovered per sample and per locus. It will also output a paralog_report.tsv whit the number of assembled copies per locus and per sample, as well as a heatmap.

See the full details on this command and heatmap examples here.

Paralogs identification

Run a quick phylogenetic reconstruction of the loci to inspect for the multi-copies loci:

cd paralogs_all
cat locus1_paralogs_all.fasta | mafft --auto | FastTree -nt -gtr > locus1_paralogs_all.tre

To loop it over all the loci present in the working directory, we first need to get the list of loci. We’ll extract this list from the file paralogs_above_threshold_report.txt created by the paralog_retriever.

cat paralogs_above_threshold_report.txt | sed '1,/The gene names are:/d' > paralogs_all/loci_with_paralog_warning.txt

Then loop it over all the loci and run it in parallel:

for locus in $(cat loci_with_paralog_warning.txt) ;
do
echo "cat "$locus"_paralogs_all.fasta | mafft --auto --quiet - | FastTree -nt -gtr > "$locus"_paralogs_all.tre"
done | parallel -j8

Once the trees are inferred for the paralogs, run the R script plot_hybpiper_paralog_trees.R to visualize the trees in a convenient way. The script creates a single .pdf file with one locus per page.

Rscript plot_hybpiper_paralog_trees.R

For each locus, on the left are 2 plots of the tree with all the samples (including the samples with single copy), and on the right, 2 plots of the tree with only the samples that have more than 1 copy for this locus. Top plots represent the trees unrooted, bottom plots represent the same trees rooted. In orange are the “.main”copy, selected by HybPiper, and in blue are all the other copies (“.0”, “.1”, etc.).

Then inspect all the trees and decide which locus present paralogy. You can store in a text file (see example here) the list of the paralogous loci either as a simple loci list:

locus_to_move_1
locus_to_move_2
...

or as a list of moving commands:

mv *locus_to_move_1* paralogs
mv *locus_to_move_2* paralogs

Using this list, you can discard the paralogous loci form the set of loci you will be inferring a tree. The removal step has to occur either before building the Astral tree (gene trees approach), or before concatenating the loci into a supermatrix (concatenation approach). An alternative to discarding the paralogous loci, is to “resolve” the paralogy (see below).

See also the associated:
👉 💻 script for local use
👉 👩‍💻 script for cluster (SLURM) use.

Paralogy resolution - Orthology inference

Instead of discarding the paralogous loci, you can undertake a paralogy resolution step.

A phylogenetic aware paralogy resolution can be viewed as more interesting than simply removing the paralogous loci, as it allows to use data that would have been discarded otherwise.

The ParaGone pipeline implements phylogenetic aware paralogy resolution as described in Yang and Smith (2014) and Morales-Briones et al. (2021).

Note that to run ParaGone, you need to first run the paralogy extraction step above (hybpiper paralog_retriever).

🚧 More to come soon here with associated template scripts 🚧
👉 Feel free to check this and this out for more details. 👈