InsilicoΣ
Drug Discovery, Cheminformatics & Bioinformatics
About Our Team Publications Mobile App Contact Us Login Register
Beta - under development

Sigma-AI

The in-platform AI assistant integrated across every InsilicoΣ tool. Helps researchers configure runs, interpret results, and extract structured data from the literature - while keeping your data yours by default.

Try Sigma-AI Read the privacy section

Three integrated surfaces

Sigma-AI shows up in three places across the platform, each tuned to a specific workflow:

Floating chat widget
Top-right corner of every tool page. Each tool has its own persona, brand colour, and conversation history. Ask questions about the form you're filling in or the result you're reading.
Extraction side-panel
For evidence-synthesis tools (RoB-Sigma, GRADE-Sigma, MetaPrep). Upload a paper PDF; the panel returns one card per field with confidence + verbatim source quote + accept / edit / skip controls.
Result narratives
Short plain-language summaries of finished tool runs. Strict numeric grounding: every number quoted in the narrative must appear in the input metrics.

Each tool has its own personality

Sigma-AI runs different personas for different tools, so QSAR-X help is tuned for medicinal chemists while RoB-Sigma extraction is tuned for systematic-review methodologists. Conversation history is kept per-tool and per-user so switching between tools never mixes contexts.

Currently supported: QSAR-X, irQSAR, ADMET-X, APPAS, REINVENT4, NEXUS, POLY-X, CRAFT, RNA-Sigma, RoB-Sigma, GRADE-Sigma, MetaPrep, Clinical ML, FORMULA-X, BoltzGen, ESMFold, DecompDiff, plus the platform-wide Teach Corner tutor.

Privacy-first by design

Sigma-AI is built around opt-in, not opt-out:

  • Training on your data is OFF by default. We will never use your accepted extractions or chat messages to fine-tune any model unless you explicitly toggle it on in your AI Settings.
  • Auto-save requires both a platform-wide switch and your personal opt-in to fire. You can review and revert any auto-saved value with full audit trail.
  • Audit retention is user-configurable between 7 and 365 days; older Sigma-AI logs are deleted automatically.
  • Admin actions are visible to you. Any time an administrator edits your settings, your AI Settings page shows who edited and when.
  • Self-hosted inference. The current beta routes to a self-hosted OpenAI-compatible endpoint inside the InsilicoΣ infrastructure - your data is not forwarded to any third-party AI vendor.

Read the full privacy policy and the terms of service for the complete contract.

Current status

Sigma-AI is in beta. The infrastructure is live in production - persona registry, audit log, accept / reject / revert, review-autosaved page, AI Settings, sitemap, terms - but the underlying language model is not yet wired. While in this state, every chat reply is the deterministic placeholder "Sigma-AI is still under development. A full LLM model will be implemented soon." No real LLM call leaves your account during beta.

The full model will be wired without changing the user-facing contract. Subscribe via the dashboard for an in-app announcement when chat goes live for your tools.

✨
AI Lab