This repository contains the computational and analytical framework used to characterize toxin–antitoxin (TA) systems in the oral microbiome using metatranscriptomic data from two publicly available cohorts: Dieguez et al. and Ev datasets.

The analysis integrates differential expression, curated functional annotations, and detailed visualizations to identify and interpret the transcriptional activity of TA gene pairs in healthy and caries-associated and caries-treated oral microbiomes. All scripts, intermediate data files, and processed outputs are included to enable reproducibility, transparency, and downstream reuse.

${\color{red}Disclaimer}$ - The raw FASTQ reads, intermediate preprocessing files, and full functional annotations are not included here due to storage constraints. However, detailed scripts and a step-by-step tutorial are provided below.

• Toxin–Antitoxin Systems in the Oral Microbiome
  • Repository Overview & Structure
  • Installation instructions
    • Clone the Repository
    • Install packages and download databases
  • Metatranscriptomic Data Processing Pipeline
    • Download Raw Reads (SRA/ENA)
    • Quality Control and Trimming (Trim Galore)
    • Host and rRNA Removal (SortMeRNA)
  • Functional Profiling (HUMAnN)
    • Directory Structure
  • Functional annotations
    • UniRef90 ID Extraction and FASTA Retrieval
    • Database annotations
  • Downstream Analysis: Differential Expression and Visualization
    • Environment Setup
    • Data Cleaning and Preprocessing
    • UniRef90 Toxin Cluster Overlap and Venn Analysis
    • Differential Abundance Analysis (ANCOM-BC)
    • Volcano Plot Visualization
    • Expression Heatmaps (ComplexHeatmap)
    • Gene Set Overlap Analysis: UpSet and Venn Panels
      • UpSet Plot of Significant TA Genes
      • Comparison Label Harmonization
      • Venn Diagram Panels (Three-Way Intersection)
      • Venn Panel Reference Table
  • Functional Evidences
    • Parsing TADB3 DIAMOND Results
    • Annotating Significant Genes with TADB3
    • Annotating Significant Genes with VFDB
    • Annotating Significant Genes with eggNOG-mapper
    • Annotating Significant Genes with InterProScan
  • Validated TA Gene Clusters
    • Summary Statistics by Condition
    • Condition-Wise Expression Barplots
  • Directory structure
  • Packages to install
  • Authors and Maintainers
  • Questions and Feedback

${\color{blue}Note}$ - All files larger than 50MB are tracked using Git LFS. Access instructions are provided below.

All analyses were developed and executed on a Linux-based HPC cluster and assume working familiarity with common bioinformatics tools plus basic R/Python scripting.

git clone https://github.com/biocoms/ta_systems_oral_mt.git
cd ta_systems_oral_mt

To streamline setup, we provide a single bash script that installs all dependencies in one step.

${\color{blue}Note}$ - Ensure this github is cloned, you are inside this folder and the conda is installed.

bash scripts/setup.sh

If any large files are missing or stubbed out, ensure Git LFS is installed. You can also install git lfs via conda and activate in git as shown below.

git lfs install
git lfs pull
git lfs track "filename or file extension"
git add .gitattributes

The following section details the entire preprocessing workflow using publicly available tools, from raw sequencing reads to functional gene and pathway profiles.

Enter these BioProject numbers in SRA-Explorer, copy the FTP links of the paired-end FASTQ files to a .txt file and download using wget

• Dieguez: BioProject PRJNA712952

• Ev: BioProject PRJNA930965

wget -i dieguez.txt
wget -i ev.txt

Make sure dieguez.txt and ev.txt each contains full FTP links for both R1 and R2 reads e.g.:

ftp://ftp.sra.ebi.ac.uk/vol1/fastq/SRR233/020/SRR23351020/SRR23351020_1.fastq.gz
ftp://ftp.sra.ebi.ac.uk/vol1/fastq/SRR233/020/SRR23351020/SRR23351020_2.fastq.gz
.......
.......

Trim Galore is a wrapper tool that combines FastQC and Cutadapt to perform quality filtering and adapter trimming.

Outputs:

• Trimmed paired FASTQ files (e.g., _1_trimmed.fastq.gz,_2_trimmed.fastq.gz)
• FastQC quality reports (_fastqc.html,_fastqc.zip)

Command Example:

trim_galore --paired --fastqc --gzip --phred33 --length 50 \
  --output_dir dieguez/trimmed_reads sample_1.fastq.gz sample_2.fastq.gz

• --paired: Indicates paired-end reads.

