MetaBAT 2: an adaptive binning algorithm for robust and efficient genome reconstruction from metagenome assemblies

Abstract

We previously reported on MetaBAT, an automated metagenome binning software tool to reconstruct single genomes from microbial communities for subsequent analyses of uncultivated microbial species. MetaBAT has become one of the most popular binning tools largely due to its computational efficiency and ease of use, especially in binning experiments with a large number of samples and a large assembly. MetaBAT requires users to choose parameters to fine-tune its sensitivity and specificity. If those parameters are not chosen properly, binning accuracy can suffer, especially on assemblies of poor quality. Here, we developed MetaBAT 2 to overcome this problem. MetaBAT 2 uses a new adaptive binning algorithm to eliminate manual parameter tuning. We also performed extensive software engineering optimization to increase both computational and memory efficiency. Comparing MetaBAT 2 to alternative software tools on over 100 real world metagenome assemblies shows superior accuracy and computing speed. Binning a typical metagenome assembly takes only a few minutes on a single commodity workstation. We therefore recommend the community adopts MetaBAT 2 for their metagenome binning experiments. MetaBAT 2 is open source software and available at https://bitbucket.org/berkeleylab/metabat.

Keywords: Clustering; Metagenome binning; Metagenomics.

Figure 1. Benchmark of several popular binning tools on CAMI challenge datasets.

The number of identified genomes are shown at two different precision levels, ≥95% (A, C and D) or ≥90% (B, D and F). The number of identified genomes recovered with a completeness (recall) level 90%, 80%, 70%, 60%, or 50% are represented by different shades of gray, with 90% being the darkest. Benchmarking results using the high complex dataset (A and B), medium complex dataset (C and D), and low complex dataset (E and F) are shown. All the tools (MyCC, CONCOCT, COCACOLA, BinSanity, MaxBin 2 and MetaBAT 2) were run using their default parameters. Completeness and precision were calculated with the ground truth of each dataset.

Figure 2. Comparing binning performance of MetaBAT 2 with alternative binning tools on real world metagenomes.

A total of 120 metagenome assemblies were obtained from IMG (IMG100, see Methods). (A) Total number of bins from each of the dataset formed by each binner. (B) Using a 5% contamination cutoff, number of genome bins that have at least 95% (c95), 70% (c70) and 50% (c50) genome completeness estimated by CheckM. Experiments that produced no bins were omitted.

Figure 3. A binning performance comparison between the default parameter set of MetaBAT 2 against several common best parameter sets found by the genetic algorithm.

The IMG100 dataset was used for searching for the best parameter set for each sample. For each parameter set (S1–S9, see Table S2), a stacked bar shows the percentages of datasets where its performance is better than (darkest gray at the bottom), the same as (medium gray in the middle), or worse than (light gray at the top) the default parameter set. Overall the default parameter set is consistently selected as the best parameter set for most samples.

Figure 4. Comparing MetaBAT 2 with two sets of MetaBAT 1 binning experiments using real world metagenomes.

IMG100 dataset was used for benchmarking experiment. The top 20 metagenomes ordered by the number of genome bins identified are shown. X-axis represents each metagenome, and Y-axis shows the number of genome bins identified using 5% contamination cutoff. Each bar represents three completeness results of 90%, 70%, and 50% by the order of color density (i.e., darkest color represents 90%). The completeness and contamination were estimated by CheckM. MetaBAT 2 outperforms both modes of MetaBAT 1 in most cases.