Yuxuan Du
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Yuxuan Du has a total of 39 co-authors. Here are the top co-authors by number of shared publications:
- Fengzhu Sun
Department of Quantitative and Computational Biology, University of Southern California, Los Angeles, USA
13 shared publications - Ruishan Liu
Department of Computer Science, University of Southern California, Los Angeles, CA 90089, USA
4 shared publications Shiyuan Wang
Department of Electrical Engineering, University of Texas at San Antonio, San Antonio, TX 78249, USA
4 shared publications- Athan Z. Li
Department of Computer Science, University of Southern California, Los Angeles, CA 90089, USA
3 shared publications - Jed A. Fuhrman
Department of Biological Sciences, University of Southern California
3 shared publications - Wenxuan Zuo
Department of Quantitative and Computational Biology, University of Southern California, Los Angeles, USA
3 shared publications
Latest Publications
- Open Access
A Taxonomy-Informed Sparse DNA Foundation Model for Microbial Genomics
: Microorganisms are crucial to Earth’s ecosystems, with genomes encoding functions important to agriculture, biotechnology and human health. Although genomic language models have advanced DNA representation learning, the extensive diversity of microorganisms and highly imbalanced taxonomic composition of existing pretraining corpora pose challenges for effective microbial genomic sequence modeling. Here we present MicroGlot, a taxonomy-informed microbial DNA foundation model pretrained on 3.70 million sequences comprising 378.3 billion nucleotides from 99,700 species. MicroGlot represents hierarchical relationships among taxa using hyperbolic embeddings and integrates this knowledge into a sparse mixture-of-experts architecture. Probing frozen embeddings across layers showed that MicroGlot captures both microbial phenotypic traits and taxonomic identity. Compared with a taxonomy-ablated variant pretrained under the same scheme, incorporating taxonomic knowledge consistently improved representation quality across model depth. Combining sparse computation with efficient architectural and training techniques, MicroGlot achieved competitive probing and fine-tuning performance with low computational requirements. Species-level expert-routing fingerprints aligned more closely with taxonomic groups than tetranucleotide-frequency profiles, indicating that expert usage recapitulates taxonomic structure. These results show that combining sparse architecture with taxonomic knowledge enables efficient and biologically informative genomic language modeling across diverse microbial taxa. MicroGlot is publicly available at https://huggingface.co/athanzli/MicroGlot .
Open AccessA Taxonomy-Informed Sparse DNA Foundation Model for Microbial Genomics
: Microorganisms are crucial to Earth’s ecosystems, with genomes encoding functions important to agriculture, biotechnology and human health. Although genomic language models have advanced DNA representation learning, the extensive diversity of microorganisms and highly imbalanced taxonomic composition of existing pretraining corpora pose challenges for effective microbial genomic sequence modeling. Here we present MicroGlot, a taxonomy-informed microbial DNA foundation model pretrained on 3.70 million sequences comprising 378.3 billion nucleotides from 99,700 species. MicroGlot represents hierarchical relationships among taxa using hyperbolic embeddings and integrates this knowledge into a sparse mixture-of-experts architecture. Probing frozen embeddings across layers showed that MicroGlot captures both microbial phenotypic traits and taxonomic identity. Compared with a taxonomy-ablated variant pretrained under the same scheme, incorporating taxonomic knowledge consistently improved representation quality across model depth. Combining sparse computation with efficient architectural and training techniques, MicroGlot achieved competitive probing and fine-tuning performance with low computational requirements. Species-level expert-routing fingerprints aligned more closely with taxonomic groups than tetranucleotide-frequency profiles, indicating that expert usage recapitulates taxonomic structure. These results show that combining sparse architecture with taxonomic knowledge enables efficient and biologically informative genomic language modeling across diverse microbial taxa. MicroGlot is publicly available at https://huggingface.co/athanzli/MicroGlot .
