Improved detection of disease-associated gut microbes using 16S sequence-based biomarkers

Brianna Sierra Chrisman, Kelley M. Paskov, Nate Tyler Stockham, Jae-Yoon Jung, Maya Varma, Peter Washington, Christine A. Tataru, Shoko Iwai, Todd Z. DeSantis, Maude M. David, Dennis P. WallBMC Bioinformatics 2021 · OpenReview

TL;DR

This paper introduces Sequence-Based Biomarkers (SBBs), a novel method for grouping 16S rRNA gene sequences to improve the detection of disease-associated gut microbes. SBBs outperform existing methods like OTU clustering and taxonomic classification in identifying microbial differences between healthy and diseased individuals and in predicting host phenotypes.

What Problem Does This Paper Address?

The human gut microbiome plays a crucial role in health and disease. 16S rRNA gene sequencing is a cost-effective way to study microbial communities, but analyzing this data presents challenges. Current methods for grouping microbial sequences—such as clustering similar sequences into Operational Taxonomic Units (OTUs), assigning taxonomic labels, or treating each unique sequence (ASV) individually—have limitations. These approaches often fail to capture the evolutionary relationships between microbes, hindering the ability to reliably identify microbial groups that are differentially abundant in disease states. This can lead to missed signals or an inability to pinpoint specific microbial contributions to conditions like autism or obesity.

Methods

The core innovation is the development of Sequence-Based Biomarkers (SBBs). Instead of relying on external databases or arbitrary similarity thresholds, SBBs leverage specific nucleotide variants (or combinations of variants) within the 16S rRNA gene sequence itself to group related microbes. The process involves:

  1. Multiple Sequence Alignment (MSA): Aligning all 16S sequences in a dataset to identify conserved and variable regions.
  2. Biomarker Computation: Identifying specific nucleotide bases at particular positions (single-locus biomarkers) or combinations of bases across multiple positions (multi-locus biomarkers) as SBBs. For example, a 2-locus SBB might be defined by a 'C' at position 10 and a 'G' at position 25.
  3. Linkage Disequilibrium (LD) Pruning: Removing redundant biomarkers that are highly correlated to avoid over-representation.
  4. Aggregation: Grouping ASVs that contain a specific SBB and then aggregating their relative abundances.

The authors compared SBBs (specifically 1-, 2-, and 3-locus SBBs) against established methods like OTU clustering (at 95%, 97%, and 99% similarity), Micropheno, and DiTaxa using three benchmarking tasks:

  • Permutation-based FDR: Assessing biomarker discovery power by comparing distributions of microbial group abundances between case and control groups.
  • Linear Discriminant Analysis (LDA) effect size (LefSe): Identifying biomarkers with significant differences in abundance.
  • Phenotype Prediction: Using random forest classifiers to predict host phenotypes based on microbial group abundances.

Datasets

The study utilized two independent gut microbiome datasets:

  1. Autism Dataset: Stool samples from 115 pairs of children with autism and their neurotypical siblings. This dataset used the V4 variable region of the 16S rRNA gene.
  2. Obesity Dataset: A published dataset of 196 obese and 61 lean twins, targeting the V3 region of the 16S rRNA gene.

These datasets represent real-world scenarios with different variable regions and sequencing facilities, demonstrating the robustness of the SBB method.

Key Insights and Ideas

  • Evolutionary Clues in 16S: The paper posits that specific nucleotide variations in the 16S rRNA gene act as "evolutionary clocks," providing a more direct way to group microbes based on their relatedness than traditional methods.
  • Beyond Taxonomy and Similarity: SBBs offer a "sweet spot" between the rigidity of taxonomic labels and the potential arbitrariness of similarity-based clustering, allowing for more nuanced microbial groupings.
  • Handling Complex Microbes: SBBs proved effective in identifying differentially abundant groups of microbes, even for challenging organisms like Blastocystis, which are difficult to classify using traditional taxonomic or OTU-based approaches due to their unique evolutionary history and lack of representation in standard databases.
  • Reduced Reliance on External Resources: A significant advantage of SBBs is that they do not require external reference databases, mitigating issues related to database incompleteness or outdated classifications.
  • Computational Efficiency: While higher-order SBBs can be computationally intensive, 2-locus SBBs offer a good balance between computational cost and microbial group specificity, making them a practical choice.

