Executable
From molecular measurements to runnable disease dynamics
A Computational Operating System for Biology
LS-OS
Living Systems OS is a research and engineering program to make biology executable, intervention-aware, and reproducible.
Built by Kalman Labs. Current focus: oncology kernel. Long-term goal: reusable biological runtime.
Perturbation-Aware
Simulate response under dose, schedule, and treatment changes
Reproducible
Versioned models, explicit assumptions, and auditable outputs
Core Thesis
Biology Should Be Run, Not Only Described.
Current State
- High-dimensional data, low executable modeling capacity
- Static models struggle under intervention shifts
- Limited causal simulation for treatment planning
LS-OS Direction
- Hybrid mechanistic-learning engines with explicit dynamics
- Counterfactual simulation under intervention schedules
- Reusable runtime interfaces across disease settings
Layer 1 MVP
Executable Tumor Response Dynamics in EGFR-mutant NSCLC
Start narrow, solve deeply. The first kernel of LS-OS is a hybrid model that maps RNA-seq to pathway state, pathway state to ODE parameters, and ODE simulation to resistance timing under EGFR TKI perturbations.
Input
Bulk RNA-seq and curated clinical progression outcomes
Abstraction
Pathway activity representation for intervention-relevant state
Dynamics
Constrained parameter net feeding differentiable tumor ODEs
Output
Trajectory behavior and resistance timing forecasts
Execution Strategy
Two Tracks, One Company Direction
Track 1: Current Work (Now)
Executable EGFR-mutant NSCLC response dynamics under EGFR TKI perturbations.
- Publishable oncology benchmark kernel
- Pilot-ready outputs for translational teams
- Near-term pharma collaboration surface
Track 2: Foundational Kernel (Next)
Perturbation-driven biological state-transition modeling for cross-disease reuse.
- Intervention-proximal dynamics over time
- Generalizable runtime abstractions
- Platform path beyond one disease
Company Build Stack
Four Layers, One Direction
Disease Instance
EGFR NSCLC executable model as first verified kernel.
Reusable Framework
Hybrid ODE + ML architecture reusable across cohorts.
Runtime Engine
Standard interfaces for state, intervention, and simulation.
Platform Layer
Programmatic integration for therapy design and trial strategy.
Path Ahead
From EGFR Kernel to Biological Runtime
Phase A: Prove Track 1
Complete and publish the EGFR kernel with rigorous validation and pilot-facing outputs.
Phase B: Launch Track 2
Build perturbation state-transition kernel and demonstrate transferability across settings.
Phase C: Runtime Layer
Unify both kernels under stable interfaces for simulation, intervention, and deployment.
Collaborate in Research
Work With Kalman Labs on LS-OS
Now: Oncology Pilots
Joint retrospective EGFR-NSCLC studies for response and resistance trajectory validation.
Next: Perturbation Data
Partnerships with pre/post-treatment transcriptomics or perturbation time-course datasets.
Co-Development
Collaborate on endpoint design, runtime interfaces, and external validation workflows.