Brian Bushnell
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Brian Bushnell has a total of 79 co-authors. Here are the top co-authors by number of shared publications:
- Juan C. Villada
DOE Joint Genome Institute, Lawrence Berkeley National Laboratory, Berkeley, USA
5 shared publications - Simon Roux
Joint Genome Institute, Lawrence Berkeley National Laboratory, Berkeley, USA
4 shared publications - Antonio Pedro Camargo
Department of Biochemistry, Institute of Chemistry, University of São Paulo, São Paulo, SP 05508-060, Brazil
3 shared publications - Emiley A. Eloe-Fadrosh
DOE Joint Genome Institute, Lawrence Berkeley National Laboratory, Berkeley, USA
3 shared publications - Natalia Ivanova
Department of Energy Joint Genome Institute, Lawrence Berkeley National Laboratory
3 shared publications - Nikos C. Kyrpides
DOE Joint Genome Institute, Lawrence Berkeley National Laboratory , 1 Cyclotron road, Berkeley, CA 94720
3 shared publications
Latest Publications
- Open Access
- 29 October 2026
- Published by Springer Nature in Nature Microbiology
- Vol. 11 (11) , 3253-3267
Planetary-scale metagenomic analysis reveals CRISPR targeting dynamics across ecosystems
- 29 October 2026
- Published by Springer Nature in Nature Microbiology
- Vol. 11 (11) , 3253-3267
Open AccessPlanetary-scale metagenomic analysis reveals CRISPR targeting dynamics across ecosystems
- Open Access
- 5 October 2026
- Published by Elsevier in Cell Genomics
RandomReadsMG: Rapid, realistic metagenome simulation at terabase scale for benchmarking and experimental design
- 5 October 2026
- Published by Elsevier in Cell Genomics
Open AccessRandomReadsMG: Rapid, realistic metagenome simulation at terabase scale for benchmarking and experimental design
- Open Access
- 25 August 2026
- Published by Springer Nature in Communications Biology
- Vol. 9 (1) , 1135
Deployable high-fidelity metagenome binning at scale with QuickBin
- 25 August 2026
- Published by Springer Nature in Communications Biology
- Vol. 9 (1) , 1135
Open AccessDeployable high-fidelity metagenome binning at scale with QuickBin
- Open Access
DynamicDemiLog: A Single Sketch for Ultrafast Similarity, Frequency, and Cardinality Estimation
Probabilistic cardinality estimators (HyperLogLog), similarity sketches (MinHash), and frequency estimators (Count-Min Sketch) are fundamental approximate data structures that each target one primary problem. We present DynamicDemiLog (DDL), a sketch that unifies cardinality estimation, set similarity, containment, element frequency and composition in one tiny data structure built from a single pass over the input stream. Using an inverted index over 200,687 RefSeq sketches (159,567 organisms), DDL performs all-to-all sketch similarity comparison of the full database in 30 seconds (128 threads, indexed) - over 375x faster per query than Mash's brute-force all-to-all comparison of 91,282 sketches, or 31x faster without the index, at double the sketch resolution. DDL extends the LogLog register with a mantissa: each register stores a floating-point-encoded hash value consisting of an integer exponent (the leading-zero count) and a fractional mantissa (the sub-leading-zero bits), rather than the integer leading-zero count alone. This preserves enough hash information for meaningful register-by-register comparison - a property that standard 6-bit registers lack - while improving on LogLog's cardinality estimation machinery, including DynamicLogLog's early exit mask for high-throughput streaming. With a default 10 mantissa bits (16-bit registers, 2,048 buckets, 4 KB), DDL achieves a per-register false-match rate of 0.018% on unrelated random same-size sets (compared to 17.0% for LL6, a basic HyperLogLog implementation), enabling Weighted Kmer Identity (WKID), Average Nucleotide Identity (ANI), containment, and completeness estimation from register comparison alone. A 16-bit per-register observation counter provides element frequency information at trivial additional computation cost, and an additional byte tracks element composition (GC content, for biological data). Furthermore, DDL's high-specificity registers enable an inverted index structure (DDLIndex) that answers similarity queries against a database of N sketches in O(B + M) time, where M is the number of matching index entries, compared to O(NxB) for pairwise comparison.
