The work describes the annotation of metagenomic assemblies of four cellulolytic microbial communities obtained from various substrates (straw, sawdust, leaf litter).

Student: Grigory Gladkov

Supervisors: Mike Raiko, Lavrentiy Danilov

• Python 3.8:

  • tqdm 4.54
  • fn 0.43
  • bio 0.23
  • biopython 1.78
  • pandas 1.1.4

• MetaBAT2 2.15

• CheckM 1.0.18

• taxonkit 0.6.2

• hhmer 3.3.1

• R 3.5.2:

  • dplyr 1.0.2
  • plyr 1.8.6
  • ggplot2 3.3

Find the differences between microbial communities that metabolising cellulose on straw and leaf litter from ONT assemblies.

Tasks:

• de novo binning (MetaBAT2), compare with reference base binning (Kaiju), compare with 16S data

• find community-specific glycoside hydrolases

Binning was done using Metabat2 with followed command:

metabat2 -i consensus.fasta -o metabat2_bins_3

Quality control was done out using CheckM:

checkm lineage_wf -x fa  -f checkm/N2_MAGs_checkm.tab -t 50 metabat2_bins_3/ out/
checkm qa ./out/lineage.ms ./out -o 2 -f res --tab_table -t 20

For ssu extraction, we use ssu_finder:

checkm ssu_finder consensus.fasta bins/ out/ -x fa -t 20

where out - out folder, metabat2_bins_3 - folder with bins, -t - number of used threads, -x format of fasta file, consensus.fasta - assemblies.

A preliminary attempt was made to binning using coverage information, which was unsuccessful. For alignment per assembly, we used two alignment algorithms:

bwa mem  -t 84 consensus.fasta all.fq | samtools sort -o all.bam
maCMD -p 'Nanopore' -x mc_out/consensus.json -i ../all.fq -t 64 -o mc_out/consensus.sam

To generate a file with sequencing depth, the following command was used:

jgi_summarize_bam_contig_depths --outputDepth true_depth.txt --referenceFasta consensus.fasta  all.bam

with result:

0 bins  formed.

To search for domain families of hydroside hydrolases by hmm profile, the following code was used, which launched a search for each .faa file in the directory:

for filename in *.faa; do
    hmmscan --noali --notextw --acc -E 0.000000000000000001 --cpu 50 -o  hm_${filename}.txt hmm/glyco.hmm $filename
done

glyco.hmm - hmm profile, acc - accuracity trashold, filename - prokka output

Post-processing of the obtained results was carried out in jupyter notebook (parse_hmm - this code is based on the code of Ilya Corvigo). Adding Kaiju results to hhmer results was done using the ultimate pipe for each community:

cat ../c1_ann.tsv | cut -f2 |grep -f /dev/stdin pr_01.gff | awk '{print $1}' | grep -f /dev/stdin ka_01.tsv | cut -f3 | taxonkit --data-dir ~/storage/temp/  -j 50 lineage | taxonkit --data-dir ~/storage/temp/ -j 50 reformat > kid_01.txt

Where - c1_ann.tsv - output from hmmer, pr_01.gff - Prokka output, ka_01.tsv - Kaiju output, taxonkit - add NCBI taxonomy. Visualisation of hmmerscan results have done with com_vis.R script.

• As a result of binning, 10 genomes were obtained from the metagenome.
• In the metagenomes found matches with catalase families responsible for cellulose/hemicellulose decomposition.
• The collected metagenomes and the most represented groups in which the necessary catalases were found do not coincide.
• Differences in taxonomy are more pronounced than differences in the representation of glycoside hydrolase families.