Welcome Back
Quick Actions
Recently Viewed
Modules
Cancer Digital Twin Dashboard
- Home
- Dashboard
Population Explorer
Configure filters to explore the patient cohort in a UMAP-projected latent space. Adjust age boundaries and select molecular subtypes to focus on specific subgroups.
Step 1 — Age & Demographics
Step 2 — Molecular Subtypes
Select which molecular subtypes to include in the visualization. Each subtype will appear as a distinct cluster in the UMAP projection.
Patient Digital Twin
Build a comprehensive digital twin for an individual patient spanning 28 clinical dimensions — 3D anatomy, molecular profiling, genomic drivers, risk models, clonal evolution, and treatment optimization.
Step 1 — Patient Selection
Choose a patient from the cohort. The twin will be constructed from their full multimodal record including genomics, imaging, pathology, and clinical history.
Step 2 — View Configuration
Select which categories of analysis to prioritize. All 28 sections will be available, but your selections appear first.
Step 3 — Analysis Settings
Fine-tune prediction settings. The defaults work well for most people — just click “View Digital Twin” to continue.
Patient Digital Twin
A complete health profile for this patient — risk assessment, tumor details, treatment options, and personalized predictions all in one place.
I
Risk Overview
Your overall risk profile, 3D body map, and early detection opportunities
Sections 1–3
▾
Risk Overview
Your overall risk profile, 3D body map, and early detection opportunities
What this shows: An overview of your risk across 8 health dimensions, a score showing how early the cancer could have been caught, a 3D body model showing where it might spread, patient-specific tumor reconstruction for surgical planning, and a mechanistic simulation of how the tumor grows and metabolizes.
Technical details
The risk radar encodes 8 latent features from a variational autoencoder. The detection opportunity score quantifies probability of earlier detection given optimal screening. The 3D model maps metastatic spread probabilities based on TNM staging, molecular subtype, and lymphovascular invasion status. Image-guided 3D reconstruction uses multi-modal imaging fusion (MRI/CT/PET) for patient-specific tumor geometry. Mechanistic modeling incorporates Gompertzian growth kinetics, Warburg metabolic flux, and cell-cycle compartment dynamics.
Multi-dimensional risk assessment synthesizing genomic, clinical, and demographic factors. The radar chart maps 8 latent dimensions, the gauge shows overall detection opportunity score, and the bar chart decomposes risk by biological pathway.
Interactive 3D human body model with metastatic pathway visualization. Right panel shows detection opportunity windows and spread probability estimates from the digital twin's latent state encoding.
3D Human Anatomy — Metastatic Spread Map
Interactive 3D body model inspired by Human3D body-part segmentation (Takmaz et al., ICCV 2023). Drag to rotate, scroll to zoom. Organ markers sized by metastatic probability.
High-fidelity, patient-specific 3D reconstruction of tumor morphology and surrounding organs for surgical planning and radiotherapy dosimetry. Mechanistic tumor model incorporating Gompertzian growth kinetics, Warburg-effect metabolic flux, and cell-cycle compartment dynamics — enabling the digital twin to behave like a real tumor rather than just identifying statistical patterns.
Image-Guided 3D Tumor Reconstruction
Patient-specific 3D tumor surface reconstructed from multi-modal imaging (MRI T1/T2, CT, PET-SUV). Color encodes tissue density; surrounding margin zones guide surgical planning and radiation field definition.
II
Tumor Biology
Cancer type classification, tumor characteristics, and genetic mutations driving growth
Sections 4–6
▾
Tumor Biology
Cancer type classification, tumor characteristics, and genetic mutations driving growth
What this shows: The type of cancer (there are 5 main subtypes, each treated differently), how aggressive the tumor cells look under a microscope, and which genetic mutations are driving the cancer’s growth.
