Turn proteomic complexity into actionable target hypotheses.
Druggx.ai connects DIA mass-spectrometry analysis, protein networks, causal reasoning and structural intelligence in one explainable discovery workflow.
Currently building the first MVP with public proteomics datasets.
Proteomics generates evidence. Turning it into decisions remains difficult.
Modern mass spectrometry can measure thousands of peptides and proteins, but the analysis remains fragmented across specialist tools, databases and manual interpretation. Researchers must connect spectral evidence, quantitative changes, biological pathways, causal relationships and structural information before identifying promising targets.
Fragmented analysis tools
Search engines, statistics packages, network databases and structure predictors each speak a different language.
Difficult biological interpretation
Lists of differentially abundant proteins rarely explain which mechanisms actually matter.
Limited traceability between evidence and conclusions
Once results are summarized, the path back to spectra, peptides and modifications is often lost.
Slow therapeutic-target prioritization
Ranking candidates by quantitative, network and structural evidence is still largely manual.
One traceable workflow from spectra to targets
Druggx.ai is designed to connect established proteomics engines with specialized AI models and biological knowledge sources. Every target hypothesis remains linked to the evidence and transformations that produced it.
- 01
Spectral processing
Convert, normalize and quality-check mass-spectrometry data in open formats such as mzML.
Technologies
- mzML
- ProteoWizard
- DIA-NN
- FragPipe/MSFragger
- 02
Identification and quantification
Identify peptides, proteins and post-translational modifications while controlling false discoveries and measuring abundance across samples.
Technologies
- DIA-NN
- MSFragger
- Casanovo
- Spectral libraries
- FASTA and UniProt
- 03
Differential biology
Compare biological conditions and identify proteins, proteoforms and pathways associated with meaningful changes.
Technologies
- MSstats
- limma
- Reactome
- Pathway enrichment
- 04
Network and causal intelligence
Build biological networks and rank candidate drivers using graph attention and causal analysis.
Technologies
- STRING
- SIGNOR
- GAT/GATv2
- PyTorch Geometric
- Causal inference
- 05
Structural druggability
Evaluate selected proteins using predicted structures, binding-pocket analysis and evidence-based druggability scoring.
Technologies
- AlphaFold/ColabFold
- P2Rank
- fpocket
- Open Targets
- ChEMBL
The final output is a ranked and explainable set of research hypotheses—not a clinical diagnosis or an experimentally confirmed drug target.
Designed for explainable target discovery
These design principles guide the platform. They describe where Druggx.ai is heading, not features that are already complete.
Evidence lineage
Trace every result back from the final target score to spectra, peptides, PTMs, quantitative evidence and biological sources.
Uncertainty propagation
Preserve confidence information across the pipeline instead of reducing every intermediate result to a simple yes-or-no decision.
Proteoform-level reasoning
Analyze biologically relevant protein forms and post-translational modifications, not only generic protein identifiers.
Human-in-the-loop AI
Keep researchers in control with inspectable evidence, configurable thresholds and documented analytical decisions.
A scientific copilot for connected biological evidence
Druggx.ai is exploring a Gemini-powered scientific interaction layer that can help researchers navigate pipeline results, biological databases and supporting literature through natural-language questions.
- Answers reference pipeline results and databases, not free-floating claims.
- No diagnostic, prognostic or treatment recommendations.
- Every interpretation stays attributable to inspectable evidence.
AI-generated interpretations must be reviewed by qualified researchers and linked to verifiable sources.
Example questions
- Why was this protein ranked as a potential driver?
- Which PTMs support this hypothesis?
- Which pathways connect this protein to the observed phenotype?
- What evidence would be needed to validate this target?
Draft answer, linked to sources
This candidate ranks highly because its phosphorylated proteoform increases consistently across replicates, it sits upstream of three enriched pathway members in the interaction network, and its predicted structure exposes a scorable pocket. Each statement links to the supporting evidence.
- Quantitative evidence
- PTM sites
- Network neighbors
- Pathway
- Structure
Built for teams working at the intersection of biology and data
Biotechnology companies
Accelerate early-stage target discovery and evidence review.
Proteomics laboratories
Transform analytical outputs into structured biological insights.
Pharmaceutical research teams
Prioritize candidates using quantitative, network and structural evidence.
Academic research groups
Build reproducible workflows around public and experimental datasets.
Cloud-native scientific computing
The platform is designed around containerized and reproducible workflows, with scalable CPU and GPU resources activated according to each analytical stage.
Google Cloud Storage
Object storage for proteomics datasets.
Compute Engine and Cloud Batch
Reproducible, containerized analysis jobs.
L4 and A100 GPUs
Specialized AI workloads, activated per stage.
Vertex AI
Model management and training pipelines.
Gemini
The scientific interaction layer.
Cloud Run
Application services and APIs.
IAM, Secret Manager and audit logging
Security, access control and traceability.
Development roadmap
The roadmap below describes planned work. Phases and features are subject to change and are not available today unless stated otherwise.
- Phase 1In progress
Proteomics MVP
- mzML ingestion
- DIA-NN/MSFragger integration
- peptide and protein results
- quality metrics and basic reporting
- Phase 2Planned
Biological intelligence
- differential analysis
- STRING and Reactome integration
- pathway analysis
- graph-based ranking
- Phase 3Planned
Target discovery
- causal evidence
- protein-structure integration
- binding-pocket analysis
- explainable druggability scoring
- Phase 4Planned
Scientific copilot
- natural-language exploration
- evidence-linked explanations
- experiment-planning assistance
- collaborative research workspace
Help shape the next generation of explainable proteomics
We are looking for proteomics laboratories, biotech teams, scientific advisors and early-stage partners interested in validating the Druggx.ai workflow.