Meta-analytic microbiome target discovery for immune checkpoint inhibitor response in advanced melanoma

This repository includes a set of analyses and visualization for a project entitled "Meta-analytic microbiome target discovery for immune checkpoint inhibitor response in advanced melanoma".

Overview

This repository contains a reproducible computational pipeline for conducting a meta-analysis of metagenomic samples in melanoma immunotherapy. Our study integrates microbial species, biosynthetic gene clusters (BGCs), and functional pathways across multiple datasets to identify key microbial signatures associated with immune checkpoint inhibitor (ICI) response.

Citation

If you use this pipeline in your research, please cite our study:

Zhang, X., Mallick, H. & Rahnavard, A. Meta-analytic microbiome target discovery for immune checkpoint inhibitor response in advanced melanoma. bioRxiv 2025, 2025.03.21.644637 (2025). https://doi.org/10.1101/2025.03.21.644637.

Key Features

• Comprehensive Meta-Analysis: The comprehensive meta-analysis of metagenomic samples in melanoma immunotherapy(N=678).
• Identification of Secondary Metabolites: Novel discovery of secondary metabolites linked to immunotherapy response.
• Biomarker Discovery: Faecalibacterium SGB15346 identified as a potential biomarker for ICI response.
• Functional Insights: RiPP biosynthetic gene clusters and Enterobacteriaceae-associated BGCs enriched in responders.
• Reproducible Analysis Pipeline: A standardized and reproducible workflow for future research applications.

Data Collection

We compiled publicly available whole metagenome shotgun sequencing (MGS) datasets from melanoma patients receiving immunotherapy. The following studies were included:

• GopalakrishnanV_2018 (PRJEB228939)
• MatsonV_2018 (PRJNA3997428)
• FrankelAE_2017 (PRJNA39790610)
• SpencerCN_2021 (PRJNA77029513)
• BaruchEN_2021 (PRJNA67873712)
• DavarD_2021 (PRJNA67286711)
• LeeKA_2022 (PRJEB431196)

Melanoma metagenomic samples

Table 1 summarizes melanoma metagenomic datasets included in this study and compares them with recent meta-analyses.

Melanoma Study / Meta-analysis Study	Our study	Cai et al. 2025	Zhu et al. 2025	Zhang et al. 2025	Olekhnovich et al. 2023	Lee et al. 2022	Limeta et al. 2020
Gopalakrishnan et al., 2017	25	25	0	0	25	25	25
Matson et al., 2018	38	38	0	38	38	38	38
McCulloch et al., 2022	46	46	46	46	0	0	0
Lee et al., 2022 - UK	55	55	0	55	55	55	0
Lee et al., 2022 - Netherlands	55	55	0	55	55	55	0
Lee et al., 2022 - Manchester	25	25	0	25	25	25	0
Lee et al., 2022 - Leeds	18	18	0	18	18	18	0
Lee et al., 2022 - Barcelona	12	12	0	12	12	12	0
Spencer et al., 2021	97	0	0	97	97	0	0
Frankel et al., 2017	39	0	0	39	39	39	39
Simpson et al., 2022	25	0	0	0	0	0	0
Glitza et al., 2024	47	0	0	0	0	0	0
Zhu et al. 2025	0	0	165	0	0	0	0
Wind et al., 2020	0	0	0	0	0	25	0
*Davar et al., 2021	181	0	181	0	181	0	0
*Baruch et al., 2020	40	0	40	0	40	0	0
*Bertrand et al., 2019	60	0	0	0	0	0	0
*ICI Total / ICI+FMT Total	482 / 281	274 / 0	211 / 221	385 / 0	364 / 221	292 / 0	102 / 0

Workflow

1. Quality Control & Preprocessing

• Removal of host DNA sequences using KneadData.

• Quality filtering and trimming with fastp.

• Taxonomic profiling via MetaPhlAn 4.

2. Functional Analysis

• Pathway Analysis: Conducted using HUMAnN3 to identify enriched pathways.

• Biosynthetic Gene Clusters (BGCs): Identified using antiSMASH 7.0 and BGCLens.

• Gene Families: Annotated using omePath.

3. Beta Diversity and Batch Effect Correction

• Bray-Curtis Dissimilarity: Used to assess inter-study variability.

• Batch Effect Correction: Performed using MMUPHin.R.

4. Meta-Analysis & Statistical Modeling

• Statistical associations analyzed using MaAsLin2 within MMUPHin.R.

• Compound Poisson Linear Model (CPLM) applied to identify significant microbiome features.

• PERMANOVA used to quantify variance explained by batch effects.

5. Data Visualization

• Volcano Plot: Run volcano_plot.R to visualize differentially abundant taxa or pathways.

• Heatmap: Use heatmap.R to generate heatmaps for significant associations.

Key Findings

• Faecalibacterium SGB15346 is significantly enriched in responders across multiple studies.

• RiPP biosynthetic gene clusters exhibit increased abundance in responders.

• dTDP-sugar biosynthesis pathways correlate with non-response.

• Batch effect correction successfully reduced inter-study variability while preserving biological signals.

Contact

For questions, please open an issue or contact Ali Rahnavard, Himel Mallick or Xinyang Zhang.

License

This project is licensed under the MIT License - see the LICENSE file for details.