Technical details
PAM50 gene expression classifier assigns intrinsic subtypes. Histological grade follows the Nottingham (Elston-Ellis) system. Imaging classification includes BI-RADS, ACR density, and pN staging. Genomic landscape maps 18 driver genes and signaling pathways (PI3K/AKT/mTOR, RAS/MAPK, Wnt, DDR, Cell Cycle).
PAM50-based molecular subtype classification, histological grade distribution (Nottingham grading), and biomarker expression panel (ER/PR/HER2/Ki-67). These determine treatment eligibility and prognosis stratification.
BI-RADS assessment category, mammographic breast density (ACR classification), and pathological lymph node staging (pN). Density is the single strongest imaging risk factor; BI-RADS guides management decisions.
Key driver gene mutations and dysregulated signaling pathways identified from whole-exome sequencing. Pathway enrichment analysis reveals which oncogenic cascades (PI3K/AKT/mTOR, RAS/MAPK, Wnt, Cell Cycle, DDR, etc.) are most active in this tumor.
III
Evidence & Research
How your case connects to medical research, global data, and treatment guidelines
Sections 7–11
▾
Evidence & Research
How your case connects to medical research, global data, and treatment guidelines
What this shows: How your specific case connects to medical knowledge — including a visual map of related genes, drugs, and pathways, worldwide cancer statistics, similar patients in research databases, and recommended treatment guidelines.
Technical details
Force-directed knowledge graph maps BI-RADS features, molecular biomarkers, signaling pathways, driver genes, and therapeutic targets. GLOBOCAN 2022 epidemiology data provides global context. AJCC 8th-edition TNM staging and NCCN-aligned treatment protocols included.
Force-directed network mapping BI-RADS mammographic features, molecular biomarkers, oncogenic pathways, driver genes, and therapeutic targets. Based on NER-KG breast cancer knowledge graph ontology. Patient-specific nodes are highlighted with pulsing rings.
Orthographic globe projection showing age-standardised breast cancer incidence (per 100k) and BRCA1/2 mutation carrier prevalence by WHO region. Data sourced from GLOBOCAN 2022. Regional statistics aggregated by WHO classification with top/bottom country rankings.
Similar patient cohort analysis, relevant clinical trial summaries from published literature, and an AI-generated clinical narrative synthesizing all available data into a personalized assessment.
Complete BI-RADS feature matrix (mass shape, margins, calcification morphology, enhancement patterns) and AJCC 8th edition TNM staging breakdown with prognostic stage grouping and 5-year survival estimates.
NCCN-aligned treatment recommendations based on stage, subtype, and biomarker status. Includes neoadjuvant/adjuvant sequencing, radiation planning, endocrine therapy duration, and 10-year follow-up surveillance schedule.
IV
Detailed Lab Results
100+ lab values, immune cell analysis, and clinically validated risk scores
Sections 12–14
▾
Detailed Lab Results
100+ lab values, immune cell analysis, and clinically validated risk scores
What this shows: Over 100 lab measurements from the tumor, including how many mutations it has, what immune cells surround it, and five well-known risk scores (like Oncotype DX and MammaPrint) shown with easy-to-read green/yellow/red zones.
Technical details
Biomarkers derived from WES, RNA-seq, and methylation arrays including TMB, MSI, HRD. CIBERSORTx immune deconvolution. Five validated risk scores: Gail, Tyrer-Cuzick, Oncotype DX, MammaPrint, and Prosigna with model-specific thresholds.
Comprehensive computational biomarker panel derived from whole-exome sequencing, RNA-seq expression profiling, and methylation arrays. Includes tumor mutational burden, microsatellite instability, homologous recombination deficiency, and 17 additional genomic features interpreted by ensemble ML models.
Somatic mutation variant allele frequencies across 18 driver genes (top-left), CIBERSORTx immune cell deconvolution showing tumor-infiltrating immune composition (top-right), CpG methylation beta values at differentially methylated probes (bottom-left), and chromosome-level copy number variation log2 ratios (bottom-right).