Open AccessDynamicDemiLog: A Single Sketch for Ultrafast Similarity, Frequency, and Cardinality Estimation
Probabilistic cardinality estimators (HyperLogLog), similarity sketches (MinHash), and frequency estimators (Count-Min Sketch) are fundamental approximate data structures that each target one primary problem. We present DynamicDemiLog (DDL), a sketch that unifies cardinality estimation, set similarity, containment, element frequency and composition in one tiny data structure built from a single pass over the input stream. Using an inverted index over 200,687 RefSeq sketches (159,567 organisms), DDL performs all-to-all sketch similarity comparison of the full database in 30 seconds (128 threads, indexed) - over 375x faster per query than Mash's brute-force all-to-all comparison of 91,282 sketches, or 31x faster without the index, at double the sketch resolution. DDL extends the LogLog register with a mantissa: each register stores a floating-point-encoded hash value consisting of an integer exponent (the leading-zero count) and a fractional mantissa (the sub-leading-zero bits), rather than the integer leading-zero count alone. This preserves enough hash information for meaningful register-by-register comparison - a property that standard 6-bit registers lack - while improving on LogLog's cardinality estimation machinery, including DynamicLogLog's early exit mask for high-throughput streaming. With a default 10 mantissa bits (16-bit registers, 2,048 buckets, 4 KB), DDL achieves a per-register false-match rate of 0.018% on unrelated random same-size sets (compared to 17.0% for LL6, a basic HyperLogLog implementation), enabling Weighted Kmer Identity (WKID), Average Nucleotide Identity (ANI), containment, and completeness estimation from register comparison alone. A 16-bit per-register observation counter provides element frequency information at trivial additional computation cost, and an additional byte tracks element composition (GC content, for biological data). Furthermore, DDL's high-specificity registers enable an inverted index structure (DDLIndex) that answers similarity queries against a database of N sketches in O(B + M) time, where M is the number of matching index entries, compared to O(NxB) for pairwise comparison.
- Open Access
GradeBins: quality control of bins from synthetic and real metagenomes
Genomes reconstructed from metagenomes support studies of microbial diversity and function, but their interpretation depends on how much DNA is recovered and whether foreign sequence is included. Quality can be measured in benchmarks, where contig origins are known, but must be estimated in routine analyses. Here we present GradeBins, which assesses individual bins (contigs assigned to one candidate genome) and summarizes the complete bin set from an assembly. For 6,374 simulated bins, its measurements agreed with AMBER and independent calculations within rounding. In control bins built from 20 kb fragments of reference genomes, GradeBins had lower observed completeness error than CheckM1 and CheckM2 and the highest observed F1 (the harmonic mean of precision and recall) for bin selection. With 2 kb fragments, GradeBins overestimated the completeness of incomplete bins, and its error increased from 3.1 to 17.6 percentage points. Contamination estimates differed among GradeBins, CheckM1, CheckM2 and GUNC, which define contamination differently. In sequenced metagenomes, the choice of estimator changed which bins passed quality cutoffs. GradeBins estimated quality faster than CheckM1, CheckM2 and GUNC on the same hardware. GradeBins thus provides fast quality assessment of prokaryotic bins and bin sets, with completeness estimates that are most accurate for long contigs.
Open AccessGradeBins: quality control of bins from synthetic and real metagenomes
Genomes reconstructed from metagenomes support studies of microbial diversity and function, but their interpretation depends on how much DNA is recovered and whether foreign sequence is included. Quality can be measured in benchmarks, where contig origins are known, but must be estimated in routine analyses. Here we present GradeBins, which assesses individual bins (contigs assigned to one candidate genome) and summarizes the complete bin set from an assembly. For 6,374 simulated bins, its measurements agreed with AMBER and independent calculations within rounding. In control bins built from 20 kb fragments of reference genomes, GradeBins had lower observed completeness error than CheckM1 and CheckM2 and the highest observed F1 (the harmonic mean of precision and recall) for bin selection. With 2 kb fragments, GradeBins overestimated the completeness of incomplete bins, and its error increased from 3.1 to 17.6 percentage points. Contamination estimates differed among GradeBins, CheckM1, CheckM2 and GUNC, which define contamination differently. In sequenced metagenomes, the choice of estimator changed which bins passed quality cutoffs. GradeBins estimated quality faster than CheckM1, CheckM2 and GUNC on the same hardware. GradeBins thus provides fast quality assessment of prokaryotic bins and bin sets, with completeness estimates that are most accurate for long contigs.
