Single-omics pipelines

BiomiX compares two groups using single and multiple omics. Input files must follow a specific format to work correctly.

Metadata file

A metadata file must be provided at launch. The first column must contain the sample ID; remaining columns contain sample variables.

Column	Description
ID Mandatory	First column. Unique sample identifier.
CONDITION Mandatory	Sample condition. These values will be available for group selection in the main interface.
SAMPLE_TYPE Optional	For convoluted matrices containing samples from different tissues or cell types (e.g., 25658_BLymphocyte). Allows analysis on a specific subset.
GENDER Optional	Used to adjust the Limma or DESeq2 model in transcriptomics.
AGE Optional	Used to adjust the Limma or DESeq2 model in transcriptomics.
Any other column Optional	Can be used to filter samples in the Metadata advanced options, or for PCA and UMAP visualization in Preview-QC.

Expression matrix

BiomiX requires a matrix where columns represent samples and rows represent variables. Accepted formats: .tsv, .xls, .xlsx.

• Transcriptomics: Variables (genes) must use Gene Symbol or Ensembl ID.
• Metabolomics: Variables (metabolites) should use KEGG, HMDB, or official compound names. For unannotated peaks, use the format peak_[number].
• Methylomics: Variables must use EPIC or 450K array probe codes.
• Undefined omics: Any variable name is accepted.

Directory system and reproducibility

BiomiX follows the FAIR principles (Findability, Accessibility, Interoperability, and Reuse).

• Save inputs in the correct location: Although BiomiX can read files from any location, we recommend saving a copy within the BiomiX shared folder — metadata in Metadata/, omics matrices in the corresponding omics subfolder.
• Analysis reports: After each completed analysis, a report file is automatically saved including filenames, analysis type, and all parameters.

Selecting Preview-QC in an omics input slot opens a Shiny interface for data exploration and quality control. It includes four sections:

• Plot section: Scatter plots of the top 50 high-variance variables, PCA, UMAP, and correlation heatmaps. Available transformations: logarithmic, median/mean centralization, Z-score, MAD, quantile, Loess, and VST.
• Data Table section: Preview of the full matrix.
• Outliers section: Filters variables by variance threshold and identifies sample outliers via PCA.
• Download section: Download the transformed matrix for external use.

Once satisfied with the transformation and QC, press the Close App button to pass the current matrix to the analysis pipeline.

The transcriptomics pipeline identifies differentially expressed genes (DEGs) and provides volcano plots and heatmaps. Default thresholds: Log2FC > 0.5 and adjusted p-value < 0.05. If GENDER and AGE columns are present in the metadata, they are used to adjust the statistical model.

If the data are transformed in Preview-QC, Limma is used for differential expression. If the matrix contains raw counts without transformation, DESeq2 is used instead. The transformed data are then passed to the integration module. To optimize speed, only the most variable genes are used as MOFA input (5,000 by default, adjustable in advanced options).

Biological process enrichment is explored using the R version of EnrichR. BiomiX also prepares export files for GSEA and EnrichR.

Additional features

Gene panel subgrouping

A gene panel can be provided via the GENE_PANEL file in the General section of advanced options. Genes must be in Gene Symbol format, one per line:

GENES_FOR_SUBPOPULATION
SIGLEC1
IFIT3
IFI6
LY6E
MX1
USP18
OAS3
IFI44L

These genes generate a heatmap showing each gene's expression Z-score across samples, enabling classification of condition samples into positive and negative subgroups. BiomiX then runs DGE analysis comparing controls with each subgroup.

Clinical or biological marker validation

Adding a MARKER column in the metadata allows sample classification based on a clinical or biological marker. This can then be compared with gene panel subgrouping to identify similarities or differences.

Peak or metabolite signals are first explored via Preview-QC. Transformed signals are then compared between groups: Log2FC is computed as the log2 ratio of median peak signals, and p-values are calculated with the Mann-Whitney test corrected by FDR. Volcano plots and heatmaps display significantly changed metabolites or peaks.

Pathway enrichment is performed using methPath v1.0.5 from TidyMass v1.0.8. BiomiX also prepares ready-to-copy input files for MetaboAnalyst. If transcriptomics results are available, BiomiX automatically prepares joint pathway and network analysis files.

Additional features

Targeted metabolomics

For data from LC-MS, GC-MS, CE-MS, or NMR where metabolite identities are already available. HMDB or KEGG identifiers are required for pathway analysis. IUPAC compound names are also accepted but will not trigger pathway analysis.

MS1 annotation (untargeted HRMS)

BiomiX uses CEU Mass Mediator to match m/z values of unannotated peaks. Default tolerance: 15 ppm for positive mode. Available adducts in positive mode: M+H, M+2H, M+Na, M+NH4, M+H-H2O. In negative mode: M-H, M-Cl, M+FA-H, M-H-H2O. Databases consulted: HMDB, LipidMaps, Metlin, KEGG (all customizable).

MS1 and DDA-MS/MS annotation (.mzML and .mgf)

Extends MS1 annotation with fragmentation spectra matching against HMDB, MassBank, and MoNA. MS2-based annotations replace MS1 ones where available, due to higher reliability.

Sample type filtering

BiomiX can filter metabolites using HMDB specimen-specific lists, retaining those identified in the relevant sample type: plasma, urine, saliva, CSF, feces, sweat, breast milk, bile, or amniotic fluid.

BiomiX performs differential methylation analysis using the ChAMP package. Output includes CpG island delta-beta values, FDR-adjusted p-values, a volcano plot, and a heatmap. Default thresholds: delta-beta > |0.15| and adjusted p-value < 0.05 (FDR).