- Open Access
- 1 July 2026
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 42 (Supplement)
VirBinn improves viral genome binning from metagenomic Hi-C through graph diffusion
Motivation: Metagenomic Hi-C provides in situ proximity signals that can improve genome binning and enable virus–host-association analysis. However, viral genome recovery remains difficult because virus–virus Hi-C contact matrices are extremely sparse. Viral genomes are small, often low-abundance, and frequently assemble into short contigs, leaving many true within-genome links unobserved and causing viral bins to fragment. Results: We present VirBinn, a graph-diffusion framework for viral binning from metagenomic Hi-C. VirBinn enhances virus–virus connectivity through two complementary mechanisms: random-walk-with-restart enhancement on the sparse virus–virus contact graph and host-guided diffusion that propagates viral seeds through the host network to infer indirect virus–virus associations. The enhanced views are integrated and clustered using Leiden community detection to produce viral metagenome-assembled genomes (vMAGs). On dataset-specific simulation benchmarks with ground truth, VirBinn consistently recovers more high-quality vMAGs than Hi-C-based and shotgun-based baselines and substantially increases the number of near-complete genomes. On four real metagenomic Hi-C datasets spanning human gut, pig gut, sheep gut (long-read assembly), and wastewater, VirBinn yields more high-completeness vMAGs under CheckV and produces bins with strong within-cluster contact support. Finally, host linkage analysis using reconstructed host MAGs reveals habitat-specific host-association patterns and plausible host taxonomic profiles. Availability and implementation: VirBinn is available at https://github.com/dyxstat/VirBinn. The scripts to reproduce the results and figures in this article are available at https://github.com/dyxstat/Reproduce_VirBinn.
- 1 July 2026
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 42 (Supplement)
Open AccessVirBinn improves viral genome binning from metagenomic Hi-C through graph diffusion
Motivation: Metagenomic Hi-C provides in situ proximity signals that can improve genome binning and enable virus–host-association analysis. However, viral genome recovery remains difficult because virus–virus Hi-C contact matrices are extremely sparse. Viral genomes are small, often low-abundance, and frequently assemble into short contigs, leaving many true within-genome links unobserved and causing viral bins to fragment. Results: We present VirBinn, a graph-diffusion framework for viral binning from metagenomic Hi-C. VirBinn enhances virus–virus connectivity through two complementary mechanisms: random-walk-with-restart enhancement on the sparse virus–virus contact graph and host-guided diffusion that propagates viral seeds through the host network to infer indirect virus–virus associations. The enhanced views are integrated and clustered using Leiden community detection to produce viral metagenome-assembled genomes (vMAGs). On dataset-specific simulation benchmarks with ground truth, VirBinn consistently recovers more high-quality vMAGs than Hi-C-based and shotgun-based baselines and substantially increases the number of near-complete genomes. On four real metagenomic Hi-C datasets spanning human gut, pig gut, sheep gut (long-read assembly), and wastewater, VirBinn yields more high-completeness vMAGs under CheckV and produces bins with strong within-cluster contact support. Finally, host linkage analysis using reconstructed host MAGs reveals habitat-specific host-association patterns and plausible host taxonomic profiles. Availability and implementation: VirBinn is available at https://github.com/dyxstat/VirBinn. The scripts to reproduce the results and figures in this article are available at https://github.com/dyxstat/Reproduce_VirBinn.
- Open Access
MurineCyto-Det: A High-Resolution Murine BALF Cytology Dataset for Leukocyte Segmentation and Detection
Automated analysis of murine bronchoalveolar lavage fluid (BALF) cytology is important for preclinical respiratory research, yet progress has been limited by the lack of publicly available, well-annotated mouse BALF image datasets. We present MurineCyto-Det, a high-resolution murine BALF cytology dataset comprising 333 image tiles of size 1024×1024 pixels, annotated across five cytological categories with both pixel-level segmentation masks and one-to-one matched bounding boxes. The dataset contains 14,551 annotated cell instances and supports two complementary analysis tasks: morphology-oriented cell segmentation and object-level cell detection. To establish reproducible benchmark baselines, we evaluated representative segmentation and detection models. The results demonstrate the practical utility of MurineCyto-Det while highlighting realistic challenges arising from class imbalance, small object size, irregular cell morphology, and ambiguous debris-like structures. MurineCyto-Det provides a standardized resource for developing, evaluating, and comparing automated methods for murine BALF cytology analysis. The dataset is publicly available at https://doi.org/10.5281/zenodo.17608677 .
