The AI platform for
computational biology
From raw data to defensible results. Cortex executes rigorous, multi-method analysis with full traceability. Synapse connects every finding to scored evidence from the literature and 20+ public databases.
From question to insight: autonomous, multi-step, fully traceable
Upload your data, ask a question, and Cortex deploys a team of specialized AI agents, each an expert in a different analytical domain. They design a multi-step plan, execute each step with the right statistical method, and cross-validate findings against Synapse's evidence from hundreds of thousands of papers and 20+ public databases.
Every method is justified. Every finding is traceable. You stay in control at every checkpoint.
Read the paper and identify gaps in the authors' analysis. What can we focus on to uncover novel insights?
I now have a thorough picture of the paper and its analytical pipeline. Let me update working memory and compile the gap analysis.
Gap Analysis: Roelands et al. 2021 — What Was Left on the Table
Paper Summary
Roelands et al. characterized ancestry-associated transcriptomic differences in breast cancer across TCGA-BRCA (n=1,051; EA/AA/AsA) and a Qatar cohort (n=24; Arab/Asian). The central finding: African-American (AA) patients in the BasalMyo subtype have worse overall survival, lower immune cytolytic response (ICR), higher AMPK signaling, and reduced immunoediting — suggesting metabolic reprogramming as a driver of immune evasion. The analysis pipeline spans 87 R scripts covering ICR clustering, IMS subtyping, ssGSEA, survival analysis, mutational load, and an XGBoost prediction model.
Gap 1 — Survival Analysis Is Unadjusted
What they did: Kaplan-Meier curves and univariate Cox regression stratified by ancestry × IMS subtype. Stage-stratified forest plots were generated (C41), but the hardcoded HR values (HR=2.39 for all BasalMyo AA vs EA, p=0.020) come from a univariate model only.
What's missing: No multivariate Cox model adjusting simultaneously for tumor stage, age at diagnosis, and ICR cluster. The cBioPortal data contains AJCC stage, age, and mutation count for all BRCA samples. The stage distribution difference between AA and EA in BasalMyo (C54 script) is acknowledged but never statistically controlled. The survival disparity could be partially or fully explained by stage imbalance — this is unresolved.
What to do: Fit a multivariate Cox model: OS ~ ancestry + stage + age + ICR_cluster within BasalMyo. Test the ancestry × ICR interaction term. This directly tests whether the immune microenvironment mediates the ancestry–survival relationship.
This multi-step computational run investigates ancestry-associated immune heterogeneity in breast cancer, wit...
Tell Cortex what you need.
Share the result.
Chat with Cortex to build interactive reports from your analysis results. Describe the story you want to tell: survival curves, feature importance, mediation diagrams, and the AI assembles a polished, navigable report in minutes.
Every report is shareable with your team members or external collaborators.
Generated by Inflexa on February 20, 2026 at 9:00 PM
Survival & Prognosis
Total ancestry effect on survival is null
After adjusting for subtype, age, and stage, African ancestry does not confer a survival advantage or disadvantage (OS endpoint, n=957, 137 events). Breast cancer survival disparities are largely explained by clinical factors, not ancestry-specific molecular features.
SHAP Feature Importance: Ancestry-Specific Survival Models
Top 20 features by mean |SHAP| value for each ancestry-specific CoxPH model. Colors indicate feature group.
EA Model (C-index = 0.652)
Mean |SHAP|
AA Model (C-index = 0.408)
Mean |SHAP|
AA model is dominated by PI3K-Akt-mTOR (SHAP=0.510), while EA model spreads importance across TF and checkpoint features. NT5E (CD73) is the top EA feature — consistent with the adenosine immunosuppression pathway identified in the checkpoint analysis.
Feature Group SHAP Contributions
TF activity contributes 37\u201348% of SHAP importance across all models, peaking in BasalMyo (47.6%). This layer was entirely absent from the original paper.
Causal Mediation: Ancestry → Survival
Indirect Effect (Δ log-HR)
TReg is the sole significant mediator (indirect = −0.089, p=0.030) via suppression. Green = p<0.05, Yellow = p<0.10.
Everything known about your target. One dossier.
Synapse reads thousands of papers and cross-references 20+ public databases to build comprehensive target dossiers. Druggability scores, disease associations, drug interactions, safety signals, clinical trial landscapes, all structured, scored, and continuously updated.
Every data point traced to its source. Every claim scored for evidence strength.
GLP1R Target Assessment
GLP1R
HGNC:4324 · gene
Top indications: Hypertension, Coronary Artery Disease, Obesity +17 more
Unknown
Precedent: DULAGLUTIDE, LIRAGLUTIDE, EXENATIDE
+14 more
2 Class Liabilities
0/0 tissues > 1000 TPM
Drug Interactions (309)
The intelligence layer meets the analysis agent
Rigorous analysis
Runs multi-track, multi-step analysis on your omics, cheminformatics, and imaging data with full provenance. Every finding traced to source.
Structured intelligence
Synthesizes scored evidence from 20+ databases and the full scientific literature into target dossiers, disease landscapes, and safety profiles.
See what the reports look like
Produced on publicly available datasets. Full methods, statistical tests, and visualizations included.
Show us your data.
We’ll show you what’s possible.
Bring a dataset you care about. We’ll run a full analysis and walk you through the output.