Five clinically validated risk stratification models: Gail (NCI 5-year absolute risk), Tyrer-Cuzick/IBIS (10-year incorporating family pedigree + density), Oncotype DX (21-gene recurrence score), MammaPrint (70-gene prognosis), and Prosigna/PAM50 (risk of recurrence). Green/yellow/red zones indicate low/intermediate/high risk thresholds per model-specific cutoffs.
V
Survival Outlook
Survival predictions, how your case compares to similar patients, and immune health indicators
Sections 15–17
▾
Survival Outlook
Survival predictions, how your case compares to similar patients, and immune health indicators
What this shows: Charts projecting survival over 10 years, broken down by cancer-specific vs. other health risks. Also shows how your immune system is responding and how your case compares to national averages for your age, weight, and background.
Technical details
Kaplan-Meier estimates for OS, DFS, and DMFS. Competing risk analysis separates cancer-specific from cardiovascular/other mortality. Time-varying hazard function included. Twenty immune metrics via TIL scoring, flow cytometry, and serum biomarkers. SEER population benchmarking by age, BMI, race/ethnicity, and stage.
Kaplan-Meier survival estimates for overall survival (OS), disease-free survival (DFS), and distant metastasis-free survival (DMFS). Competing risk cumulative incidence separates breast cancer mortality from cardiovascular and other causes. Time-varying hazard function captures changing risk dynamics over the treatment timeline.
20 immune and tumor microenvironment metrics computed from digital pathology H&E analysis (TIL scoring), flow cytometry (CD8+/CD4+/Treg quantification), and serum inflammatory biomarkers (CRP, IL-6, NLR). Immunoscore classification per Galon framework.
This patient's profile benchmarked against population data: SEER age-specific incidence rates, BMI-associated relative risk (postmenopausal Women's Health Initiative data), race/ethnicity incidence disparities, and stage distribution at diagnosis. Your patient's position is contextualized within these population references.
VI
Health Monitoring
24-month tracking of hormones, immune health, and metabolism with 20 key biomarkers
Sections 18–19
▾
Health Monitoring
24-month tracking of hormones, immune health, and metabolism with 20 key biomarkers
What this shows: How key health markers are expected to change over the next 24 months during treatment — including hormone levels, immune system strength, and metabolic health (blood sugar, weight). Plus a panel of 20 important blood test values.
Technical details
Hormonal axis tracks estradiol and progesterone for endocrine therapy assessment. Immune surveillance monitors TIL and NLR. Metabolic health follows HbA1c and BMI. Biomarker panel includes ER/PR/HER2/AR, estradiol, FSH, SHBG, HOMA-IR, IGF-1, leptin, adiponectin, vitamin D, folate, and lipid components.
24-month projected trajectories for three critical monitoring domains: hormonal axis (estradiol + progesterone tracking endocrine therapy effect), immune surveillance (TIL score + neutrophil-lymphocyte ratio detecting immune escape), and metabolic health (HbA1c + BMI monitoring treatment-related metabolic syndrome).
20 hormonal and metabolic biomarkers including receptor status (ER/PR/HER2/AR), circulating hormone levels (estradiol, FSH, SHBG), insulin resistance markers (HOMA-IR, HbA1c), growth factors (IGF-1), adipokines (leptin, adiponectin), micronutrients (vitamin D, folate), and lipid panel components interpreted through breast cancer risk models.
VII
Treatment Options
Which treatments are most likely to work, predicted side effects, and clinical scoring
Sections 20–22
▾
Treatment Options
Which treatments are most likely to work, predicted side effects, and clinical scoring
What this shows: How likely each treatment is to work for your specific tumor, predicted side effects ranked by severity, and 20 validated clinical scores that doctors use to guide treatment decisions.