• --fastqc: Runs FastQC before and after trimming for QC reports.

• --gzip: Compresses the output files in .gz format.

• --phred33: Specifies base quality encoding (standard for Illumina).

• --length 50: Discards reads shorter than 50 bp after trimming.

• --output_dir: Output folder for trimmed reads and FastQC reports.

This step is applied to both the Dieguez and Ev datasets. FastQC reports are generated to evaluate sequence quality before and after trimming.

SortMeRNA filters out unwanted rRNA and host (human) reads using curated databases.

Databases Used:

• SILVA rRNA databases (16S, 23S, 5S, etc.)
• Rfam (non-coding RNAs)
• GRCh38 (human genome and transcriptome)

All reference FASTA files are stored in: dbs/sortmerna_databases/

Inputs:

Trimmed paired FASTQ files (trimmed_reads/_1_trimmed.fastq.gz and trimmed_reads/_2_trimmed.fastq.gz)

Outputs:

• *_aligned.fastq.gz — reads matching rRNA/host
• *_unaligned.fastq.gz — filtered reads for downstream analysis

Command Template:

sortmerna --ref dbs/ref_sortmerna/silva-bac-16s-id90.fasta \
          --reads sample_1_trimmed.fastq.gz \
          --reads sample_2_trimmed.fastq.gz \
          --fastx \
          --aligned output_dir/sample_aligned \
          --other output_dir/sample_unaligned \
          --threads 32                        #flexible as per available cores

• --ref: Specifies each reference database for filtering.

• --reads: Input FASTQ files (both forward and reverse).

• --fastx: Ensures FASTQ format is preserved for input/output.

• --aligned: Output prefix for reads that matched references.

• --other: Output prefix for reads that did not match references.

• --threads: Number of threads to use for parallel processing.

Filtered reads are saved into:

• dieguez/trimmed_reads/sortmerna_unaligned/
• ev/trimmed_reads/sortmerna_unaligned/

These unaligned reads are used as input for HUMAnN.

HUMAnN3 (The HMP Unified Metabolic Analysis Network) is used to profile gene families and metabolic pathways from host-filtered microbial reads.

Databases Required:

• ChocoPhlAn: nucleotide-level mapping
• UniRef90 or UniRef50: translated protein database for function

Inputs:

• .fq.gz files from sortmerna_unaligned/ directory

Outputs (for each sample):

• {sample}_genefamilies.tsv
• {sample}_pathabundance.tsv
• {sample}_pathcoverage.tsv

Command Template:

humann -i sample_unaligned.fq.gz \
       -o humann_output/sample_name \
       --threads 32 \
       --nucleotide-database humann/dbs/chocophlan \
       --protein-database humann/dbs/uniref \
       --verbose

• -i: Input FASTQ file (filtered reads).
• -o: Output directory for results.
• --threads: Number of parallel threads for speed.
• --nucleotide-database: Path to ChocoPhlAn nucleotide database.
• --protein-database: Path to UniRef protein database.
• --verbose: Prints detailed progress during execution.

Run separately for each dataset (Dieguez and Ev).

Run separately for each dataset (Dieguez and Ev). This step generates functional profiles that can be used for downstream gene analysis and comparison towards TA gene clusters.

ta_systems_oral_mt/
├── dieguez/
│   ├── trimmed_reads/
│   │   ├── *_1_trimmed.fastq.gz
│   │   ├── *_2_trimmed.fastq.gz
│   │   ├── *_fastqc.html
│   │   ├── *_fastqc.zip
│   │   ├── sortmerna/
│   │   │   ├── <sample>/
│   │   │   │   ├── <sample>_aligned.fastq.gz
│   │   │   │   ├── <sample>_unaligned.fastq.gz
│   │   └── sortmerna_unaligned/
│   │       └── *.fq.gz                     # Filtered (host & rRNA removed) reads
│   └── humann/
│       └── <sample>/
│           ├── <sample>_genefamilies.tsv
│           ├── <sample>_pathabundance.tsv
│           └── <sample>_pathcoverage.tsv
│
├── ev/
│   ├── trimmed_reads/
│   │   ├── *_1_trimmed.fastq.gz
│   │   ├── *_2_trimmed.fastq.gz
│   │   ├── *_fastqc.html
│   │   ├── *_fastqc.zip
│   │   ├── sortmerna/
│   │   │   ├── <sample>/
│   │   │   │   ├── <sample>_aligned.fastq.gz
│   │   │   │   ├── <sample>_unaligned.fastq.gz
│   │   └── sortmerna_unaligned/
│   │       └── *.fq.gz                     # Filtered reads
│   └── humann/
│       └── <sample>/
│           ├── <sample>_genefamilies.tsv
│           ├── <sample>_pathabundance.tsv
│           └── <sample>_pathcoverage.tsv
│
├── dbs/
│   └── sortmerna_databases/                     # SortMeRNA reference databases
│       ├── *.fasta
│   └── humann_databases/
│       ├── chocophlan/                    # HUMAnN nucleotide database
│       └── uniref/                        # HUMAnN protein database
│
├── dieguez.txt                            # FTP links for Dieguez dataset
├── ev.txt                                 # FTP links for Ev dataset
├── ta_process.sh                          # Main preprocessing script
└── log/
    └── processing_pipeline.log            # Cluster job log output

