Automatic revision of MAGs to remove likely chimeric contigs

• Introduction
• Quick start
• Why this pipeline matters
• References

Introduction

This is basic pipeline to automatically identify and remove contigs from metagenome-assembled genomes (MAGs) that have a high chance of being chimeric and do not actually belong to the MAG. It is written in Snakemake and makes use of containers (Singularity and Docker) and conda environments to ensure its reproducibility. It has been previously been used by Klapper et al. (2023) and has received some further updates since then.

The pipeline implements three different steps:

1. Taxonomic assignment of all contigs of a MAG using MMseqs2’s taxonomy workflow (Mirdita et al. 2021) against the GTDB and identifying of the most common lineage at the rank family or higher
2. Taxonomic assignment of every gene of a contig that could only be assigned to the taxonomic rank kingdom or phylum using MMseqs2’s taxonomy workflow (Mirdita et al. 2021) against the GTDB to discriminate contigs carrying preserved genes from contigs carrying genes from multiple lineages
3. Evaluation of the average depth of all contigs

Afterwards, the information of these three steps are combined and contigs are removed because they are likely chimeric when:

1. A contig was assigned to a lineage at the taxonomic level class or lower that does not overlap with the most common lineage of the MAG. E.g. the main lineage was d_Bacteria;p_Firmicutes A;c_Clostridia;o_Lachnospirales;f_Lachnospiraceae but a contig was assigned to d_Bacteria;p_Firmicutes_A;c_Clostridia;o_Peptostreptococcales;f_Filifactoraceae;g_Peptoanaerobacter;s_Peptoanaerobacter_stomatis.
2. A contig could only be assigned to the taxomic rank kingdom and phylum and the per-gene analysis revealed that genes on this contig can be assigned to a lineage other than the main lineage.
3. A contig could only be assigned to the taxomic rank kingdom and phylum and its coverage was deviating from the average coverage of all contigs assigned to the main lineage by more than two standard deviations.

Finally, the pipeline runs multiple quality evaluation steps on the refined MAGs:

1. Functional annotation using Bakta (Schwengers et al. 2022)
2. Evaluation of the quality of the MAG using

• checkM (Parks et al. 2015)
• checkM2 (Chklovski et al. 2023)
• GUNC (Orakov et al. 2021)
• the presence of alternative alleles with a minimal allele frequency of 20% within coding sequences that would lead to a non-synonymous mutation (Pasolli et al. 2019)

3. Taxonomic assignment using

• GTDBTK (Chaumeil et al. 2020)
• PhyloPhlAn (Asnicar et al. 2020)

A detailed description of the methodology and some results can be found in the Supplementary Material Section 3 “Reference-free binning of the de novo assembled contigs” of Klapper et al. (2023).

Quick start

To be able to run the pipeline, Snakemake with a minimal version of 7.0 is necessary. The easiest way to install the dependencies of this program and to have reproducible results is to create a new conda environment using the environment file provided with it.

wget https://raw.githubusercontent.com/alexhbnr/automatic_MAG_refinement/main/environment.yml
conda env create -f environment.yml

After activating the environment using

conda activate automatic_MAG_refinement

the necessary Python environment has been created.

Next, the pipeline itself needs to be downloaded. This can be easily down by cloning this repository to your computer via git

git clone https://github.com/alexhbnr/automatic_MAG_refinement.git

or by downloading the zip file and extracting it:

wget -O automatic_MAG_refinement.zip https://github.com/alexhbnr/automatic_MAG_refinement/archive/refs/heads/main.zip
unzip automatic_MAG_refinement.zip

Afterwards, we can change into pipeline directory:

cd automatic_MAG_refinement

Finally, the configuration file and sample table have to be provided. Templates for these can be found in config/config.yaml for the configuration file and in test/samples.tsv for the sample table.