Technical details
Predicted pCR rates across 5 chemotherapy regimens using tumor genomics, PD-L1 CPS, and GEP signatures. Endocrine therapy and checkpoint inhibitor ORR estimates. CTCAE-stratified toxicity prediction. Genomic visualizations include mutation waterfall, volcano plot, and expression heatmap. Twenty scoring systems including NPI, CTS5, PREDICT v3.0, RCB, and ctDNA MRD.
Predicted pathological complete response (pCR) rates across 5 chemotherapy regimens, endocrine therapy benefit estimates for 5 agents, checkpoint inhibitor objective response rates (ORR), and pharmacogenomics-based toxicity profiling stratified by CTCAE grade. Based on tumor genomics, PD-L1 CPS, and GEP signatures.
Mutation frequency waterfall across top 20 driver genes, differential expression volcano plot (log2FC vs -log10 p-value with significance thresholds), and gene expression heatmap of 12 prognostic genes (Oncotype DX panel subset) across 8 tumor samples clustered by molecular subtype.
20 validated clinical scoring systems and prognostic indices: Gail, Tyrer-Cuzick, BRCAPRO (carrier probability), Oncotype DX, MammaPrint, Prosigna/ROR, EndoPredict, NPI, CTS5 (late recurrence), PREDICT v3.0 (10yr OS), Adjuvant!, Magee Equation, RCB classification, ctDNA MRD status, ECOG performance, Charlson comorbidity, and ASA surgical fitness scores.
VIII
Lifestyle & Imaging
Imaging analysis, lifestyle impact on risk, health conditions, and environmental factors
Sections 23–26
▾
Lifestyle & Imaging
Imaging analysis, lifestyle impact on risk, health conditions, and environmental factors
What this shows: Detailed imaging analysis, how your daily habits (exercise, diet, alcohol, sleep) affect your cancer risk, the impact of other health conditions, and environmental exposure assessment. Includes a breakdown showing what contributes most to your overall risk.
Technical details
Radiomic features (GLCM texture, shape, histograms) via PyRadiomics. Twenty imaging metrics including BI-RADS, MRI dimensions, enhancement kinetics, DWI-ADC. Lifestyle dose-response analysis. Charlson-weighted comorbidity modeling. Environmental exposure scoring for 8 agents. Composite risk decomposition across 10 factor categories.
Quantitative radiomic features extracted from diagnostic imaging: GLCM/GLRLM texture descriptors (entropy, contrast, correlation, homogeneity), shape morphology radar (sphericity, compactness, spiculation, elongation), and intensity histogram distribution. These PyRadiomics-compliant features feed into the ML malignancy classifier.
20 imaging-derived quantitative metrics: BI-RADS assessment, MRI lesion dimensions, enhancement kinetics (Type I/II/III curves), DWI apparent diffusion coefficient, PET/CT SUVmax, volumetric breast density, contralateral asymmetry index, and deep learning malignancy probability score. Radiomic signature computed from 107 PyRadiomics features.
Dose-response relationships between modifiable risk factors and breast cancer relative risk: physical activity level (WCRF continuous update), dietary pattern (Mediterranean diet meta-analysis), alcohol consumption (Hamajima et al. collaborative reanalysis), and sleep duration (prospective cohort data). Personalized to this patient's current lifestyle profile.
Comorbidity impact radar integrating 10 chronic disease risk modifiers via Charlson-weighted interaction modeling. Environmental exposure assessment scoring 8 carcinogenic/endocrine-disrupting agents. Composite risk decomposition pie chart showing the relative contribution of genetic, hormonal, lifestyle, environmental, clinical, imaging, molecular, immune, metabolic, and demographic factors to overall breast cancer risk.
IX
Predictions & Insights
Future risk projections, treatment planning, and explanations of what drives your results
Sections 27–31
▾
Predictions & Insights
Future risk projections, treatment planning, and explanations of what drives your results
What this shows: Your predicted risk over time using 5 different models, how the tumor may evolve, which treatment strategy gives the best outcome for the cost, exactly which factors are raising or lowering your risk score, and a personalized treatment schedule.
