RxAI | Drug Screen

Ready

Select the Machine Learning model for drug sensitivity prediction.

Workflow. Run a screen, select a cell line, inspect the model explanation, then send a curated gene set into unified evidence.
Cell line Awaiting selection
Disease Awaiting selection
Drug input Awaiting screen
Model used Awaiting run
Run a screen and select a cell line to populate the prediction summary and explanation workflow.
Current gene-set status No candidate genes selected yet.

Local Feature Importance

Top genes driving prediction for Select Cell

Select a cell line from the left to view analysis.
Purpose. Use these views to understand the model attribution, then decide which genes should enter downstream evidence analysis.
Selected gene set 0 genes selected
Selection mode: auto top SHAP until you choose genes manually.
Support view: no evidence run yet.
No genes selected yet.

Gene Impact Bubbles

Bubble position shows SHAP effect and expression. Select genes here to define the candidate set for evidence review.

Candidate gene set from model explanation Use manual bubble selection for expert curation. If nothing is selected, unified evidence will fall back to top SHAP-ranked genes.
Run Unified Evidence to overlay literature and pathway support on the chart.
Resistance-associated (SHAP > 0) Sensitivity-associated (SHAP < 0) Literature evidence Lit + Pathway Size = |SHAP| importance  ·  Click a bubble to select gene

Prediction Breakdown

Detailed positive and negative contributions to the predicted response.

Select a cell line from the left to view analysis.

Summary Dot Plot

Global feature impact distribution across the cohort.

Select a cell line and wait for analysis.

Gene Dependence

Use this to see where the selected cell sits in the cohort for a given gene.

Select a cell line from the left to view analysis.

Force Plot

Interactive visualization of the additive feature effects behind the prediction.

Select a cell line from the left to view analysis.

Biological Evidence Engine

Contextualize the selected gene set with literature, pathways, and cohort/sample evidence. Unified evidence is the primary path; literature-only controls remain as an advanced option.

Unified evidence run. This combines SHAP-based gene selection, pathway analysis, cohort grouping, and literature mining into one cited evidence session. Manual bubble selection overrides the auto-selected top SHAP genes.
0 labels loaded
Using top SHAP genes once an analysis is available.
0 selected Showing the highest-priority SHAP genes. Use the bubble chart to select a custom set.
Genes queried 0
Genes with papers 0
Genes with evidence 0
Evidence records 0
Top supported genes
Run mining to populate supported genes.
Run mining to populate evidence counts, retrieved papers, and hypothesis output.
Run unified evidence to view pathway enrichment and pathway warnings.
Run unified evidence to view sample focus and cohort group summaries.
Run mining to view retrieved papers.
Run mining to view extracted evidence.
Run mining to view ranked candidates.
Run mining to generate evidence-grounded hypotheses.
Outcome. This section should answer one question: why is this cell predicted to respond, and what evidence supports that claim?
Cell line Awaiting selection
Disease Awaiting selection
Drug input Awaiting screen
Model used Awaiting run
Run unified evidence to generate a final integrated hypothesis with pathway and literature support.
Run unified evidence to generate sample, cohort, and integrated hypotheses.
Report Builder

Bio-Assistant

Use the assistant to interrogate the current prediction, evidence, and report context.

Hello! I am your research assistant. Select a cell line and ask me to explain the results, check facts, or write a report.