- Open Access
Deployable high-fidelity metagenome binning at scale with QuickBin
1. Reconstructing genomes from metagenomic assemblies is foundational to microbiome research, yet metagenome binning remains constrained by a persistent trade-off between genome fidelity and deployable throughput. Many high-accuracy approaches rely on GPU-intensive workflows, marker-gene-informed postprocessing, or resource demands that limit reproducible use at catalog scale. Here we present QuickBin , a CPU-native, marker-free binning algorithm designed to recover near-complete, ultra-low-contamination metagenome-assembled genomes (MAGs) under practical compute constraints. QuickBin couples a GC-coverage spatial index (BinMap) with a hierarchical, early-exit Oracle cascade of similarity tests (scalar composition/coverage filters followed by SIMD-accelerated k-mer comparisons), invoking a compact neural network only for the small fraction of ambiguous candidate merges. Across synthetic communities evaluated by both marker-based estimates and contig-origin ground truth, QuickBin yields high recovery while maximizing the fraction of sequence assigned to uncontaminated bins. In a benchmark of 297 diverse real metagenomes, QuickBin completed all runs and recovered more MAGs meeting ≥95% completeness and ≤1% contamination than widely used CPU binners and compute-intensive alternatives that frequently failed to finish under standardized limits. QuickBin provides a practical path to reproducible, high-fidelity genome-resolved metagenomics at scale for downstream comparative analyses and data products that depend on low-artifact genome reconstructions. The full software package is open-source and available for download at https://bbmap.org , GitHub repo ( https://github.com/bbushnell/BBTools ), or Docker container ( https://hub.docker.com/r/bryce911/bbtools ).
Open AccessDeployable high-fidelity metagenome binning at scale with QuickBin
1. Reconstructing genomes from metagenomic assemblies is foundational to microbiome research, yet metagenome binning remains constrained by a persistent trade-off between genome fidelity and deployable throughput. Many high-accuracy approaches rely on GPU-intensive workflows, marker-gene-informed postprocessing, or resource demands that limit reproducible use at catalog scale. Here we present QuickBin , a CPU-native, marker-free binning algorithm designed to recover near-complete, ultra-low-contamination metagenome-assembled genomes (MAGs) under practical compute constraints. QuickBin couples a GC-coverage spatial index (BinMap) with a hierarchical, early-exit Oracle cascade of similarity tests (scalar composition/coverage filters followed by SIMD-accelerated k-mer comparisons), invoking a compact neural network only for the small fraction of ambiguous candidate merges. Across synthetic communities evaluated by both marker-based estimates and contig-origin ground truth, QuickBin yields high recovery while maximizing the fraction of sequence assigned to uncontaminated bins. In a benchmark of 297 diverse real metagenomes, QuickBin completed all runs and recovered more MAGs meeting ≥95% completeness and ≤1% contamination than widely used CPU binners and compute-intensive alternatives that frequently failed to finish under standardized limits. QuickBin provides a practical path to reproducible, high-fidelity genome-resolved metagenomics at scale for downstream comparative analyses and data products that depend on low-artifact genome reconstructions. The full software package is open-source and available for download at https://bbmap.org , GitHub repo ( https://github.com/bbushnell/BBTools ), or Docker container ( https://hub.docker.com/r/bryce911/bbtools ).
- Open Access
Rapid terabase-scale simulation of realistic metagenomes for experimental design and pathogen detection with RandomReadsMG
1. Accurate simulation of metagenomic sequencing data is determinant for benchmarking new algorithms, guiding experimental design and generation of synthetic data to develop and train models for AI/ML. However, existing simulators often struggle to reproduce the uneven coverage depth patterns seen in real microbial communities, can be difficult to install, or incur long runtimes. Here we present RandomReadsMG , a fast and scalable read simulator designed to generate synthetic metagenomes with realistic, user-defined depth distributions from hundreds of genomes in a single command. The tool supports multiple abundance models and intra-genome depth variability to better mimic real data. Benchmarks show RandomReadsMG runs orders of magnitude faster than comparable software while maintaining constant memory usage for input datasets of unbounded size. We demonstrate its utility by determining pathogen detection thresholds in a complex microbiome, showcasing its value for optimizing experimental design and creating robust training data sets for bioinformatics and AI/ML. RandomReadsMG is open-source software, distributed as part of the BBTools suite. The full software package is available for download at https://sourceforge.net/projects/bbmap/ . For containerized deployment, a Docker image is also available from Docker Hub at https://hub.docker.com/r/bryce911/bbtools .
