ProcessDSL + FlowCell-10 Proposal

This proposal outlines a pilot initiative to integrate the "genome as program" concept and cellular process flowcharting into the Virtual Cell project. The goal is to formalize biological processes as executable, interpretable programs that can be learned, simulated, and manipulated by AI.

1. ProcessDSL Specification

ProcessDSL is a domain-specific language for representing cellular processes. It compiles human-readable flowcharts into machine-executable forms such as stochastic rule systems, Petri nets, or hybrid ODE/event simulators.

Key features:

Reactions as rules with explicit guards and rate constants.
Conditional logic (IF/ELSE) for regulation.
Iterative loops (WHILE) for cyclic processes.
Event triggers for environmental or signaling changes.
Support for compartments (nucleus, cytosol, organelles).

2. FlowCell-10 Pilot Dataset

FlowCell-10 is a curated set of ten well-characterized yeast pathways, each represented as:

A canonical flowchart
A ProcessDSL file
Reference simulation outputs from literature data

Example pathways:

Glycolysis
TOR nutrient sensing pathway
Heat shock response
Autophagy initiation
Unfolded protein response (UPR)
Cell cycle G1/S transition
Mitochondrial respiration control
Amino acid biosynthesis regulation
Gluconeogenesis
Alcoholic fermentation

3. Example ProcessDSL (Glycolysis)

process Glycolysis in Cytosol:
  state: [Glucose, G6P, F6P, F16BP, G3P, DHAP, PEP, Pyruvate, ATP, ADP, NAD+, NADH]
  rule Hexokinase: Glucose + ATP -> G6P + ADP  [guard: ATP>θ1]
  rule PFK: F6P + ATP -> F16BP + ADP           [guard: ATP<θ2 & AMP>θ3]
  rule Aldolase: F16BP -> G3P + DHAP
  rule TPI: DHAP <-> G3P
  rule PyruvateKinase: PEP + ADP -> Pyruvate + ATP [allosteric: F16BP activates]
  event GlucosePulse(t=0..T): inflow rate r_in
        

4. Expanded Glycolysis Flowchart

Below is an example from FlowCell-10 showing Glycolysis in Yeast with branch and loop structure:

5. Deliverables

ProcessDSL specification and parser.
FlowCell-10 diagrams, DSL files, and simulation benchmarks.
Jupyter notebook demo: diagram → ProcessDSL → simulation → data comparison.
Documentation for extending the dataset.

6. Benefits to the Virtual Cell Project

Provides an interpretable, executable representation of cellular processes.
Bridges molecular prediction tools (e.g., AlphaFold 3) to systems-level dynamics.
Enables counterfactual simulations and intervention planning.
Creates training data for AI models to learn biological program induction.

This proposal demonstrates how computational biology and artificial intelligence can converge to create interpretable, executable models of biological systems.