Open AccessMurineCyto-Det: A High-Resolution Murine BALF Cytology Dataset for Leukocyte Segmentation and Detection
Automated analysis of murine bronchoalveolar lavage fluid (BALF) cytology is important for preclinical respiratory research, yet progress has been limited by the lack of publicly available, well-annotated mouse BALF image datasets. We present MurineCyto-Det, a high-resolution murine BALF cytology dataset comprising 333 image tiles of size 1024×1024 pixels, annotated across five cytological categories with both pixel-level segmentation masks and one-to-one matched bounding boxes. The dataset contains 14,551 annotated cell instances and supports two complementary analysis tasks: morphology-oriented cell segmentation and object-level cell detection. To establish reproducible benchmark baselines, we evaluated representative segmentation and detection models. The results demonstrate the practical utility of MurineCyto-Det while highlighting realistic challenges arising from class imbalance, small object size, irregular cell morphology, and ambiguous debris-like structures. MurineCyto-Det provides a standardized resource for developing, evaluating, and comparing automated methods for murine BALF cytology analysis. The dataset is publicly available at https://doi.org/10.5281/zenodo.17608677 .
- Open Access
- 4 May 2026
- Published by Oxford University Press (OUP) in Briefings in Bioinformatics
- Vol. 27 (3)
Benchmarking computational methods for multi-omics biomarker discovery in cancer
Multi-omics profiling characterizes cancer biology and supports biomarker discovery for prognosis and therapy selection. Although numerous computational multi-omics biomarker identification methods have been proposed, their ability to identify clinically relevant biomarkers has not been systematically evaluated, leaving it unclear whether the resulting biomarker nominations are reliable for downstream validation. Here, we systematically benchmark 20 representative statistical, machine learning and deep learning methods using curated gold-standard prognostic and therapeutic biomarkers across five real-world datasets. We evaluate performance in terms of both biomarker identification accuracy and stability. Overall, DeePathNet and DeepKEGG achieve the best performance. Across methods, effective biomarker recovery is associated with the integration of biological knowledge, global feature interactions, multivariate feature attribution, and effective regularization. Analysis of omics type contributions reveals method- and modality-specific biases, highlighting the importance of broader omics integration. We further evaluate methods on simulated datasets to probe sensitivity with controlled signal and noise. By aggregating results from top-performing methods, we construct consensus biomarker panels that nominate candidates for potential investigations. Finally, we provide user-friendly interfaces to allow researchers to benchmark new methods against the 20 baselines or apply selected methods for biomarker identification on custom multi-omics datasets. Our benchmark is publicly available at https://github.com/athanzli/CancerMOBI-Bench.
- 4 May 2026
- Published by Oxford University Press (OUP) in Briefings in Bioinformatics
- Vol. 27 (3)
Open AccessBenchmarking computational methods for multi-omics biomarker discovery in cancer
Multi-omics profiling characterizes cancer biology and supports biomarker discovery for prognosis and therapy selection. Although numerous computational multi-omics biomarker identification methods have been proposed, their ability to identify clinically relevant biomarkers has not been systematically evaluated, leaving it unclear whether the resulting biomarker nominations are reliable for downstream validation. Here, we systematically benchmark 20 representative statistical, machine learning and deep learning methods using curated gold-standard prognostic and therapeutic biomarkers across five real-world datasets. We evaluate performance in terms of both biomarker identification accuracy and stability. Overall, DeePathNet and DeepKEGG achieve the best performance. Across methods, effective biomarker recovery is associated with the integration of biological knowledge, global feature interactions, multivariate feature attribution, and effective regularization. Analysis of omics type contributions reveals method- and modality-specific biases, highlighting the importance of broader omics integration. We further evaluate methods on simulated datasets to probe sensitivity with controlled signal and noise. By aggregating results from top-performing methods, we construct consensus biomarker panels that nominate candidates for potential investigations. Finally, we provide user-friendly interfaces to allow researchers to benchmark new methods against the 20 baselines or apply selected methods for biomarker identification on custom multi-omics datasets. Our benchmark is publicly available at https://github.com/athanzli/CancerMOBI-Bench.
