Profiler v1.2 — Open Multi-Omics Analysis Platform

Where Omics Meet Clarity
Developed by PRISM U1192 Laboratory, Université de Lille — Protected by INSERM Transfert

🌐 Try Profiler Live  ·  💻 Profiler Desktop  ·  📦 Test Datasets  ·  📄 Publication

---

Profiler is a free, open, web-based platform for end-to-end multi-omics data analysis, from raw instrument files to publication-ready figures and biological interpretation. No programming required.

---

Citation

If you use Profiler in your research, please cite:

Zirem, Y., Ledoux, L., Fournier, I., & Salzet, M.
Profiler: an open web platform for multi-omics analysis.
Bioinformatics, Oxford University Press, 2025.
DOI: 10.1093/bioinformatics/btaf644
PMID: 41324558

---

What's New in v1.2

Feature	Details
GSEA enrichment	Gene Set Enrichment Analysis from any output — volcano, heatmap, Venn/UpSet — joining ORA across 100+ databases
Regression modeling	ML and MLP for continuous targets (R², RMSE, residual plots, cross-validation)
Longitudinal analysis	Mixed-effects models, trajectory plots, repeated-measures stats (Subject_ID + Time)
HTML report generation	One-click self-contained export of all session plots, tables, and metrics
Clinical metadata support	any column ending with _meta is treated as clinical metadata; available as alternative classification/regression targets, and used to colour heatmap annotations, PCA/UMAP points, and model training
Extended format support	auto-detection extended to Spectronaut, FragPipe, DESeq2/edgeR, Salmon, kallisto, MetaboAnalyst, XCMS, MZmine...
New graphical interface	Redesigned UI with improved layout, cleaner navigation and a live pipeline status indicator updated at each step so you always know where you are in the workflow

• Extended format support — auto-detection extended to Spectronaut, FragPipe, DESeq2/edgeR, Salmon, kallisto, MetaboAnalyst, XCMS, MZmine... |

---

Features

Raw Data Conversion

Convert raw mass spectrometry files from major vendors directly in Profiler's sidebar — no external tool (e.g. MSConvert) required.

Supported: Bruker .d · Waters .raw · Thermo Fisher .raw → .mzML · .mzXML · .mzDB · .mz5

---

Multi-Omics Support

Load and analyse data from any omics layer — all in one platform, no format conversion needed.

Omics type	Supported parsers / formats
Proteomics	MaxQuant proteinGroups.txt, DIA-NN pg_matrix.tsv, Spectronaut, FragPipe, PEAKS, Perseus, Proteome Discoverer, Progenesis
Metabolomics	MetaboAnalyst, XCMS, MZmine, generic CSV (m/z + retention time)
Lipidomics	Generic CSV, LipidSearch output
Transcriptomics	DESeq2, edgeR, Salmon, kallisto, featureCounts, STAR, HTSeq
Mass Spectrometry Imaging	MSI2Profiler CSV output (from imzML, MALDI-MSI, DESI-MSI)
Generic	Any CSV / TSV / XLSX with a Class column

Auto-detection of format, sample columns and software origin on upload. The _meta column system allows embedding clinical or batch metadata directly in the data file.

---

Data Lab — QC & Exploration

Instant dataset overview before any analysis:

• Missing value heatmap & per-sample % report
• Feature distribution plots & CV analysis
• Isolation Forest outlier detection
• Class balance visualisation (SMOTE / ADASYN ready)
• Sample rename, edit & metadata management (_meta columns)

---

Preprocessing Pipeline

A complete, ordered preprocessing workflow with data-driven auto-suggestions:

• Imputation: KNN, median, min/2
• Normalisation: log₂, Z-score, quantile, robust
• Batch correction: ComBat (NeuroCombat)
• Variance filtering: configurable threshold
• Class balancing: SMOTE, ADASYN
• Post-QC validation dashboard

---

Data Visualisation

Every plot is fully interactive (Plotly) — zoom, pan, hover, lasso selection, export at publication resolution (PNG · SVG · PDF @2×).

• PCA · UMAP · t-SNE with class overlays
• Hierarchical clustering heatmap (Ward, complete, average…)
• Correlation matrix & cosine similarity heatmap
• Violin, boxplot, density distributions
• Signal profile (multi-feature line plot)

---

Differential Analysis & Biomarker Discovery

• Volcano plot — binary and multi-class, interactive, configurable thresholds
• Statistical tests: t-test, Wilcoxon, Mann-Whitney, ANOVA, Kruskal-Wallis
• FDR correction: Benjamini-Hochberg, Bonferroni, Holm
• Heatmap of significant features with clustering
• Venn & UpSet plots — exclusive and shared feature sets across conditions
• Feature boxplots per group with significance bars
• Direct export of significant feature lists → enrichment module

---

AI Modeling — Classification & Regression (v1.2)

Clustering:

• K-Means to detect new groups
• Silhouette criterion to find the optimal number of clusters, avoiding over- or under-segmentation

Classification:

• Random Forest, XGBoost, SVM, Gradient Boosting, k-NN...
• Logistic Regression, PLS-DA...
• Cross-validation, model comparaison and optimal model, ROC curves, confusion matrices...