To run a test case using the assembly results of sample FUM003, a Neanderthal dental calculus sample, that was first published by Fellows Yates et al. (2021) and later de novo assembled in Klapper et al. (2023):

wget -O test/FUM003-megahit.fasta.gz https://share.eva.mpg.de/index.php/s/nQ7Df5Z4T2EFQrA/download/FUM003-megahit.fasta.gz
wget -O test/FUM003.sorted.dedup.bam https://share.eva.mpg.de/index.php/s/fbQNLGs74AGit6J/download/FUM003.sorted.dedup.bam
wget -O test/FUM003.sorted.dedup.bam.bai https://share.eva.mpg.de/index.php/s/B2nMAWLZCw5kK6y/download/FUM003.sorted.dedup.bam.bai
wget -O test/metawrap_50_10_bins.stats https://share.eva.mpg.de/index.php/s/dkqeA2fNMksdqsk/download/metawrap_50_10_bins.stats

To start the pipeline, we run

snakemake --use-conda --conda-prefix conda \
          --use-singularity --singularity-prefix singularity -j 8

This will automatically evaluate the entries in the configuration file config/config.yaml and use the sample FUM003 as input. The temporary files are written into the folder tmp and the results in the folder results.

By activating the options --use-conda and --use-singularity, snakemake will download and install the programs necessary to run the pipeline via conda or pull the container images via singularity. This step will only happen once, at the first time or when you change the folder for storing these conda environments using --conda-prefix or --singularity-prefix, respectively.

Why this pipeline matters

The de novo assembly of short-read metagenomic sequencing data has become the de-facto standard when studying the microbiome of complex samples. In contrast to genomes derived from cultured microbial isolates, the task of inferring which contig belongs to which genome is not trivial. Therefore, multiple approaches have been developed to use the sequence composition and the sequencing depth along the contigs to cluster similar contigs together (Alneberg et al. 2013; Wu, Simmons, and Singer 2016; Kang et al. 2019). While using the sequence composition and the sequencing depth allows to cluster all contigs in a short amount of time, this information is not always sufficient to correctly separate all contigs in the correct genome bins. To improve these clusters, workflows have been developed that infer the presence of lineage-specific, single-copy genes along the contigs and use these to revise the assignment of contigs into clusters (Sieber et al. 2018; Uritskiy, DiRuggiero, and Taylor 2018). However, the performance of the different combination of tools is dependent on the underlying set of sequencing data (Yue et al. 2020).

There were attempts to standardise the evaluation of the minimum information that is to be provided for new metagenome-assembled genomes (MAGs) (Bowers et al. 2017). CheckM (Parks et al. 2015) has established itself as the de-facto standard for estimating the completeness and the contamination of a MAG based on marker genes. However, Orakov et al. (2021) could show that checkM’s approach only identifies the surplus of contigs from other taxa but is not able to identify chimeric contigs and therefore overestimates the quality. Instead, the authors developed their own tool, GUNC, that can evaluate whether a MAG is likely chimeric.

How to proceed with MAGs that were either assigned to high or medium quality using checkM but returned a higher than expected GUNC score is yet unclear. In the presence of a large number of additional genomes from the same habitat some researchers tended to discard these MAGs as chimeric (e.g. Saheb Kashaf et al. (2021)). However, for samples that are limited in quantity and underlie strong ethical considerations, such as ancient DNA samples, this is not an adequate solution. Suggestions have been put forward to manually curate the contigs of chimeric MAGs (Chen et al. 2020) in programs such like anvi’o (Eren et al. 2015) and discard the problematic contigs. This manual process ranges from time-consuming to infeasible, when a dataset consists out of many samples with each a large number of MAGs.

Here, we developed an automatic workflow that is heavily influenced by the suggestions by Chen et al. (2020) and automatises many steps that can be manually done in anvi’o. In brief, the pipeline written in Snakemake (Mölder et al. 2021) expects MAGs refined by MetaWRAP (Uritskiy, DiRuggiero, and Taylor 2018) as input and identifies contigs that are likely chimeric by inferring the majority lineage across all contigs using MMSeqs2 (Steinegger and Söding 2017) against the GTDB reference database (Parks et al. 2020) using the command mmseqs taxonomy and discard contigs that diverge either by average sequencing depth or lineage assignment. For the revised contigs, a standard set of assembly information including an updated estimate for the genome completeness and the contamination is determined and reported.

