hub:
  repo_type: generator
  builtin: gard
  tags:
  - banzhaf-yamamoto
chemistries:
- id: gard
  name: GARD (Graded Autocatalysis Replication Domain)
  intuition: 'Heredity without a sequence. An assembly of lipid types grows by absorbing lipids from outside,
    and lipids already inside catalyse the entry of particular types, so the assembly''s *composition*
    biases which lipids it takes up next. Split it in two and each half tends to reconstruct the same
    composition - a composome. The information is in the multiset, not in an ordered polymer. It also
    shows cleanly that catalysis alone cannot shift an equilibrium: without the external supply, nothing
    happens.

    '
  aliases:
  - Lipid World model
  - polymer GARD
  - EE-GARD
  - spatial GARD
  origin: Segre, Lancet et al., 1998-2001
  book: 6.2.4
  refs:
  - '[764]'
  - '[765]'
  - '[760]'
  - '[761]'
  - '[762]'
  - '[763]'
  - '[771]'
  - '[772]'
  - '[773]'
  - '[490]'
  - '[726]'
  - '[884]'
  family: origin-of-life
  kind: generator
  constructive: false
  S:
    repr: lipid types L_1..L_N outside (buffered) and A_1..A_N inside one assembly; the assembly's COMPOSITION
      (a 'compositional genome' / composome) carries the information
    definition: explicit
  R:
    definition: explicit
    arity:
    - 1
    - 2
    scheme: L_i --k_f--> A_i ; A_i --k_b--> L_i ; L_i + A_j --beta_ij--> A_i + A_j (catalysis by lipids
      already in the assembly)
  A:
    reactor: ssa
    dilution: outside lipids buffered; the assembly splits in two when it reaches N_max molecules
    notes: simulated with Gillespie SSA in the original papers
  params:
  - name: N_G
    type: int
    default: 100
    min: 2
    max: 200
    role: structural
    meaning: number of lipid types; the network has N_G^2 catalysed joining reactions
  - name: k_f
    type: float
    default: 0.01
    min: 0
    role: kinetic
    meaning: basal joining rate constant
  - name: k_b
    type: float
    default: 0.0001
    min: 0
    role: kinetic
    meaning: basal leaving rate constant
  - name: rho
    type: float
    default: 0.01
    min: 0
    role: population
    meaning: buffered outside concentration of every lipid type
  - name: beta_mu
    type: float
    default: -4.0
    role: kinetic
    meaning: mean of ln(beta_ij); beta is the receptor-affinity-distribution rate enhancement
  - name: beta_sigma
    type: float
    default: 4.0
    min: 0
    role: kinetic
    meaning: standard deviation of ln(beta_ij)
  - name: N_max
    type: int
    default: 100
    min: 2
    role: population
    meaning: assembly size at which fission occurs; also the crowding capacity of joining
  - name: catalysed_leaving
    type: bool
    default: false
    role: structural
    meaning: also accelerate leaving by beta_ij (the book's 'in both directions' reading), adding A_i
      + A_j -> L_i + A_j at k_b beta_ij
  provides:
  - topology
  - stoichiometry
  - catalysts
  - rate-constants
  - compartments
  - initial-state
  fidelity: reconstructed
  sources:
  - 'Markovitch, O. & Krasnogor, N. (2018). Predicting species emergence in simulated complex pre-biotic
    networks. PLoS ONE 13(2): e0192871, doi:10.1371/journal.pone.0192871 - GARD rate equation, ln(beta)
    ~ Normal(-4, 4), k_f = 1e-2, k_b = 1e-4, rho = 1e-2, N_max = N_G = 100'
  - 'Segre, D., Ben-Eli, D. & Lancet, D. (2000). Compositional genomes: prebiotic information transfer
    in mutually catalytic noncovalent assemblies. PNAS 97:4112-4117 (original model; parameters taken
    from the 2018 paper; this paper itself used k_b = 1e-5 and split at 2 N_min = 80)'
  decisions:
  - The book gives the reaction scheme but no numbers. The rate equation and all parameter values are
    from the 2018 PLoS ONE GARD paper; each term becomes one mass-action reaction with outside lipids
    buffered at rho (checked in tests).
  - That equation accelerates only joining, while the book says catalysis acts 'in both directions'. The
    default follows the paper's equation; catalysed_leaving = true adds the book's reverse catalysis.
  - The crowding factor (1 - N/N_max) is not mass action. It is recorded on every joining rate as crowding_capacity,
    and fission at N_max is described in extras.compartments. beta is sampled from the seed, so the network
    is a random instance.
  - The polymer, EE and spatial variants are not generated (the former 'variant' parameter is dropped);
    split_size is N_max; k_i, k_minus_i and beta are replaced by the distribution parameters above. thermodynamic-consistency
    is not claimed, because the default catalyses only one direction.
  phenomena:
  - homeostatically stable composomes (quasi-stationary states)
  - composome succession as mutations accumulate
  - spontaneous symmetry breaking akin to homochirality (spatial variant)
  - 'known evolvability limits: replication fidelity is too low'
  notes: 'A rare case where information lives in a MULTISET COMPOSITION rather than a sequence. Also a
    clean demonstration that catalysis alone cannot shift equilibrium - the flow term is mandatory.

    '