- Open Access
- 2 March 2026
- Published by Wiley in Current Protocols
- Vol. 6 (3)
Metagenomic Hi‑C Protocols for Viral Genome Binning, Taxonomic Annotation, and Interaction Network Visualization
Metagenomic Hi‐C (metaHi‐C) links mobile genetic elements to their cellular hosts directly within complex microbial communities. Once shotgun and Hi‐C libraries have been generated, however, the main challenges shift to the bioinformatics required for preprocessing, genome binning, taxonomic annotation, and network‐level interpretation. Here, we present metaHi‐C protocols that span from raw reads to downstream data analyses. Basic Protocol 1 describes quality control of shotgun and Hi‐C reads, metagenomic assembly, Hi‐C read mapping, and viral contig identification from assembled contigs. Basic Protocol 2 details the use of ViralCC to recover viral metagenome‐assembled genomes (vMAGs) and infer virus–host linkages. Support Protocol 1 introduces NormCC and ImputeCC for normalization of raw Hi‐C contacts and host genome binning. Support Protocols 2 and 3 describe taxonomic annotation of host MAGs with GTDB‐Tk and viral bins with Virgo, respectively. Support Protocol 4 shows how to integrate these outputs in MetaHiCNet to generate cross‐taxa and cross‐bin Hi‐C interaction networks. Together, these protocols provide a reproducible workflow for reconstructing viral and host genomes, assigning consistent taxonomies, and visualizing metaHi‐C–derived virus–host interaction structure across diverse microbiomes. © 2026 The Author(s). Current Protocols published by Wiley Periodicals LLC. Basic Protocol 1 : Preprocessing raw metagenomic Hi‐C data Basic Protocol 2 : Viral genome binning and virus–host interaction inference using ViralCC Support Protocol 1 : Host genome binning using ImputeCC Support Protocol 2 : Host MAG taxonomic annotation with GTDB‑Tk Support Protocol 3 : Viral bin taxonomic annotation with Virgo Support Protocol 4 : Visualization of virus–host interaction networks with MetaHiCNet
- 2 March 2026
- Published by Wiley in Current Protocols
- Vol. 6 (3)
Open AccessMetagenomic Hi‑C Protocols for Viral Genome Binning, Taxonomic Annotation, and Interaction Network Visualization
Metagenomic Hi‐C (metaHi‐C) links mobile genetic elements to their cellular hosts directly within complex microbial communities. Once shotgun and Hi‐C libraries have been generated, however, the main challenges shift to the bioinformatics required for preprocessing, genome binning, taxonomic annotation, and network‐level interpretation. Here, we present metaHi‐C protocols that span from raw reads to downstream data analyses. Basic Protocol 1 describes quality control of shotgun and Hi‐C reads, metagenomic assembly, Hi‐C read mapping, and viral contig identification from assembled contigs. Basic Protocol 2 details the use of ViralCC to recover viral metagenome‐assembled genomes (vMAGs) and infer virus–host linkages. Support Protocol 1 introduces NormCC and ImputeCC for normalization of raw Hi‐C contacts and host genome binning. Support Protocols 2 and 3 describe taxonomic annotation of host MAGs with GTDB‐Tk and viral bins with Virgo, respectively. Support Protocol 4 shows how to integrate these outputs in MetaHiCNet to generate cross‐taxa and cross‐bin Hi‐C interaction networks. Together, these protocols provide a reproducible workflow for reconstructing viral and host genomes, assigning consistent taxonomies, and visualizing metaHi‐C–derived virus–host interaction structure across diverse microbiomes. © 2026 The Author(s). Current Protocols published by Wiley Periodicals LLC. Basic Protocol 1 : Preprocessing raw metagenomic Hi‐C data Basic Protocol 2 : Viral genome binning and virus–host interaction inference using ViralCC Support Protocol 1 : Host genome binning using ImputeCC Support Protocol 2 : Host MAG taxonomic annotation with GTDB‑Tk Support Protocol 3 : Viral bin taxonomic annotation with Virgo Support Protocol 4 : Visualization of virus–host interaction networks with MetaHiCNet
- Open Access
- 1 March 2026
- Vol. 6 (3) , e70341
Metagenomic Hi‑C Protocols for Viral Genome Binning, Taxonomic Annotation, and Interaction Network Visualization.