We first extracted unique UniRef90 gene family IDs from the *_genefamilies.tsv outputs of HUMAnN. These IDs were then used to fetch FASTA sequences for downstream functional annotation using InterProScan, eggNOG-mapper, TADB3, and VFDB.

Inputs:

• Directory of *_genefamilies.tsv files from HUMAnN for each sample
  (raw_data/dieguez_genefamilies.tsv, raw_data/ev_genefamilies.tsv)

Steps

1. Extract UniRef90 IDs
   • Parsed each genefamilies file to collect only those gene families starting with UniRef90_.
   • Saved as one ID list per sample in:
     uniref_ids/extracted_ids/*_uniref90_ids.txt
2. Download FASTA Sequences
   • Queried the UniProt REST API in chunks (default = 500 IDs/query).
   • Fetched .fasta sequences corresponding to UniRef90 IDs per sample.
   • Logs were saved for each sample to monitor download status.

Script:

• scripts/uniref_idmapping.py

python scripts/uniref_idmapping.py \
  --input_dir raw_data/dieguez_genefamilies.tsv \
  --output_dir dieguez_uniref_mapped/

python scripts/uniref_idmapping.py \
  --input_dir raw_data/ev_genefamilies.tsv \
  --output_dir ev_uniref_mapped/

Outputs:

• uniref_ids/extracted_ids/*.txt – Sample-wise UniRef90 ID lists
• uniref_ids/fastas/*.fasta – Fetched amino acid sequences
• uniref_ids/logs/*.log – Chunk-wise log of downloads per sample

The shell script database_annotations.sh automates batch functional annotation using four major tools: VFDB, TADB3, eggNOG-mapper, and InterProScan, applied to UniRef90-mapped FASTA sequences from the Dieguez and EV datasets.

Inputs

• FASTA Sequences: Protein-coding genes extracted using UniRef90 IDs from significant TA genes.

  • Located in:
    • dieguez_uniref_mapped/fastas/
    • ev_uniref_mapped/fastas/

• Databases:

  • vfdb/VFDB_prot.dmnd
  • tadb3/tadb3_combined.fasta/*.dmnd
  • dbs/eggnog_data/eggnog_proteins.dmnd
  • dbs/interproscan-5.72-103.0/interproscan.sh

• Threads: VFDB/TADB3 (64), eggNOG (80), InterProScan uses all available CPUs.

Steps

1. VFDB (DIAMOND Blastp)
   Matches each UniRef90 FASTA file to the VFDB protein database.
   Output format: tab-delimited DIAMOND blastp (.tsv)

2. TADB3 (DIAMOND Blastp to Multiple DBs)
   Iteratively matches each FASTA to all *.dmnd toxin/antitoxin databases from TADB3.

3. eggNOG-mapper
   Runs emapper.py with:

   • Minimum 50% identity (--pident)
   • Minimum 80% query coverage (--query_cover)
   • DIAMOND mode, using custom eggNOG directory

4. InterProScan
   Annotates each FASTA using InterProScan with GO term and IPR lookups.
   Output format: tab-separated .tsv with all domain and functional annotations.

Outputs

Execution

To run all annotations:

bash database_annotations.sh

All downstream analyses were conducted in R using the ta_systems_oral_mt.Rproj environment with strict reproducibility ensured via renv.

Activate the project-local R environment using:

# From within the R project root
install.packages("renv")
renv::restore()

This installs all required packages at pinned versions as declared in renv.lock.