Technical details
Multi-model risk projections combining Gail, Tyrer-Cuzick, BCSC, Claus, and ensemble ML via Bayesian model averaging. Clonal evolution phylogenetic structure identifying truncal vs branch-specific mutations. Treatment optimization on cost-effectiveness frontier with tornado sensitivity analysis. SHAP feature contribution decomposition and treatment timeline integration.
Projected 10-year cumulative risk trajectory from 5 validated statistical models (Gail, Tyrer-Cuzick, BCSC, Claus, and ensemble ML). Each model uses different input features and makes different independence assumptions. The “ensemble” combines predictions via Bayesian model averaging, weighting each by its calibration accuracy on the BCSC validation cohort. Confidence bands reflect parametric uncertainty in each model’s coefficients.
Inferred clonal architecture from multi-region or longitudinal sequencing. The clonal frequency plot shows the evolution of distinct subclones over time, with driver mutations annotated at acquisition events. The tumor phylogenetic tree depicts branching evolution including truncal (present in all cells) vs branch-specific (subclonal) mutations. Understanding clonal dynamics is critical for predicting therapy resistance — truncal mutations are best drug targets, while branch mutations drive heterogeneity-mediated escape.
Decision-analytic modeling comparing treatment strategies on the efficiency frontier. Each point represents a distinct treatment plan (surgery type + systemic therapy combination); the x-axis shows incremental cost (USD) relative to watchful waiting, the y-axis shows incremental quality-adjusted life-years (QALYs). Strategies on the frontier are the most cost-effective at their willingness-to-pay threshold. The tornado diagram identifies which model parameters most influence the incremental cost-effectiveness ratio (ICER).
Shapley Additive Explanations (SHAP) decompose the model's risk prediction into individual feature contributions. The waterfall chart shows how each clinical variable pushes the prediction up (red) or down (blue) from the base rate. The bar chart summarizes global feature importance by mean absolute SHAP value.
Personalized treatment schedule generated from the patient's stage, molecular subtype, receptor status, and current clinical guidelines (NCCN 2024). Each bar represents a therapeutic phase — surgery, chemotherapy, radiation, targeted therapy, endocrine therapy — with follow-up appointment markers. Hover for drug protocols and dosing details.
Imaging Console
Review mammography and MRI imaging data with saliency overlays highlighting regions of clinical interest. View BI-RADS classification, breast density assessment, and comprehensive screening summary.
Step 1 — Patient Selection
Select a patient to load their mammography (L-CC, L-MLO) and MRI imaging records. Saliency maps will be generated using gradient-weighted class activation.
Step 2 — Display Options
Configure how imaging data is displayed. Saliency overlays highlight model-identified regions of interest.
L-CC Mammography
L-MLO Mammography
MRI Slice
Pathology and Biology
Molecular subtype probabilities, oncogenic pathway importance scores, gene expression profiles, and key biomarker panel for the selected patient.
Step 1 — Patient Selection
Choose a patient to analyze. Pathology data includes molecular subtype classification, pathway activation scores, and expression profiling.
Step 2 — Analysis Scope
Select which analyses to include in the pathology report.
Recurrence Forecast
Kaplan-Meier survival curves with confidence intervals and time-varying hazard function. Includes similar-case matching for clinical context and comparison against population baselines.
Step 1 — Patient Selection
Select a patient to generate their personalized recurrence forecast using multi-model survival analysis.
Step 2 — Forecast Parameters
Configure the survival analysis model and comparison baseline.
Lifestyle Simulator
See how lifestyle changes could affect your breast cancer risk over time. Enter your details, pick a scenario, and run a simulation.
Step 1 — About You
Basic information about your age, body, and health status.
Step 2 — Genetics & Habits
Your genetic background, family history, and daily habits.
Step 3 — Run Settings
Choose a scenario and how far into the future to simulate.