Metagenomic Hi-C (metaHi-C) links mobile genetic elements to their cellular hosts directly within complex microbial communities. Once shotgun and Hi-C libraries have been generated, however, the main challenges shift to the bioinformatics required for preprocessing, genome binning, taxonomic annotation, and network-level interpretation. Here, we present metaHi-C protocols that span from raw reads to downstream data analyses. Basic Protocol 1 describes quality control of shotgun and Hi-C reads, metagenomic assembly, Hi-C read mapping, and viral contig identification from assembled contigs. Basic Protocol 2 details the use of ViralCC to recover viral metagenome-assembled genomes (vMAGs) and infer virus-host linkages. Support Protocol 1 introduces NormCC and ImputeCC for normalization of raw Hi-C contacts and host genome binning. Support Protocols 2 and 3 describe taxonomic annotation of host MAGs with GTDB-Tk and viral bins with Virgo, respectively. Support Protocol 4 shows how to integrate these outputs in MetaHiCNet to generate cross-taxa and cross-bin Hi-C interaction networks. Together, these protocols provide a reproducible workflow for reconstructing viral and host genomes, assigning consistent taxonomies, and visualizing metaHi-C-derived virus-host interaction structure across diverse microbiomes. © 2026 The Author(s). Current Protocols published by Wiley Periodicals LLC. Basic Protocol 1: Preprocessing raw metagenomic Hi-C data Basic Protocol 2: Viral genome binning and virus-host interaction inference using ViralCC Support Protocol 1: Host genome binning using ImputeCC Support Protocol 2: Host MAG taxonomic annotation with GTDB‑Tk Support Protocol 3: Viral bin taxonomic annotation with Virgo Support Protocol 4: Visualization of virus-host interaction networks with MetaHiCNet.
- 1 March 2026
- Vol. 6 (3) , e70341
Open AccessMetagenomic Hi‑C Protocols for Viral Genome Binning, Taxonomic Annotation, and Interaction Network Visualization.
Metagenomic Hi-C (metaHi-C) links mobile genetic elements to their cellular hosts directly within complex microbial communities. Once shotgun and Hi-C libraries have been generated, however, the main challenges shift to the bioinformatics required for preprocessing, genome binning, taxonomic annotation, and network-level interpretation. Here, we present metaHi-C protocols that span from raw reads to downstream data analyses. Basic Protocol 1 describes quality control of shotgun and Hi-C reads, metagenomic assembly, Hi-C read mapping, and viral contig identification from assembled contigs. Basic Protocol 2 details the use of ViralCC to recover viral metagenome-assembled genomes (vMAGs) and infer virus-host linkages. Support Protocol 1 introduces NormCC and ImputeCC for normalization of raw Hi-C contacts and host genome binning. Support Protocols 2 and 3 describe taxonomic annotation of host MAGs with GTDB-Tk and viral bins with Virgo, respectively. Support Protocol 4 shows how to integrate these outputs in MetaHiCNet to generate cross-taxa and cross-bin Hi-C interaction networks. Together, these protocols provide a reproducible workflow for reconstructing viral and host genomes, assigning consistent taxonomies, and visualizing metaHi-C-derived virus-host interaction structure across diverse microbiomes. © 2026 The Author(s). Current Protocols published by Wiley Periodicals LLC. Basic Protocol 1: Preprocessing raw metagenomic Hi-C data Basic Protocol 2: Viral genome binning and virus-host interaction inference using ViralCC Support Protocol 1: Host genome binning using ImputeCC Support Protocol 2: Host MAG taxonomic annotation with GTDB‑Tk Support Protocol 3: Viral bin taxonomic annotation with Virgo Support Protocol 4: Visualization of virus-host interaction networks with MetaHiCNet.