Open AccessRapid terabase-scale simulation of realistic metagenomes for experimental design and pathogen detection with RandomReadsMG
1. Accurate simulation of metagenomic sequencing data is determinant for benchmarking new algorithms, guiding experimental design and generation of synthetic data to develop and train models for AI/ML. However, existing simulators often struggle to reproduce the uneven coverage depth patterns seen in real microbial communities, can be difficult to install, or incur long runtimes. Here we present RandomReadsMG , a fast and scalable read simulator designed to generate synthetic metagenomes with realistic, user-defined depth distributions from hundreds of genomes in a single command. The tool supports multiple abundance models and intra-genome depth variability to better mimic real data. Benchmarks show RandomReadsMG runs orders of magnitude faster than comparable software while maintaining constant memory usage for input datasets of unbounded size. We demonstrate its utility by determining pathogen detection thresholds in a complex microbiome, showcasing its value for optimizing experimental design and creating robust training data sets for bioinformatics and AI/ML. RandomReadsMG is open-source software, distributed as part of the BBTools suite. The full software package is available for download at https://sourceforge.net/projects/bbmap/ . For containerized deployment, a Docker image is also available from Docker Hub at https://hub.docker.com/r/bryce911/bbtools .
- Open Access
Planetary-scale metagenomic search reveals new patterns of CRISPR targeting
Summary: Interactions between microbes and their mobile genetic elements (MGEs), including viruses and plasmids, are critical drivers of microbiome structures and processes. CRISPR-Cas systems are important regulators of these host-MGE interactions, but a global understanding of CRISPR loci diversity, activity, and roles across Earth’s biomes is still lacking. Here, we use an optimized computational approach to search short-read data and collect ∼800 million CRISPR spacers across ∼450,000 public metagenomes. From this extensive CRISPR spacer dataset, we identified 1.18 billion hits between 41 million spacers and 2.5 million viruses and plasmids. Focusing on the role of CRISPR as anti-phage defense, we observed an over-representation of prevalent and conserved spacers consistent with a positive selection pressure associated with phage targeting. Identification of sustained CRISPR targeting by microbes not expected to be viable hosts suggested that some phage genomes may frequently enter non-host microbial cells. Taken together, these results outline the extensive diversity of CRISPR loci across microbiomes, and highlight key parameters driving CRISPR targeting patterns likely to influence strain-level diversification and selection processes within microbial populations.
Open AccessPlanetary-scale metagenomic search reveals new patterns of CRISPR targeting
Summary: Interactions between microbes and their mobile genetic elements (MGEs), including viruses and plasmids, are critical drivers of microbiome structures and processes. CRISPR-Cas systems are important regulators of these host-MGE interactions, but a global understanding of CRISPR loci diversity, activity, and roles across Earth’s biomes is still lacking. Here, we use an optimized computational approach to search short-read data and collect ∼800 million CRISPR spacers across ∼450,000 public metagenomes. From this extensive CRISPR spacer dataset, we identified 1.18 billion hits between 41 million spacers and 2.5 million viruses and plasmids. Focusing on the role of CRISPR as anti-phage defense, we observed an over-representation of prevalent and conserved spacers consistent with a positive selection pressure associated with phage targeting. Identification of sustained CRISPR targeting by microbes not expected to be viable hosts suggested that some phage genomes may frequently enter non-host microbial cells. Taken together, these results outline the extensive diversity of CRISPR loci across microbiomes, and highlight key parameters driving CRISPR targeting patterns likely to influence strain-level diversification and selection processes within microbial populations.
- Open Access
Tool Choice drastically Impacts CRISPR Spacer-Protospacer Detection
1 CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) systems are a fundamental defense mechanism in prokaryotes, where short sequences called spacers are stored in the host genome to recognize and target exogenous genetic elements. Viromics, the study of viral communities in environmental samples, relies heavily on identifying these spacer-target interactions to understand host-virus relationships. However, the choice of sequence search tool to identify putative spacer targets is often overlooked, leading to an unknown impact of downstream inferences in virus-host analysis. Here, we utilize simulated and real datasets to compare popular sequence alignment and search tools, revealing critical differences in their ability to detect multiple matches and handle varying degrees of sequence identity between spacers and potential targets. Finally, we provide general guidelines that may inform future research regarding matching, which is a common practice in studying the complex nature of host-MGE interactions.
Open AccessTool Choice drastically Impacts CRISPR Spacer-Protospacer Detection
1 CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) systems are a fundamental defense mechanism in prokaryotes, where short sequences called spacers are stored in the host genome to recognize and target exogenous genetic elements. Viromics, the study of viral communities in environmental samples, relies heavily on identifying these spacer-target interactions to understand host-virus relationships. However, the choice of sequence search tool to identify putative spacer targets is often overlooked, leading to an unknown impact of downstream inferences in virus-host analysis. Here, we utilize simulated and real datasets to compare popular sequence alignment and search tools, revealing critical differences in their ability to detect multiple matches and handle varying degrees of sequence identity between spacers and potential targets. Finally, we provide general guidelines that may inform future research regarding matching, which is a common practice in studying the complex nature of host-MGE interactions.