Alternatively, for manual setup, install the following CRAN and Bioconductor packages:

# CRAN packages
install.packages(c(
  "tidyverse", "dplyr", "ggplot2", "gridExtra", "patchwork", 
  "VennDiagram", "RColorBrewer", "pheatmap", "UpSetR"
))

# Bioconductor
if (!requireNamespace("BiocManager", quietly = TRUE))
  install.packages("BiocManager")
BiocManager::install(c(
  "phyloseq", "ANCOMBC", "ComplexHeatmap", "circlize", "microbiome"
))

Raw gene family expression files (*_genefamilies.tsv) generated by HUMAnN were cleaned and normalized as follows:

• Retained only gene families prefixed with UniRef90_
• Removed unmapped or low-abundance entries (e.g., UNMAPPED)
• Standardized sample identifiers by removing .RPKs suffixes
• Ensured uniqueness of gene family rows

Output Files:

• processed_data/processed_dieguez_genes.csv
• processed_data/processed_ev_genes.csv

We intersected processed gene family tables with the curated UniRef90 toxin cluster list (uniref90_toxin_toxic.tsv) to identify shared and unique transcriptional signals across datasets.

• Computed overlaps between Dieguez and EV gene families with UniRef90 toxin genes
• Separated results into:
  • Dieguez-only overlaps
  • EV-only overlaps
  • Shared UniRef90-positive genes

Output Matrices:

• uniref_tox_abundance/dieguez_final_uniref90_mapped.csv
• uniref_tox_abundance/ev_final_uniref90_mapped.csv
• uniref_tox_abundance/intersection_dieguez_final_uniref90_mapped.csv
• uniref_tox_abundance/intersection_ev_final_uniref90_mapped.csv

Venn Diagram

• venn_diagrams/venn_dieguez_ev_uniref.png

Differential abundance testing was conducted using ANCOM-BC independently within each dataset.

Workflow:

• Constructed phyloseq objects from UniRef90-filtered matrices
• Enumerated all valid pairwise condition comparisons
• Applied ancombc() with the model formula ~ condition
• Exported full and filtered (significant) result tables

Output Tables:

• DA_results/da_complete/*.csv
• DA_results/da_sig/*.csv

For each ANCOM-BC comparison, volcano plots were generated to highlight gene-level statistical significance and fold-change magnitude.

• x-axis: $\log_{2}$ fold-change (lfc)
• y-axis: –\log_{10}$(p-value)
• Color scheme: red (significant), gray (non-significant)
• Vertical and horizontal thresholds marked at ±1 $\log_{2}$(FC) and p = 0.05

Output Directory:

• DA_results/volcano_plot/*.png

Significantly differentially expressed TA gene clusters were visualized using scaled heatmaps.

• Expression values were $\log_{2}$-transformed with pseudocounts
• Row-wise scaling was applied to highlight relative expression shifts
• Columns were annotated using condition and host_subject_id
• Color palettes were customized to reflect biological and sample group distinctions

Output Directory:

• DA_results/heatmaps/*.png

To summarize and visualize shared transcriptional activity of UniRef90-mapped toxin–antitoxin genes across conditions, we generated both UpSet plots and Venn diagrams.

The UpSet plot captures the intersection of all significantly expressed UniRef90 genes across validated pairwise comparisons.

Highlights:

• Rows: Gene clusters
• Columns: Comparisons (Dieguez and EV)
• Bars: Number of shared genes across comparison sets
• Queries:
  • Red: genes annotated as toxins
  • Green: genes annotated as antitoxins

Color Mapping:

• Dieguez comparisons: #E82561
• EV comparisons: purple

Output:

• DA_results/upset_plot/upset_plot.png

To ensure consistency in comparison labels across datasets, condition names were programmatically cleaned and standardized. Reverse contrasts (e.g., B vs A and A vs B) were collapsed.

Nine thematic panels (B to J) were designed to illustrate 3-way overlap across clinically meaningful comparisons.

Examples:

• Panel B: Caries–CA* vs Healthy–CF (AAa, CAi, CAs)
• Panel H: Healthy-only timepoint comparisons
• Panel J: Healthy vs Caries across all stages

Each Venn diagram shows:

• Size-labeled sets with word-wrapped condition labels
• Color-coded ellipses from RColorBrewer::Set2
• Gene counts in intersections and unique sections

Output Directory:

• DA_results/venn_panels/

File Naming Convention:

• Venn_Panel_<Letter>.png

Supporting Tables:

• All_Venn_Sets_Combined.csv: All individual sets per panel
• All_Venn_Intersections_3way.csv: Genes shared by all three comparisons per panel

tadb3_master_table.py parses DIAMOND alignment outputs against the TADB3 database and generates a consolidated annotation table for all matched genes across Dieguez and Ev datasets.