References

Alneberg, Johannes, Brynjar Smári Bjarnason, Ino de Bruijn, Melanie Schirmer, Joshua Quick, Umer Z Ijaz, Nicholas J Loman, Anders F Andersson, and Christopher Quince. 2013. “CONCOCT: Clustering Contigs on Coverage and Composition.” arXiv Preprint arXiv:1312.4038.

Asnicar, Francesco, Andrew Maltez Thomas, Francesco Beghini, Claudia Mengoni, Serena Manara, Paolo Manghi, Qiyun Zhu, et al. 2020. “Precise Phylogenetic Analysis of Microbial Isolates and Genomes from Metagenomes Using PhyloPhlAn 3.0.” Nature Communications 11 (1): 2500. https://doi.org/10.1038/s41467-020-16366-7.

Bowers, Robert M, Nikos C Kyrpides, Ramunas Stepanauskas, Miranda Harmon-Smith, Devin Doud, TBK Reddy, Frederik Schulz, et al. 2017. “Minimum Information about a Single Amplified Genome (MISAG) and a Metagenome-Assembled Genome (MIMAG) of Bacteria and Archaea.” Nature Biotechnology 35 (8): 725–31.

Chaumeil, Pierre-Alain, Aaron J Mussig, Philip Hugenholtz, and Donovan H Parks. 2020. “GTDB-Tk: A Toolkit to Classify Genomes with the Genome Taxonomy Database.” Bioinformatics 36 (6): 1925–27. https://doi.org/10.1093/bioinformatics/btz848.

Chen, Lin-Xing, Karthik Anantharaman, Alon Shaiber, A Murat Eren, and Jillian F Banfield. 2020. “Accurate and Complete Genomes from Metagenomes.” Genome Research 30 (3): 315–33.

Chklovski, Alex, Donovan H. Parks, Ben J. Woodcroft, and Gene W. Tyson. 2023. “CheckM2: A Rapid, Scalable and Accurate Tool for Assessing Microbial Genome Quality Using Machine Learning.” Nature Methods 20 (8): 1203–12. https://doi.org/10.1038/s41592-023-01940-w.

Eren, A Murat, Özcan C Esen, Christopher Quince, Joseph H Vineis, Hilary G Morrison, Mitchell L Sogin, and Tom O Delmont. 2015. “Anvi’o: An Advanced Analysis and Visualization Platform for ‘Omics Data.” PeerJ 3: e1319.

Fellows Yates, James A., Irina M. Velsko, Franziska Aron, Cosimo Posth, Courtney A. Hofman, Rita M. Austin, Cody E. Parker, et al. 2021. “The Evolution and Changing Ecology of the African Hominid Oral Microbiome.” Proceedings of the National Academy of Sciences of the United States of America 118 (20): e2021655118. https://doi.org/10.1073/pnas.2021655118.

Kang, Dongwan D, Feng Li, Edward Kirton, Ashleigh Thomas, Rob Egan, Hong An, and Zhong Wang. 2019. “MetaBAT 2: An Adaptive Binning Algorithm for Robust and Efficient Genome Reconstruction from Metagenome Assemblies.” PeerJ 7: e7359.

Klapper, Martin, Alexander Hübner, Anan Ibrahim, Ina Wasmuth, Maxime Borry, Veit G. Haensch, Shuaibing Zhang, et al. 2023. “Natural Products from Reconstructed Bacterial Genomes of the Middle and Upper Paleolithic.” Science 380 (6645): 619–24. https://doi.org/10.1126/science.adf5300.

Mirdita, M, M Steinegger, F Breitwieser, J Söding, and E Levy Karin. 2021. “Fast and Sensitive Taxonomic Assignment to Metagenomic Contigs.” Bioinformatics 37 (18): 3029–31. https://doi.org/10.1093/bioinformatics/btab184.

