iNAP 2.0: Harnessing metabolic complementarity in microbial network analysis

Xi Peng, Kai Feng, Xingsheng Yang, Qing He, Bo Zhao, Tong Li, Shang Wang, Ye DengiMeta 2024 · DOI

TL;DR

This paper introduces iNAP 2.0, an advanced bioinformatics pipeline that leverages genome-scale metabolic models (GSMMs) to analyze metabolic complementarity and infer interspecies interactions within microbial communities. It offers novel methods for constructing and analyzing these metabolic networks, providing deeper insights into microbial ecology than traditional co-occurrence analyses.

What Problem Does This Paper Address?

Understanding microbial communities is crucial for fields ranging from environmental science to human health. Traditional methods often rely on marker gene sequencing and co-occurrence networks, which can reveal statistical relationships but struggle to uncover the underlying functional mechanisms driving these interactions. While genome-scale metabolic modeling (GSMM) offers a more mechanistic approach, analyzing the vast amount of data generated and inferring complex interactions between numerous microbes has been a significant challenge. Researchers often lack integrated tools and standardized methods for this analysis, especially when dealing with metagenome-assembled genomes (MAGs). This paper addresses this gap by providing a comprehensive, user-friendly pipeline (iNAP 2.0) that integrates multiple state-of-the-art methods for metabolic interaction network construction and analysis.

Methods

iNAP 2.0 is built around a four-module process:

  1. Prepare Genome-Scale Metabolic Models (GSMMs): The pipeline can automatically reconstruct GSMMs from genome or protein sequences using tools like Prokka for annotation and CarveMe for model building. It also supports importing pre-built models or manually curated ones. A key feature is the ability to perform "gap-filling" for MAGs, which is crucial for models derived from incomplete genomic data.
  2. Infer Pairwise Interactions: iNAP 2.0 offers three primary methods to quantify metabolic interactions between pairs of GSMMs:
  3. PhyloMint: Calculates metabolic complementarity and competition indices, considering phylogenetic relationships. It also identifies "potentially transferable metabolites" (PTMs).
  4. SMETANA (Species Metabolic Interaction Analysis): Quantifies metabolic resource overlap and interaction potential within communities. It offers both global and iterative analyses to account for potential inconsistencies.
  5. Metabolic Distance (based on pFBA): Calculates the dissimilarity between metabolic flux profiles of GSMMs using parsimonious flux balance analysis.
  6. Construct Metabolic Interaction Networks: A critical innovation in iNAP 2.0 is the integration of Random Matrix Theory (RMT) to determine statistically robust thresholds for network construction. This helps overcome the challenge of arbitrary threshold selection common in previous methods. It also supports Z-score outlier detection for SMETANA and metabolic distance indices.
  7. Analyze Metabolic Interaction Networks: Once constructed, networks can be analyzed for global properties, individual node centrality, module separation, and identification of keystone or hub nodes. iNAP 2.0 also facilitates the visualization of microbe-metabolite bipartite networks and metabolic distance heatmaps.

Datasets

The paper uses demo datasets for demonstration purposes:

  • Demo data set 1: A seven-strain microbiome genome set (SIHUMI) from the NCBI RefSeq database.
  • Demo data set 2: A genome set of 100 metagenome-assembled genomes (MAGs) from a hot spring habitat.

These datasets are used to illustrate the pipeline's functionality and the interpretation of results, rather than for benchmarking against established datasets.

Key Insights and Ideas

  • Bridging the gap between genomics and function: iNAP 2.0 moves beyond simple co-occurrence to infer functional metabolic interactions, providing a more mechanistic understanding of microbial communities.
  • The power of RMT for network construction: Using Random Matrix Theory to set network thresholds is a significant improvement, offering a more objective and statistically sound approach compared to arbitrary cutoff values.
  • Identifying transferable metabolites: The ability to pinpoint potentially transferable metabolites (PTMs) and visualize them in microbe-metabolite bipartite networks offers a novel way to understand nutrient exchange and dependencies.
  • Integrated workflow for complex data: The pipeline consolidates multiple advanced tools into a single, user-friendly platform, significantly lowering the barrier to entry for researchers wanting to perform complex metabolic network analyses, especially with MAGs.
  • Addressing MAG limitations: The inclusion of gap-filling for GSMMs derived from MAGs is crucial, as these often lack complete genomic information.