- Open Access
The DOE JGI Metagenome Workflow
The DOE JGI Metagenome Workflow performs metagenome data processing, including assembly, structural, functional, and taxonomic annotation, and binning of metagenomic datasets that are subsequently included into the Integrated Microbial Genomes and Microbiomes (IMG/M) comparative analysis system (I. Chen, K. Chu, K. Palaniappan, M. Pillay, A. Ratner, J. Huang, M. Huntemann, N. Varghese, J. White, R. Seshadri, et al, Nucleic Acids Rsearch, 2019) and provided for download via the Joint Genome Institute (JGI) Data Portal ( https://genome.jgi.doe.gov/portal/ ). This workflow scales to run on thousands of metagenome samples per year, which can vary by the complexity of microbial communities and sequencing depth. Here we describe the different tools, databases, and parameters used at different steps of the workflow, to help with interpretation of metagenome data available in IMG and to enable researchers to apply this workflow to their own data. We use 20 publicly available sediment metagenomes to illustrate the computing requirements for the different steps and highlight the typical results of data processing. The workflow modules for read filtering and metagenome assembly are available as a Workflow Description Language (WDL) file ( https://code.jgi.doe.gov/BFoster/jgi_meta_wdl.git ). The workflow modules for annotation and binning are provided as a service to the user community at https://img.jgi.doe.gov/submit and require filling out the project and associated metadata descriptions in Genomes OnLine Database (GOLD) (S. Mukherjee, D. Stamatis, J. Bertsch, G. Ovchinnikova, H. Katta, A. Mojica, I Chen, and N. Kyrpides, and T. Reddy, Nucleic Acids Research, 2018). IMPORTANCE: The DOE JGI Metagenome Workflow is designed for processing metagenomic datasets starting from Illumina fastq files. It performs data pre-processing, error correction, assembly, structural and functional annotation, and binning. The results of processing are provided in several standard formats, such as fasta and gff and can be used for subsequent integration into the Integrated Microbial Genome (IMG) system where they can be compared to a comprehensive set of publicly available metagenomes. As of 7/30/2020 7,155 JGI metagenomes have been processed by the JGI Metagenome Workflow.
Open AccessThe DOE JGI Metagenome Workflow
The DOE JGI Metagenome Workflow performs metagenome data processing, including assembly, structural, functional, and taxonomic annotation, and binning of metagenomic datasets that are subsequently included into the Integrated Microbial Genomes and Microbiomes (IMG/M) comparative analysis system (I. Chen, K. Chu, K. Palaniappan, M. Pillay, A. Ratner, J. Huang, M. Huntemann, N. Varghese, J. White, R. Seshadri, et al, Nucleic Acids Rsearch, 2019) and provided for download via the Joint Genome Institute (JGI) Data Portal ( https://genome.jgi.doe.gov/portal/ ). This workflow scales to run on thousands of metagenome samples per year, which can vary by the complexity of microbial communities and sequencing depth. Here we describe the different tools, databases, and parameters used at different steps of the workflow, to help with interpretation of metagenome data available in IMG and to enable researchers to apply this workflow to their own data. We use 20 publicly available sediment metagenomes to illustrate the computing requirements for the different steps and highlight the typical results of data processing. The workflow modules for read filtering and metagenome assembly are available as a Workflow Description Language (WDL) file ( https://code.jgi.doe.gov/BFoster/jgi_meta_wdl.git ). The workflow modules for annotation and binning are provided as a service to the user community at https://img.jgi.doe.gov/submit and require filling out the project and associated metadata descriptions in Genomes OnLine Database (GOLD) (S. Mukherjee, D. Stamatis, J. Bertsch, G. Ovchinnikova, H. Katta, A. Mojica, I Chen, and N. Kyrpides, and T. Reddy, Nucleic Acids Research, 2018). IMPORTANCE: The DOE JGI Metagenome Workflow is designed for processing metagenomic datasets starting from Illumina fastq files. It performs data pre-processing, error correction, assembly, structural and functional annotation, and binning. The results of processing are provided in several standard formats, such as fasta and gff and can be used for subsequent integration into the Integrated Microbial Genome (IMG) system where they can be compared to a comprehensive set of publicly available metagenomes. As of 7/30/2020 7,155 JGI metagenomes have been processed by the JGI Metagenome Workflow.