About You
Genetics & Background
What-If Scenario
Day-by-Day Trajectory
Drill into the Lifestyle simulation at daily resolution. Zoom into specific time windows and explore individual biological subsystems to understand how risk evolves over time.
Step 1 — Time Window
A simulation must be run on the Lifestyle Simulator tab first. Then select a time window to zoom into for daily-resolution analysis.
Step 2 — Subsystem Focus
Choose a biological subsystem to highlight. The focused subsystem gets its own dedicated chart alongside the all-subsystems overview.
Scenario Comparison
Compare multiple intervention scenarios head-to-head. Quantify absolute risk reductions via waterfall analysis and reveal which biological subsystems benefit most from each intervention through radar comparison.
Step 1 — Scenario Selection
Select the intervention scenarios you want to compare. At least 2 scenarios are recommended for meaningful comparison.
Step 2 — Comparison Parameters
Configure the comparison horizon and statistical parameters.
Scenarios to Compare
Population Epidemiology ABM
Simulate up to 2 million heterogeneous agents through multi-decade breast cancer natural history. Each agent carries unique demographic, genetic, lifestyle, and screening traits drawn from real-world distributions (US Census, SEER, GLOBOCAN).
Step 1 — Population Size
Choose the number of agents to simulate. Larger populations produce smoother statistical distributions but require more computation time.
Step 2 — Demographics & Age Range
Define the age range of the simulated population. Agents will be drawn from US Census age distributions within this range.
Step 3 — Public Health Intervention
Select a public health intervention to apply across the simulated population. Results will show the population-level impact on incidence, mortality, and stage distribution.
Output Sections
The simulation produces 12 analytical sections: risk distribution, age-specific incidence, stage at diagnosis, mortality curves, genetics breakdown, lifestyle impact, screening effectiveness, temporal trends, density analysis, intervention comparison, survival analysis, and comprehensive summary statistics.
Distribution of 10-year absolute breast cancer risk across the entire simulated population. The histogram reveals the heterogeneity of individual risk profiles arising from the interplay of age, genetics, density, lifestyle, and family history. Percentile cutoffs (P5, P25, P50, P75, P95) identify low-risk, average, and high-risk strata for targeted intervention design.
Incidence per 100,000 women-years by decade-of-life age bins, calibrated against SEER registry data. The characteristic exponential rise from age 40, plateau around 70-79, and slight decline after 80 reflects the interplay of cumulative exposure, competing mortality, and hormonal changes at menopause. Population age pyramid shows the denominator at risk.
Incidence and mortality rates stratified by self-reported race/ethnicity. While White women have the highest incidence, Black women experience 42% higher mortality driven by higher rates of aggressive subtypes (triple-negative), later stage at diagnosis, and social determinants. Ashkenazi Jewish women carry elevated BRCA prevalence (1:40 vs 1:400 general population).
AJCC stage distribution of all diagnosed cancers. The proportion diagnosed at Stage 0 (DCIS) and Stage I directly reflects screening programme effectiveness. Screening interventions produce a leftward “stage shift” — diagnosing cancers earlier when they are curable. Stage IV at diagnosis (6%) represents metastatic disease with dramatically different prognosis.
PAM50 intrinsic subtype distribution across all diagnosed cases. Luminal A (40%) represents the most favorable biology. Triple-negative (15%) disproportionately affects younger and Black women, lacks targeted therapy, and requires chemotherapy. HER2-enriched (15%) is now treatable with anti-HER2 agents (trastuzumab) but historically carried poor prognosis.
Head-to-head comparison of screened vs. unscreened populations. Screening produces measurable stage shift (more Stage 0/I diagnoses), earlier mean age at detection, and improved 5-year survival. The “Number Needed to Screen” (NNS) quantifies how many women must be screened to detect one early-stage cancer that would otherwise have been diagnosed later.