- Open Access
- 30 January 2026
- Published by Springer Nature in Genome Biology
- Vol. 27 (1)
Benchmarking alignment strategies for Hi-C reads in metagenomic Hi-C data
Background: Metagenomics combined with High-throughput Chromosome Conformation Capture (Hi-C) provides a powerful approach to study microbial communities by linking genomic content with spatial interactions. Hi-C complements shotgun sequencing by revealing taxonomic composition, functional interactions, and genomic organization within a single sample. However, aligning Hi-C reads to metagenomic contigs is challenging due to variable insert sizes of Hi-C paired-end reads, multi-species complexity, and gaps in assemblies. Although several benchmark studies have evaluated general alignment tools and Hi-C data alignment, none have specifically focused on metagenomic Hi-C data. Results: We evaluated seven alignment strategies commonly used in Hi-C analyses: BWA MEM -5SP, BWA MEM default, BWA aln default, Bowtie2 default, Bowtie2 –very-sensitive-local, Minimap2 default, and Chromap Hi-C default. We benchmarked these tools on one synthetic dataset and seven real-world environments. Performance was assessed based on the number of inter-contig Hi-C read pairs and their impact on downstream tasks, such as binning quality. Conclusions: We show that BWA MEM -5SP generally outperformed all other tools across most environments in terms of inter-contig read pairs and binning quality, followed by BWA MEM default. Chromap and Minimap2, while less effective in these metrics, demonstrated the highest computational efficiency.
- 30 January 2026
- Published by Springer Nature in Genome Biology
- Vol. 27 (1)
Open AccessBenchmarking alignment strategies for Hi-C reads in metagenomic Hi-C data
Background: Metagenomics combined with High-throughput Chromosome Conformation Capture (Hi-C) provides a powerful approach to study microbial communities by linking genomic content with spatial interactions. Hi-C complements shotgun sequencing by revealing taxonomic composition, functional interactions, and genomic organization within a single sample. However, aligning Hi-C reads to metagenomic contigs is challenging due to variable insert sizes of Hi-C paired-end reads, multi-species complexity, and gaps in assemblies. Although several benchmark studies have evaluated general alignment tools and Hi-C data alignment, none have specifically focused on metagenomic Hi-C data. Results: We evaluated seven alignment strategies commonly used in Hi-C analyses: BWA MEM -5SP, BWA MEM default, BWA aln default, Bowtie2 default, Bowtie2 –very-sensitive-local, Minimap2 default, and Chromap Hi-C default. We benchmarked these tools on one synthetic dataset and seven real-world environments. Performance was assessed based on the number of inter-contig Hi-C read pairs and their impact on downstream tasks, such as binning quality. Conclusions: We show that BWA MEM -5SP generally outperformed all other tools across most environments in terms of inter-contig read pairs and binning quality, followed by BWA MEM default. Chromap and Minimap2, while less effective in these metrics, demonstrated the highest computational efficiency.
- Open Access
Benchmarking computational methods for multi-omics biomarker discovery in cancer
Multi-omics profiling characterizes cancer biology and supports biomarker discovery for prognosis and therapy selection. Although numerous computational multi-omics biomarker identification methods have been proposed, their ability to identify clinically relevant biomarkers has not been systematically evaluated, leaving it unclear whether the resulting biomarker nominations are reliable for downstream validation. Here we systematically benchmark 20 representative statistical, machine learning and deep learning methods using curated gold-standard prognostic and therapeutic biomarkers across five real-world datasets. We evaluate performance in terms of both biomarker identification accuracy and stability. Overall, DeePathNet and Deep-KEGG achieve the best performance. Across methods, effective biomarker recovery is associated with the integration of biological knowledge, global feature interactions, multivariate feature attribution, and effective regularization. Analysis of omics type contributions reveals method- and modality-specific biases, highlighting the importance of broader omics integration. We further evaluate methods on simulated datasets to probe sensitivity with controlled signal and noise. By aggregating results from top-performing methods, we construct consensus biomarker panels that nominate candidates for potential investigations. Finally, we provide user-friendly interfaces to allow researchers to benchmark new methods against the 20 baselines or apply selected methods for biomarker identification on custom multi-omics datasets. Our benchmark is publicly available at https://github.com/athanzli/CancerMOBI-Bench .