Regression (new in v1.2):

• Linear Regression, Random Forest Regressor, XGBoost Regressor, MLP
• R², RMSE, MAE metrics; residual plots; cross-validation

Explainability:

• SHAP: beeswarm, bar, force plots (sample-level & global)
• LIME: sample-level and global feature importance

Deployment:

• Export trained models as .pkl
• Real-time prediction on new unseen samples

---

ORA + GSEA Pathway Enrichment (v1.2)

• GSEA (new in v1.2) — ranked gene set enrichment from any Profiler output
• ORA — over-representation analysis
• 100+ databases: GO BP · MF · CC, Reactome, KEGG, WikiPathways, DrugBank, MSigDB, DisGeNET and more (via gseapy)
• Auto-import from volcano plots, heatmap clusters, Venn/UpSet exclusive sets, or manual paste
• Visualisations: bar plot, dot plot, heatmap, gene–pathway network

---

Survival Analysis

• Kaplan–Meier estimator per group with confidence intervals
• Log-rank test with p-value annotation
• Cox proportional hazards model with forest plot of hazard ratios
• Risk stratification from continuous features
• Survival table & at-risk annotations

---

Longitudinal Analysis (v1.2)

Dedicated module for repeated-measures and time-series omics data.
Load a dataset with Subject_ID and Time columns to:

• Visualise molecular trajectories per feature and per subject
• Compare group-level dynamics with confidence intervals
• Run repeated-measures ANOVA and mixed-effects models
• Perform time-point pairwise comparisons
• Compatible with all omics types

---

HTML Report Generation (v1.2)

Generate a complete, self-contained HTML report at any point in your session:

• All interactive Plotly figures embedded
• Statistical tables, model metrics, enrichment results
• Auto-generated table of contents
• Works offline — no server, no dependencies
• Timestamped and branded with Profiler + PRISM

---

Test Datasets

All test datasets are open on GitHub: yanisZirem/Profiler_v1_requests_datatests

Profiler_v1_requests_datatests/
│
├── MaxQuant_data/                    # proteinGroups.txt — LFQ, auto-parsed
├── DIA-NN_data/                      # report.pg_matrix.tsv — DIA proteomics
├── Bruker_data/                      # Raw .d folders → convert in Profiler sidebar
├── Waters_data/                      # Raw .raw folders → convert in Profiler sidebar
│
├── Tabular_data_multi_omics/
│   ├── Binary_classes/               # Aggressive vs NonAggressive (4 omics types)
│   │   ├── toy_proteomics_tumor_aggressiveness.csv
│   │   ├── toy_metabolomics_tumor_aggressiveness.csv
│   │   ├── toy_lipidomics_tumor_aggressiveness.csv
│   │   └── toy_rnaseq_tumor_aggressiveness.csv
│   └── Multi_classes/                # Tumor vs Necrosis vs Healthy (4 omics types)
│       ├── toy_proteomics_tumor_necrosis_healthy.csv
│       ├── toy_metabolomics_tumor_necrosis_healthy.csv
│       ├── toy_lipidomics_tumor_necrosis_healthy.csv
│       └── toy_rnaseq_tumor_necrosis_healthy.csv
│
├── Survival_data/
│   ├── clinical_and_LipidsMarkers.csv   # Clinical variables + lipid markers (Cox model)
│   └── statuts_patients.csv             # Pre-processed data for Kaplan–Meier analysis
│
└── data_for_peerReview_paper/        # Exact datasets used in Bioinformatics 2025 figures

Tabular CSVs (Binary & Multi-class): upload directly into Profiler — the Class column is auto-detected.
Raw instrument data (Bruker / Waters): convert to .mzML / .mzXML using the Data Conversion tab in Profiler's sidebar — no external tool needed.

---

Additional Tool: MSI2Profiler

MSI2Profiler is a companion desktop tool for preprocessing Mass Spectrometry Imaging data.

• Load .imzML files from MALDI-MSI and DESI-MSI experiments
• Normalise spectra (TIC, Median, RMS), bin m/z features, concatenate ROIs
• Export a Profiler-ready CSV matrix for immediate statistical analysis

Download directly from the Profiler homepage.

python MSI2profiler.py
# Dependencies: pip install pandas numpy plotly pyimzml

Full instructions: MSI2Profiler README

---

Who Should Use Profiler?

• 🔬 Researchers needing reproducible, end-to-end omics workflows
• 🧑‍⚕️ Clinicians exploring biomarkers and survival outcomes
• 🎓 Students & bioinformaticians learning omics data science methods
• 🏛️ Core facilities seeking robust, shareable analytical pipelines

---

Getting Started

1. Go to https://prism-profiler.univ-lille.fr/
2. Upload a dataset from the test repository or your own data
3. Follow the pipeline: QC → Preprocessing → Visualisation → Modeling → Enrichment → HTML Report

✅ No installation — fully browser-based
✅ No account required
✅ Free and open access

---

Authors & Contact

Yanis Zirem 📧 yanis.zirem@univ-lille.fr

Supervised by:

• Prof. Michel Salzet — michel.salzet@univ-lille.fr
• Prof. Isabelle Fournier — isabelle.fournier@univ-lille.fr

PRISM U1192 Laboratory — Protéomique, Réponse Inflammatoire, Spectrométrie de Masse
INSERM — Université de Lille

Protected by INSERM Transfert — APP/IDDN.FR2.0013.0300044.0005.S6.C7.20258.0009.312301

---

Profiler — Empowering researchers to transform omics data into discovery.