Dose-response relationship between BMI and breast cancer incidence (by population decile). Breast density categories show a 4x gradient from fatty (A) to extremely dense (D). Dense tissue both increases biological risk and masks tumors on mammography, creating a double penalty that justifies supplemental MRI screening for Category D women.
Population-level impact of BRCA1/BRCA2 germline pathogenic variants. Carriers represent ~0.25% of the population but account for a disproportionate share of early-onset and bilateral cases. Cascade genetic testing (testing blood relatives of known carriers) is the most cost-effective cancer prevention strategy, with a Number Needed to Test (NNT) of ~20 to identify one additional carrier.
Year-by-year cumulative incidence showing the growing cancer burden across the simulation horizon. The curve shape reflects the population age structure (aging population = accelerating incidence) and any intervention effects (screening detects existing cancers earlier → transient incidence spike; prevention reduces slope). Annualized incidence rate provides the per-year event velocity.
Population Attributable Fraction (PAF) for each modifiable and non-modifiable risk factor. PAF estimates the proportion of total cancers that would be prevented if the factor were eliminated from the population. PAF depends on both the relative risk and the prevalence of the factor. Dense breasts, physical inactivity, and obesity typically dominate modifiable PAF.
GLOBOCAN-derived regional estimates of annual breast cancer cases and deaths worldwide. Sub-Saharan Africa shows the worst mortality-to-incidence ratio due to late presentation and limited treatment access. Screening rate heatmap highlights the global screening coverage gap. The intervention column shows projected cases preventable if the selected public health programme were deployed regionally.
Comprehensive tabular summary of all simulation outputs including population demographics, cancer burden metrics, screening programme performance, genetic testing yields, and intervention cost-effectiveness estimates. All rates reported per 100,000 women-years with 95% simulation intervals.
Agent Simulation (TinyTroupe)
Observe thousands of autonomous agents navigate a stochastic cancer state machine in real time. Each agent follows one of 16 archetypes with unique risk profiles. Visualize scatter dynamics, cluster formation, and knowledge graph emergence.
Step 1 — Agent Population
Set the number of agents and simulation duration. Each agent is assigned one of 16 archetypes (e.g., Young Low-Risk, BRCA1 Carrier, Inflammatory BC) with archetype-specific state transition probabilities.
Step 2 — Visualization
Configure the real-time visualization. Three modes are available: Scatter (individual agents as points), Clusters (archetype groupings), and Knowledge Graph (causal network).
Step 3 — Dashboard Sections
The post-simulation dashboard includes 8 analytical sections. All sections are generated automatically after the animation completes.
Final-state distribution across all 16 archetypes after 5-year simulation. Each row is an archetype; each column is a terminal state. Cell intensity shows the fraction of agents in that archetype reaching that outcome. BRCA carriers show dramatically higher progression to treatment and metastasis columns vs. low-risk archetypes.
Aggregate flow of agents through the state machine: healthy → screening/symptomatic → diagnosis → treatment → remission/recurrence → metastasis/death/survivor. Width of each flow band is proportional to the number of agents traversing that transition. Reveals bottlenecks and the dominant disease pathways.
Simulated Kaplan-Meier survival curves stratified by archetype. Each line tracks the proportion of agents still alive month-by-month. BRCA1 carriers and Inflammatory BC phenotypes show the steepest drops; Mediterranean Diet Adherents and Young Low-Risk agents maintain near-flat survival plateaus.
2D embedding of outcome scenario clusters. Each bubble represents a group of agents with the same archetype and final state. Bubble size ∝ cluster population; color = archetype. Proximity encodes similarity — nearby clusters share demographic and risk profiles. The knowledge graph overlay shows causal links between risk factors, diagnoses, treatments, and outcomes.
Full knowledge graph of the cancer simulation ontology. Nodes are concepts (risk factors, archetypes, diagnoses, treatments, outcomes). Edges encode causal and associative relationships (risk_factor_for, protective_against, treated_by, leads_to). Node size ∝ importance; edge thickness ∝ strength. Force-directed layout reveals structural clusters.