Open AccessBenchmarking computational methods for multi-omics biomarker discovery in cancer
Multi-omics profiling characterizes cancer biology and supports biomarker discovery for prognosis and therapy selection. Although numerous computational multi-omics biomarker identification methods have been proposed, their ability to identify clinically relevant biomarkers has not been systematically evaluated, leaving it unclear whether the resulting biomarker nominations are reliable for downstream validation. Here we systematically benchmark 20 representative statistical, machine learning and deep learning methods using curated gold-standard prognostic and therapeutic biomarkers across five real-world datasets. We evaluate performance in terms of both biomarker identification accuracy and stability. Overall, DeePathNet and Deep-KEGG achieve the best performance. Across methods, effective biomarker recovery is associated with the integration of biological knowledge, global feature interactions, multivariate feature attribution, and effective regularization. Analysis of omics type contributions reveals method- and modality-specific biases, highlighting the importance of broader omics integration. We further evaluate methods on simulated datasets to probe sensitivity with controlled signal and noise. By aggregating results from top-performing methods, we construct consensus biomarker panels that nominate candidates for potential investigations. Finally, we provide user-friendly interfaces to allow researchers to benchmark new methods against the 20 baselines or apply selected methods for biomarker identification on custom multi-omics datasets. Our benchmark is publicly available at https://github.com/athanzli/CancerMOBI-Bench .
- Open Access
METAHIT enables comprehensive and flexible genome-resolved microbiome analysis with metagenomic Hi-C
Metagenomic Hi-C (metaHi-C) augments shotgun sequencing with in-cell proximity information, enabling genome-resolved analysis of complex communities. However, computational tools for metaHi-C remain fragmented and rarely offer end-to-end, comprehensive analysis, and existing pipelines use only chimeric Hi-C pairs while ignoring non-chimeric reads, which often constitute a large fraction of Hi-C libraries. Here, we present METAHIT, an accessible and modular workflow that standardizes metaHi-C from raw reads to genome-resolved outputs. The pipeline provides alignment-derived, assumption-light quality checks, integrates three state-of-the-art Hi-C-based binners by consolidating their outputs into a single, non-redundant metagenome-assembled genome (MAG) set, and, for the first time, reuses informative intra-contig read pairs that other Hi-C workflows discard by identifying shotgun-like reads with an EM model on gap distances and using them for per-bin reassembly. METAHIT also supports Hi-C-guided scaffolding, focused visualizations for scaffold structure, MAG annotation, and mobile genetic element (MGE)–host interactions. Across six habitats spanning host-associated and environmental microbiomes, METAHIT increases the recovery of near-complete and high-quality MAGs relative to established Hi-C baselines, while per-bin reassembly lowers contamination and maintains completeness. Applied to a single sheep-gut long-read metaHi-C sample, METAHIT recovered 929 high-quality genomes, representing, to our knowledge, the highest species richness reported from a single sample, and revealed expanded diversity within Erysipelotrichales . In the human gut, METAHIT improved contiguity for an abundant Bacteroides vulgatus MAG via Hi-C–guided scaffolding, identified candidate novel Faecalibacterium lineages, and resolved MGE–host links involving F. prausnitzii and the novel Faecalibacterium MAG. Together, METAHIT delivers standardized, inspection-ready, genome-resolved outputs for comparative, hypothesis-driven microbiome studies across protocols and sequencing modalities.
