Benchmarking genome assembly methods on metagenomic sequencing data

Zhang, Zhenmiao; Yang, Chao; Veldsman, Werner Pieter; Fang, Xiaodong; Zhang, Lu

doi:10.1093/bib/bbad087

Cited by 19 publications

(28 citation statements)

References 67 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We downloaded the metagenomic sequencing datasets from CAMI I with low, medium (two datasets), and high complexity and from a human stool sample ( S 1 ) [25]. The contigs of these datasets were assembled by metaSPAdes with default parameters.…”

Section: Methodsmentioning

confidence: 99%

“…We observed that Deepurify outperformed two state-of-the-art tools MAGpurify and MDMcleaner in simulated, CAMI I challenge (high, medium1, medium2, and low) [24] and human gut metagenomic sequencing data [25, 26]. For simulated data, chimeric MAGs were created by mixing the sequences from two microbial genomes at various taxonomic ranks, with contamination rates from 5% to 20%.…”

Section: Introductionmentioning

confidence: 99%

“…Additionally, Deepurify demonstrated outstanding generalization capabilities, where it achieved excellent accuracy in identifying contaminated contigs even if their source genomes were absent from the training set (Figure 4, Supplementary Table 4). For CAMI I and a human gut metagenomic sequencing dataset, S 1 [25], we applied Deepurify to the results of four mainstream contig binning tools (VAMB [8], CONCOCT [9], MetaBAT2 [11], and MaxBin [10]), and the results showed that it could substantially improve MAG quality, surpassing both the MAGpurify and MDMcleaner for all binning tools. Next, we applied Deepurify to a large metagenomic sequencing dataset derived from a diarrhea-predominant Irritable Bowel Syndrome (IBS-D) cohort, including 290 patients and 89 healthy controls [26].…”

Section: Introductionmentioning

confidence: 99%

“…Deepurify improves the qualities of MAGs from different contig binning toolsWe applied MAGpurify, MDMcleaner, and Deepurify to the MAGs generated by MaxBin, MetaBAT2, VAMB, and CONCOCT to examine if they could increase the number of medium-and high-quality MAGs. The contigs of CAMI I and human gut metagenomic sequencing (S1) were downloaded from our previous study[25] (Methods). We used two criteria to evaluate the performance of MAG decontamination: 1. the increased number of medium-(IN M mq ) and high-quality MAGs (IN M hq ); 2. the improved quality score (IQS ), which measures the overall MAG quality improvement (Methods).Deepurify consistently outperformed the other two tools in nearly all datasets and contig binning methods (Figure5, Supplementary Table8).…”

mentioning

confidence: 99%

See 3 more Smart Citations

Deepurify: a multi-modal deep language model to remove contamination from metagenome-assembled genomes

Zou,

Wang,

Ding

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Metagenome-assembled genomes (MAGs) offer valuable insights into the exploration of microbial dark matter using metagenomic sequencing data. However, there is a growing concern that contamination in MAGs may significantly impact the downstream analysis results. Existing MAG decontamination methods heavily rely on marker genes but do not fully leverage genomic sequences. To address the limitations, we have introduced a novel decontamination approach named Deepurify, which utilizes a multi-modal deep language model employing contrastive learning to learn taxonomic similarities of genomic sequences. Deepurify utilizes inferred taxonomic lineages to guide the allocation of contigs into a MAG-separated tree and employs a tree traversal strategy for maximizing the total number of medium- and high-quality MAGs. Extensive experiments were conducted on two simulated datasets, CAMI I, and human gut metagenomic sequencing data. These results demonstrate that Deepurify significantly outperforms other decontamination methods.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Deepurify: a multi-modal deep language model to remove contamination from metagenome-assembled genomes

Zou,

Wang,

Ding

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…One main strategy is to assemble short reads and then bridge gaps with long reads, leading to assemblies that are accurate but less contiguous 1,5,8,11,14,18,19 .…”

mentioning

confidence: 99%

Simple, reference-independent analyses help optimize hybrid assembly of microbial community metagenomes

Smith,

van Alen,

van Kessel

et al. 2023

Preprint

View full text Add to dashboard Cite

Hybrid metagenomic assembly, leveraging both long- and short-read sequencing technologies, of microbial communities is becoming an increasingly accessible approach yet offers challenges that existing benchmarks and pipelines do not comprehensively handle well. One specific methodological knowledge gap is selecting the ideal number of iterations of long read correction and/or short read polishing, and in particular which the features of an assembly to examine to empirically determine them. To account for biological and technological variation, in this study the microbial communities of two laboratory-scale bioreactors were sequenced with both short and long read platforms and assembled with two different, commonly used software packages. Following assembly, long read error correction was iterated ten times, with subsequent ten rounds of short read polishing after common (0 and 2) and extensive (5 and 10) correction iterations. Several assembly characteristics were tracked for each assembly and iteration of correction and polishing: gene fragmentation, short read recruitment, automated binning yields, as well as observed beta-diversity. While long read correction was necessary, the greatest improvements occurred within the first few iterations of short read polishing, and extensive correction and polishing did not tangibly improve assembly quality. Furthermore, essentially all tracked characteristics followed the same patterns; simpler statistics can serve as decent proxies for more complex analyses to save computational resources assessing assembly quality. Hybrid sequencing approaches will likely remain relevant due to the low costs of short read sequencing, therefore it is imperative users are equipped to estimate assembly quality prior to downstream gene- and genome-centric analyses.

show abstract