Key Results

The paper primarily focuses on presenting the iNAP 2.0 pipeline and its capabilities. It demonstrates how the pipeline can be used to:

  • Generate GSMMs from raw genomic data.
  • Calculate various metabolic interaction indices (PhyloMint, SMETANA, metabolic distance).
  • Construct statistically robust metabolic interaction networks using RMT.
  • Analyze these networks to identify key players and community structures.
  • Visualize complex relationships, such as microbe-metabolite interactions and metabolic distance heatmaps.

The paper provides detailed step-by-step instructions and example outputs (Figures 3A and 3B) to showcase the pipeline's functionality.

Limitations and Future Work

  • GSMM Quality: The authors acknowledge that iNAP 2.0 does not support GSMM reconstruction quality evaluation. They emphasize that the credibility of GSMMs depends heavily on the quality of the input genomes (prioritizing complete genomes over MAGs, and applying stricter criteria for MAG refinement).
  • SMETANA Computational Cost: SMETANA is noted as a "very time-consuming program," and the limit of 300 genomes/models for SMETANA analysis is highlighted. For larger datasets, users are advised to reduce the number of models or use high-performance computing (HPC).
  • Iterative SMETANA: The paper mentions an "inconsistency" in SMETANA results due to the CPLEX solver and recommends running multiple iterations to obtain an average value.
  • Future Work: While not explicitly detailed as future work, the comprehensive nature of iNAP 2.0 suggests potential for further integration of new metabolic modeling and network analysis tools as they emerge. The availability of the pipeline as a free, web-accessible tool also encourages community adoption and potential contributions.

Summary updated

Abstract

With the widespread adoption of metagenomic sequencing, new perspectives have emerged for studying microbial ecological networks, yielding metabolic evidence of interspecies interactions that traditional co-occurrence networks cannot infer. This protocol introduces the integrated Network Analysis Pipeline 2.0 (iNAP 2.0), which features an innovative metabolic complementarity network for microbial studies from metagenomics sequencing data. iNAP 2.0 sets up a four-module process for metabolic interaction analysis, namely: (I) Prepare genome-scale metabolic models; (II) Infer pairwise interactions of genome-scale metabolic models; (III) Construct metabolic interaction networks; and (IV) Analyze metabolic interaction networks. Starting from metagenome-assembled or complete genomes, iNAP 2.0 offers a variety of methods to quantify the potential and trends of metabolic complementarity between models, including the PhyloMint pipeline based on phylogenetic distance-adjusted metabolic complementarity, the SMETANA (species metabolic interaction analysis) approach based on cross-feeding substrate exchange prediction, and metabolic distance calculation based on parsimonious flux balance analysis (pFBA). Notably, iNAP 2.0 integrates the random matrix theory (RMT) approach to find the suitable threshold for metabolic interaction network construction. Finally, the metabolic interaction networks can proceed to analysis using topological feature analysis such as hub node determination. In addition, a key feature of iNAP 2.0 is the identification of potentially transferable metabolites between species, presented as intermediate nodes that connect microbial nodes in the metabolic complementarity network. To illustrate these new features, we use a set of metagenome-assembled genomes as an example to comprehensively document the usage of the tools. iNAP 2.0 is available at https://inap.denglab.org.cn for all users to register and use for free.
BibTeX
@inproceedings{peng2024inap,
  author = {Xi Peng and Kai Feng and Xingsheng Yang and Qing He and Bo Zhao and Tong Li and Shang Wang and Ye Deng},
  title = {{iNAP 2.0: Harnessing metabolic complementarity in microbial network analysis}},
  booktitle = {iMeta},
  year = {2024},
  doi = {10.1002/imt2.235},
}
Figure