Open AccessMETAHIT enables comprehensive and flexible genome-resolved microbiome analysis with metagenomic Hi-C
Metagenomic Hi-C (metaHi-C) augments shotgun sequencing with in-cell proximity information, enabling genome-resolved analysis of complex communities. However, computational tools for metaHi-C remain fragmented and rarely offer end-to-end, comprehensive analysis, and existing pipelines use only chimeric Hi-C pairs while ignoring non-chimeric reads, which often constitute a large fraction of Hi-C libraries. Here, we present METAHIT, an accessible and modular workflow that standardizes metaHi-C from raw reads to genome-resolved outputs. The pipeline provides alignment-derived, assumption-light quality checks, integrates three state-of-the-art Hi-C-based binners by consolidating their outputs into a single, non-redundant metagenome-assembled genome (MAG) set, and, for the first time, reuses informative intra-contig read pairs that other Hi-C workflows discard by identifying shotgun-like reads with an EM model on gap distances and using them for per-bin reassembly. METAHIT also supports Hi-C-guided scaffolding, focused visualizations for scaffold structure, MAG annotation, and mobile genetic element (MGE)–host interactions. Across six habitats spanning host-associated and environmental microbiomes, METAHIT increases the recovery of near-complete and high-quality MAGs relative to established Hi-C baselines, while per-bin reassembly lowers contamination and maintains completeness. Applied to a single sheep-gut long-read metaHi-C sample, METAHIT recovered 929 high-quality genomes, representing, to our knowledge, the highest species richness reported from a single sample, and revealed expanded diversity within Erysipelotrichales . In the human gut, METAHIT improved contiguity for an abundant Bacteroides vulgatus MAG via Hi-C–guided scaffolding, identified candidate novel Faecalibacterium lineages, and resolved MGE–host links involving F. prausnitzii and the novel Faecalibacterium MAG. Together, METAHIT delivers standardized, inspection-ready, genome-resolved outputs for comparative, hypothesis-driven microbiome studies across protocols and sequencing modalities.
- Open Access
Benchmarking Alignment Strategies for Hi-C Reads in Metagenomic Hi-C Data
Background: Metagenomics combined with High-throughput Chromosome Conformation Capture (Hi-C) offers a powerful approach to study microbial communities by linking genomic content with spatial interactions. Hi-C enhances shotgun sequencing by revealing taxonomic composition, functional interactions, and genomic organization from a single sample. However, aligning Hi-C reads to metagenomic contigs presents challenges, including the varying insert size of Hi-C paired-end reads, multi-species complexity, and gaps in assemblies. Although many benchmark studies have evaluated general alignment tools and Hi-C data alignment, none have specifically addressed metagenomics Hi-C data. Results: Here, we selected seven alignment strategies that have been used in Hi-C analyses: BWA MEM -5SP, BWA MEM default, BWA aln default, Bowtie2 default, Bowtie2 –very-sensitive-local, Minimap2 default, and Chromap Hi-C default. We benchmarked them on one synthetic and seven real-world environments, and evaluated these tools based on the number of inter-contig Hi-C read pairs and their influence on downstream tasks, such as binning quality. Conclusion: Our findings show that BWA MEM -5SP consistently outperforms other tools across all environments in terms of inter-contig read pairs and binning quality, followed by BWA MEM default. Chromap and Minimap2, while less effective in these metrics, demonstrate the highest computational efficiency.
Open AccessBenchmarking Alignment Strategies for Hi-C Reads in Metagenomic Hi-C Data
Background: Metagenomics combined with High-throughput Chromosome Conformation Capture (Hi-C) offers a powerful approach to study microbial communities by linking genomic content with spatial interactions. Hi-C enhances shotgun sequencing by revealing taxonomic composition, functional interactions, and genomic organization from a single sample. However, aligning Hi-C reads to metagenomic contigs presents challenges, including the varying insert size of Hi-C paired-end reads, multi-species complexity, and gaps in assemblies. Although many benchmark studies have evaluated general alignment tools and Hi-C data alignment, none have specifically addressed metagenomics Hi-C data. Results: Here, we selected seven alignment strategies that have been used in Hi-C analyses: BWA MEM -5SP, BWA MEM default, BWA aln default, Bowtie2 default, Bowtie2 –very-sensitive-local, Minimap2 default, and Chromap Hi-C default. We benchmarked them on one synthetic and seven real-world environments, and evaluated these tools based on the number of inter-contig Hi-C read pairs and their influence on downstream tasks, such as binning quality. Conclusion: Our findings show that BWA MEM -5SP consistently outperforms other tools across all environments in terms of inter-contig read pairs and binning quality, followed by BWA MEM default. Chromap and Minimap2, while less effective in these metrics, demonstrate the highest computational efficiency.