Mölder, Felix, Kim Philipp Jablonski, Brice Letcher, Michael B Hall, Christopher H Tomkins-Tinch, Vanessa Sochat, Jan Forster, et al. 2021. “Sustainable Data Analysis with Snakemake.” F1000Research 10.

Orakov, Askarbek, Anthony Fullam, Luis Pedro Coelho, Supriya Khedkar, Damian Szklarczyk, Daniel R Mende, Thomas SB Schmidt, and Peer Bork. 2021. “GUNC: Detection of Chimerism and Contamination in Prokaryotic Genomes.” Genome Biology 22 (1): 1–19.

Parks, Donovan H, Maria Chuvochina, Pierre-Alain Chaumeil, Christian Rinke, Aaron J Mussig, and Philip Hugenholtz. 2020. “A Complete Domain-to-Species Taxonomy for Bacteria and Archaea.” Nature Biotechnology 38 (9): 1079–86.

Parks, Donovan H, Michael Imelfort, Connor T Skennerton, Philip Hugenholtz, and Gene W Tyson. 2015. “CheckM: Assessing the Quality of Microbial Genomes Recovered from Isolates, Single Cells, and Metagenomes.” Genome Research 25 (7): 1043–55.

Pasolli, Edoardo, Francesco Asnicar, Serena Manara, Moreno Zolfo, Nicolai Karcher, Federica Armanini, Francesco Beghini, et al. 2019. “Extensive Unexplored Human Microbiome Diversity Revealed by Over 150,000 Genomes from Metagenomes Spanning Age, Geography, and Lifestyle.” Cell 176 (3): 649–662.e20. https://doi.org/10.1016/j.cell.2019.01.001.

Saheb Kashaf, Sara, Diana M Proctor, Clay Deming, Paul Saary, Martin Hölzer, Monica E Taylor, Heidi H Kong, Julia A Segre, Alexandre Almeida, and Robert D Finn. 2021. “Integrating Cultivation and Metagenomics for a Multi-Kingdom View of Skin Microbiome Diversity and Functions.” Nature Microbiology, 1–11.

Schwengers, Oliver, Lukas Jelonek, Marius Alfred Dieckmann, Sebastian Beyvers, Jochen Blom, and AlexanderYR 2021 Goesmann. 2022. “Bakta: Rapid and Standardized Annotation of Bacterial Genomes via Alignment-Free Sequence Identification.” Microbial Genomics 7 (11): 000685. https://doi.org/10.1099/mgen.0.000685.

Sieber, Christian MK, Alexander J Probst, Allison Sharrar, Brian C Thomas, Matthias Hess, Susannah G Tringe, and Jillian F Banfield. 2018. “Recovery of Genomes from Metagenomes via a Dereplication, Aggregation and Scoring Strategy.” Nature Microbiology 3 (7): 836–43.

Steinegger, Martin, and Johannes Söding. 2017. “MMseqs2 Enables Sensitive Protein Sequence Searching for the Analysis of Massive Data Sets.” Nature Biotechnology 35 (11): 1026–28.

Uritskiy, Gherman V, Jocelyne DiRuggiero, and James Taylor. 2018. “MetaWRAP—a Flexible Pipeline for Genome-Resolved Metagenomic Data Analysis.” Microbiome 6 (1): 1–13.

Wu, Yu-Wei, Blake A Simmons, and Steven W Singer. 2016. “MaxBin 2.0: An Automated Binning Algorithm to Recover Genomes from Multiple Metagenomic Datasets.” Bioinformatics 32 (4): 605–7.

Yue, Yi, Hao Huang, Zhao Qi, Hui-Min Dou, Xin-Yi Liu, Tian-Fei Han, Yue Chen, Xiang-Jun Song, You-Hua Zhang, and Jian Tu. 2020. “Evaluating Metagenomics Tools for Genome Binning with Real Metagenomic Datasets and CAMI Datasets.” BMC Bioinformatics 21 (1): 1–15.