Month-by-month cumulative diagnosed cases and new monthly incident cases per 1,000 agents. Rising incidence curves reflect population aging and evolving risk. Steep early rises indicate high-risk archetypes reaching diagnosis quickly.
Distribution of 10-year risk scores across all simulated agents, colored by archetype. Box plot shows quartiles, outliers, and median risk per archetype. Reveals risk stratification across the population and identifies which archetypes drive the tails of the distribution.
Comprehensive statistical summary of the multi-agent simulation including total events, archetype-level breakdown, transition counts, and key epidemiological metrics.
Other Ideas
Experimental modules exploring clinical trial matching, genomic circos plots, drug interaction analysis, longitudinal biomarker tracking, and federated learning across hospital networks.
Step 1 — Module Selection
Choose which experimental modules to load. All five can be viewed simultaneously or individually.
Step 2 — Analysis Parameters
Configure analysis depth and display preferences for the selected modules.
Other Ideas — Experimental Modules
AI-powered matching of the patient's molecular profile, staging, and prior treatments against active clinical trials. Eligibility scores are computed from inclusion/exclusion criteria overlap. Green = eligible, yellow = partial match, red = excluded.
Circular genome visualization showing somatic mutations, copy number alterations, and structural variants across all chromosomes. Inner rings display variant allele frequency; outer ring shows gene annotations. Arcs indicate translocations and fusions.
Pairwise interaction matrix for the patient's oncology regimen plus supportive medications. Color intensity indicates severity: red = contraindicated, orange = major, yellow = moderate, green = minor/none. Click any cell for mechanism details.
Sparkline dashboard tracking key biomarkers (CA 15-3, CEA, Ki-67, ctDNA, CTC count, LDH) over 12 months. Trend arrows indicate direction. Reference ranges shown as shaded bands. Alerts fire when values cross clinical thresholds.
Visualization of a federated learning network where 8 hospital sites collaboratively train a cancer outcome model without sharing patient data. Shows per-site training loss convergence, global model accuracy, communication rounds, and differential privacy budget (epsilon) consumption.
OmniScan Digital Twin
Upload every piece of data you have — imaging, pathology, lab work, genomics, RNA sequencing — and we’ll build a complete digital twin of your cancer risk profile.
Step 1 — Clinical & Lab Data
Your basic clinical information and blood work results.
Step 2 — Imaging & Pathology
Upload mammograms, MRIs, ultrasound images, or pathology slides. We’ll analyze density, lesions, and tissue characteristics.
Step 3 — Genomics & Molecular
Upload sequencing data, gene panels, or enter receptor / mutation status manually. The more data, the more precise your digital twin.
Step 4 — Family, Lifestyle & Environment
Rounding out your twin with family history, lifestyle, and environmental exposures.
Risk Breakdown
Multi-Omic Data Integration
Every data source you provided has been analyzed and scored. Completeness shows how much data was available for each modality.
Genomic Landscape — Deep Mutation Analysis
Comprehensive somatic and germline mutation profiling across 60 cancer driver genes with variant annotations, allele frequencies, and clinical classifications.
Clonal Architecture & Tumor Evolution
Variant allele frequency distribution revealing clonal vs. subclonal populations, tumor mutational burden benchmarking, and clonal composition analysis.
HRD Score & Pharmacogenomics
Homologous recombination deficiency assessment and actionable therapeutic targets based on your genomic profile with clinical trial evidence.
Transcriptomic Profile
Imaging AI Analysis
Blood & Biomarker Panel
Digital Twin Body Map
Organ-level risk attribution: how each body system contributes to your overall cancer risk.
BRCA Family Pedigree
Three-generation family tree showing inherited mutation carrier status and cancer diagnoses. Filled shapes indicate affected individuals; half-filled indicates carrier without diagnosis.