Key Results

  • Improved Biomarker Discovery: SBBs consistently identified differentially abundant microbial groups with similar or higher statistical significance compared to OTU clustering, Micropheno, and DiTaxa across both datasets. For example, in the obesity dataset, 1- and 2-locus SBBs identified nearly twice as many significant groupings as the next best strategy.
  • Phenotype Prediction: SBBs performed on par with or better than other methods in predicting host phenotypes. The autism phenotype was predicted with an ROC-AUC of 0.64, and the obesity phenotype with an ROC-AUC of 0.84. 3-locus SBBs were particularly effective for obesity prediction.
  • Identification of Disease-Associated Microbes: SBBs successfully identified known and novel microbial associations with autism (e.g., enrichment of Porphyromonadaceae, Ruminococcaceae; depletion of Veillonellaceae) and obesity (e.g., enrichment of Megasphaera, Sutterellaceae; depletion of Phocaeicola).
  • Robustness Across Datasets: The SBB method performed well on both datasets, despite differences in the 16S variable regions (V3 vs. V4) and sequencing facilities, highlighting its generalizability.

Limitations and Future Work

  • Causality: The authors acknowledge that SBBs, like other association-based methods, cannot infer causal relationships between microbes and disease.
  • Computational Cost: While 2-locus SBBs are efficient, higher-order SBBs (3-locus and above) can be computationally expensive due to the combinatorial explosion of possible biomarkers.
  • Hyperparameter Tuning: While SBBs require less hyperparameter tuning than some other methods, the choice of MSA algorithm and the number of loci for SBBs are still parameters to consider.
  • Future Directions: The authors suggest that future autism microbiome studies should consider including Blastocystis in their analysis, given its strong differential prevalence and abundance observed. They also propose 2-locus SBBs as a standard selection for feature extraction due to their balance of specificity and computational efficiency. The work paves the way for future analyses on the relationship between the gut microbiome and complex diseases using 16S sequencing.

Summary updated

Abstract

Background Sequencing partial 16S rRNA genes is a cost effective method for quantifying the microbial composition of an environment, such as the human gut. However, downstream analysis relies on binning reads into microbial groups by either considering each unique sequence as a different microbe, querying a database to get taxonomic labels from sequences, or clustering similar sequences together. However, these approaches do not fully capture evolutionary relationships between microbes, limiting the ability to identify differentially abundant groups of microbes between a diseased and control cohort. We present sequence-based biomarkers (SBBs), an aggregation method that groups and aggregates microbes using single variants and combinations of variants within their 16S sequences. We compare SBBs against other existing aggregation methods (OTU clustering and Microphenoor DiTaxa features) in several benchmarking tasks: biomarker discovery via permutation test, biomarker discovery via linear discriminant analysis, and phenotype prediction power. We demonstrate the SBBs perform on-par or better than the state-of-the-art methods in biomarker discovery and phenotype prediction. Results On two independent datasets, SBBs identify differentially abundant groups of microbes with similar or higher statistical significance than existing methods in both a permutation-test-based analysis and using linear discriminant analysis effect size. . By grouping microbes by SBB, we can identify several differentially abundant microbial groups (FDR <.1) between children with autism and neurotypical controls in a set of 115 discordant siblings. Porphyromonadaceae, Ruminococcaceae, and an unnamed species of Blastocystis were significantly enriched in autism, while Veillonellaceae was significantly depleted. Likewise, aggregating microbes by SBB on a dataset of obese and lean twins, we find several significantly differentially abundant microbial groups (FDR<.1). We observed Megasphaera andSutterellaceae highly enriched in obesity, and Phocaeicola significantly depleted. SBBs also perform on bar with or better than existing aggregation methods as features in a phenotype prediction model, predicting the autism phenotype with an ROC-AUC score of .64 and the obesity phenotype with an ROC-AUC score of .84. Conclusions SBBs provide a powerful method for aggregating microbes to perform differential abundance analysis as well as phenotype prediction. Our source code can be freely downloaded from http://github.com/briannachrisman/16s_biomarkers .
BibTeX
@article{chrisman2021improved,
  author = {Brianna Sierra Chrisman and Kelley M. Paskov and Nate Tyler Stockham and Jae-Yoon Jung and Maya Varma and Peter Washington and Christine A. Tataru and Shoko Iwai and Todd Z. DeSantis and Maude M. David and Dennis P. Wall},
  title = {{Improved detection of disease-associated gut microbes using 16S sequence-based biomarkers}},
  journal = {BMC Bioinformatics},
  year = {2021},
  url = {https://openreview.net/forum?id=CcSAcZqfT2},
}
Figure