TRuST is a transcriptomics-driven drug repurposing platform integrating differential expression analysis, LINCS L1000 perturbational signatures, and connectivity scoring approaches to identify candidate therapeutic compounds capable of reversing disease-associated transcriptional profiles.
Workflow
- 1 Import Data — Upload transcriptomic datasets (.csv /.tsv /.txt),or retrieve datasets directly from GEO.
- 2 Process — Perform differential expression analysis using limma, gene annotation, and LINCS signature matching.
- 3 Score — Compute connectivity scores using fgsea/CoreGx methodologies together with correlation-based prioritization.
- 4 Explore — Identify top-ranked therapeutic compounds capable of reversing or mimicking disease signatures.
- 5 Export — Download ranked candidate drugs, transcriptomic signatures, and analysis results.
Scientific Notes
- LINCS L1000 signatures are accessed through a chunked backend architecture, loading only the signature chunks required for each query rather than the full resource.
- Differential expression analysis is performed using limma with configurable statistical thresholds.
-
Connectivity scores are computed using
CoreGx::connectivityScorebased on fgsea enrichment methodologies. - Two-stage prioritization (correlation-based screening followed by enrichment-based confirmation) for robust therapeutic candidate ranking.
- Automatic phenotype inference supports many GEO studies, but datasets with uncommon metadata structures may require manual group annotation.
Input Requirements
- Supported input formats: GEO accession numbers, CSV, TSV and TXT files.
- Expression matrix format: genes × samples.
- Gene identifiers: row names should contain HGNC symbols, Affymetrix probe IDs or Entrez IDs.
- Metadata requirements: must contain either a sample column or GEO accession identifiers.
- Group information: A group column is recommended. Case/Control or Disease/Control labels are automatically detected when available. For cancer datasets, Tumor/Normal annotations are also recognized.
Transcriptomic Data Import
Load transcriptomic datasets from GEO or upload custom expression matrices.
GEO accession
Upload custom dataset
Expression matrix: genes in rows, samples in columns.
Metadata: either a sample column or a GEO-style geo_accession column.
Optional: group column (Case / Control, Disease / Control, or Tumor / Normal).
Expression matrix preview
Metadata preview
Transcriptomic Drug Repurposing Pipeline
Integrated transcriptomic analysis workflow for identifying candidate therapeutic compounds using differential expression profiling and LINCS perturbational signatures.
Differential Expression
Identification of significantly dysregulated genes associated with the disease phenotype using transcriptomic statistical analysis workflows.
LINCS Signature Matching
Comparison of disease-associated transcriptomic signatures against LINCS perturbational profiles to identify reversal candidates.
Drug Prioritization
Consensus ranking and prioritization of candidate therapeutic compounds based on connectivity and correlation metrics.
Large transcriptomic analyses may require several minutes depending on dataset size and computational complexity.
Interpretation Guide
- Negative connectivity scores suggest potential transcriptional reversal of the disease-associated gene signature.
- Negative Spearman correlations suggest transcriptomic reversal of the disease-associated signature and may indicate potential therapeutic activity.
- Positive Spearman correlations indicate transcriptomic similarity and may reflect disease-mimicking perturbations.
- Two-stage prioritization (correlation-based screening followed by enrichment-based confirmation) is used for robust candidate ranking.
- Top-ranked reversal compounds may represent promising therapeutic repurposing candidates for downstream validation.
Differentially Expressed Genes
Genes identified as significantly dysregulated based on limma differential expression analysis with FDR-adjusted statistical filtering.
Spearman Drug Scores
Correlation analysis between disease-associated transcriptomic signatures and LINCS perturbational profiles.
Connectivity Scores (fgsea)
Negative enrichment scores indicate potential reversal of disease-associated transcriptomic patterns.
Drug Repurposing Results
Candidate compounds are ranked using combined Spearman correlation and fgsea-based connectivity scoring.
Interactive Results Tables
Interpretation: Negative connectivity and negative correlation scores suggest transcriptional reversal of the disease signature and may indicate potential therapeutic repurposing candidates.
About TRuST
TRuST (Transcriptomic Repurposing Signature Tool) is an academic bioinformatics platform developed for transcriptomics-driven drug repurposing using differential expression analysis, LINCS L1000 perturbational signatures, and connectivity-based prioritization approaches.
The platform integrates rank-based correlation scoring and fgsea-based connectivity analysis to identify candidate therapeutic compounds capable of reversing disease-associated transcriptomic signatures.
TRuST supports both user-supplied expression datasets and public GEO studies. Automatic phenotype inference is available for many GEO datasets, although studies with highly customized metadata may require manual group annotation.
How to Cite TRuST
If you use TRuST in your research, please cite:
Software & Packages
TRuST relies on several open-source R/Bioconductor packages:
LINCS Backend Resource
TRuST queries a chunked LINCS L1000 signature backend, derived from Bioconductor ExperimentHub (EH3226) and partitioned into 90 indexed chunks for scalable, on-demand querying (12,328 genes; 45,956 signatures). This backend is archived independently on Zenodo, both for reproducibility and as an automatic recovery source if the locally installed backend is missing or incomplete.
Backend version 2.0.0 · 90 chunks · DOI: 10.5281/zenodo.21865289
LINCS / Connectivity Map Resource
TRuST integrates LINCS L1000 perturbational transcriptomic signatures and Connectivity Map methodologies to prioritize compounds capable of reversing disease-associated expression profiles.
Compound Name & Annotation Resolution
LINCS L1000 profiles many compounds using internal Broad Institute reference codes (e.g. "BRD-K12345678") rather than a public name. Where possible, TRuST resolves these codes to a common compound name, and enriches results with mechanism of action, clinical phase, and disease area, using the Broad Institute's Drug Repurposing Hub.
Coverage is necessarily partial: the Repurposing Hub curates roughly 7,500 well-characterized compounds, while LINCS L1000 profiles a much larger screening library. In practice, only a small minority of LINCS's BRD-coded compounds overlap with the Hub's curated set — most are proprietary or uncharacterized screening compounds with no public name anywhere. Unresolved codes are always shown as the original BRD identifier rather than hidden, preserving traceability back to the exact LINCS perturbagen.
Data portal: repo-hub.broadinstitute.org/repurposing
Research Team
Bioinformatics & Biostatistics research group — Institut Pasteur de Tunis
Contact & Collaboration
For scientific collaboration, bug reports, feature requests, or academic inquiries:
- Developed by: Ghada Mahjoub & Rihab Mahjoub
- Designed by: Kais Ghedira & Oussema Khamessi
- Email: trust.drugrepo@gmail.com
- Institution: Institut Pasteur de Tunis // ISBST // Manouba University
- Research Unit: BIMS Laboratory
- Platform status: Research prototype / academic platform