Inputs:

• DIAMOND outputs:
  • tadb3/diamond_dieguez/*.tsv
  • tadb3/diamond_ev/*.tsv
• Reference FASTA:
  • tadb3/tadb3_combined.fasta

Steps:

1. Parse TADB3 FASTA to extract Protein_Accession, Genome_Accession, and Organism.
2. Read DIAMOND result files per sample; each file contains UniRef90 hits to TADB3 proteins.
3. Merge DIAMOND hits with metadata extracted from the FASTA.
4. Annotate each hit with:
   • Toxin_Antitoxin: inferred from ID patterns (_tox_, _antitox_)
   • Validation_Type: Experimental or Computational based on UniRef90 ID
5. Record Sample_Name and dataset (dieguez or ev) for each entry.

Output:

• tadb3/tadb3_tox_antitox_main_table_1.tsv: master table with annotation metadata for all TADB3-aligned genes across both datasets.

tadb3_sig_genes.py filters and annotates TADB3-aligned genes that are differentially expressed in the Dieguez or EV datasets. This script integrates DIAMOND-based hits with metadata from the TADB3 master table and augments entries with protein descriptions retrieved from NCBI.

Inputs:

• DA_results/sig_genes.txt: List of significant UniRef90 gene IDs
• tadb3/master_table_tadb3.txt: Annotated master table from tadb3_master_table.py

Steps:

1. Load significant UniRef90 IDs.
2. Subset TADB3 hits for these IDs and retain only high-confidence matches (Bit Score ≥ 90).
3. Standardize organism names to species-level.
4. Query NCBI Entrez for full protein descriptions using Protein_Accession and clean the output.
5. For each gene, perform annotation collapse (e.g., most common Organism, validation type, aggregated annotations).

Output:

• tadb3/sig_genes_annotations.csv: Final annotation table for significant TA genes enriched with curated metadata and Entrez descriptions.

vfdb_sig_genes.py filters DIAMOND alignments against the VFDB (Virulence Factor Database) to extract and annotate significant genes identified from the Dieguez and EV datasets.

Inputs:

• DA_results/sig_genes.txt: List of significant UniRef90 gene IDs.
• DIAMOND outputs:
  • vfdb/dieguez/*_vs_VFDB.tsv
  • vfdb/ev/*_vs_VFDB.tsv

Steps:

1. Load significant UniRef90 IDs and subset the DIAMOND VFDB alignments to retain only matches with these genes.
2. Extract GenBank accession numbers from Query_ID using regex.
3. Query NCBI Entrez using Bio.Entrez to fetch gene names and protein descriptions for each accession.
4. Append sample name, dataset, gene name, and description to each annotated entry.
5. Collapse annotations by gene using custom aggregation logic.

Output:

• vfdb/sig_genes_annotations.csv: Annotated table of significant genes aligned to VFDB with NCBI-derived metadata.

eggnog_sig_genes.py extracts and summarizes functional annotations for significant UniRef90 gene families using EggNOG-mapper outputs from the Dieguez and EV datasets.

Inputs:

• DA_results/sig_genes.txt: List of significant UniRef90 gene IDs (dot-truncated)
• Annotated .emapper.annotations files from EggNOG-mapper:
  • eggnog/dieguez/*_eggnog.emapper.annotations
  • eggnog/ev/*_eggnog.emapper.annotations

Steps:

1. Parse each .emapper.annotations file while ignoring comment lines and correctly identifying the column header.
2. Filter rows where the query matches a significant UniRef90 gene ID.
3. Annotate each record with dataset and sample origin.
4. Collapse annotations per UniRef90 ID by computing the most frequent (mode) entry for each functional field (e.g., KEGG_ko, EC, eggNOG_OGs, GO_terms).
5. Track how many and which samples contributed to each gene’s annotation.

Output:

• eggnog/sig_genes_annotations.csv: Collapsed annotation table of significant genes, including sample metadata and representative functional terms.

interproscan_sig_genes.py parses and annotates InterProScan .tsv results for significant UniRef90 gene families, enriching each entry with GO term descriptions and ontology categories.

Inputs:

• DA_results/sig_genes.txt: List of significant UniRef90 gene IDs.
• InterProScan output files:
  • interproscan/dieguez/*.tsv
  • interproscan/ev/*.tsv

Steps:

1. Load .tsv files and assign dataset/sample labels.
2. Filter records based on significant gene IDs.
3. Extract GO identifiers (GO:XXXXXXX) from the GO column.
4. Query the EBI QuickGO API to retrieve term names and ontology categories (BP, MF, CC).
5. Annotate each record with:
   • Full GO term descriptions
   • Separated Biological Process, Molecular Function, Cellular Component
6. Aggregate and collapse annotations for each gene across samples and datasets.

Output:

• interproscan/sig_genes_annotations.csv: Clean annotation table with GO terms and structured categories for significant genes.

This section summarizes and visualizes the expression behavior of experimentally validated or computationally predicted toxin–antitoxin (TA) system genes based on curated annotations from TADB3. Two components are included:

The script valid_gene_summary.R compiles condition-wise expression summaries for all validated TA system genes across the Dieguez and Ev datasets.

Inputs:

• processed_data/processed_dieguez_genes.csv
• processed_data/processed_ev_genes.csv
• raw_data/metadata_dieguez.csv
• raw_data/metadata_ev.csv
• tadb3/tadb3_tox_antitox_main_table.tsv

Steps:

1. Subset expression matrices to retain only TA genes annotated in the TADB3 table.
2. Merge with metadata to assign each sample to a condition group.
3. Apply a pseudocount (0.01) and compute log10(expression + 0.01) per sample.
4. For each gene × condition × dataset, compute:
   • n, non-zero sample count
   • min, max, mean, median
   • mean_log10, median_log10
   • Gene label (Toxin or Antitoxin)
5. Merge Dieguez and EV summaries into a combined summary table.

Output:

• DA_results/debug_gene_condition_summary_combined.csv

The script barplots.R creates annotated barplots for each significant TA gene identified in sig_genes_combined.csv, stratified by dataset and gene type.

Inputs:

• processed_data/processed_dieguez_genes.csv
• processed_data/processed_ev_genes.csv
• raw_data/metadata_dieguez.csv
• raw_data/metadata_ev.csv
• DA_results/sig_genes_combined.csv
• tadb3/tadb3_tox_antitox_main_table.tsv

Steps:

1. Subset significant genes by type (Toxin, Antitoxin).
2. Plot per-gene barplots using:
   • Y-axis: log1p(expression)
   • X-axis: condition group (ordered)
   • Mean ± standard error
3. Overlay ANCOM-BC adjusted p-values as:
   • Brackets for significant comparisons only
   • Significance stars in black
4. Customize plot aesthetics:
   • Red text for toxins, green text for antitoxins
   • One dot per condition per gene (no jitter)

Outputs:

• DA_results/barplots/toxin_barplots/*.png
• DA_results/barplots/antitoxin_barplots/*.png

ta_systems_oral_mt/
├── DA_results/              # Differential expression results - tables and visualizations
├── dieguez_uniref_mapped/   # Sample-wise mapping of the functional profiled UniRef90 Ids and Fastas
├── eggnog/                  # Functional annotations from eggNOG-mapper
├── ev_uniref_mapped/        # Sample-wise mapping of the functional profiled UniRef90 Ids and Fastas
├── interproscan/            # InterProScan annotations for significant genes
├── processed_data/          # Preprocessed and cleaned expression matrices
├── raw_data/                # Raw input files (metadata, UniRef90 toxin-antitoxin files, functional gene tables)
├── scripts/                 # All analysis scripts (R/Python)
├── tadb3/                   # TADB3 toxin–antitoxin annotation tables and sequences
├── uniref_tox_abundance/    # Abundance tables intersecting Dieguez, ev and UniRef90 toxin–antitoxin gene clusters
├── venn_diagrams/           # Venn diagrams used in the manuscript
├── vfdb/                    # VFDB annotations (dieguez/ev-specific)
├── README.md                # Project description and documentation
├── LICENSE                  # License for usage
└── ta_systems_oral_mt.Rproj # R project file for downstream analysis

Shri Vishalini Rajaram, Priyanka Singh and Erliang Zeng

Shri Vishalini Rajaram, Priyanka Singh, & Erliang Zeng. (2025). biocoms/ta_systems_oral_mt: Bacterial Toxin-Antitoxin Systems in Oral Metatranscriptome (v1.0.0). Zenodo. https://doi.org/10.5281/zenodo.16763308

This repository is maintained by the Bioinformatics and Computational Systems Biology Lab at the University of Iowa.

If you notice any errors, have suggestions for improvement, or simply want to discuss aspects of the analysis, we genuinely welcome your feedback. Please feel free to open an issue or contact us. We are always happy to engage with the community.