Next: Defining properties of biocomplexity?


8 Conclusion


In this course we have looked at modelling (complex) biotic systems and what can and cannot be done (sensibly). As a consequence we have obtained both methodological (how can I model complex biology) and biological (what have I learned about complex biology) results. Thus we see that when we form theory, it is possible to model complex biological systems. However, while we can model a lot about them, can we tackle the defining propoerties of a biological system?

Clearly a model must be simple enough to understand, however if it has no parameters at all there will be no understanding. Thus a model should be simple as possible, but not more than that (cf Einstein, Ashby), i.e. if the obervables one uses are trivial enough then it is very easy!. We should keep this in mind: many observables are too general, i.e. it is an empty cross-section / comparison. Therefore, we should develop modeling approaches such that biological complexity becomes feasible. In this way we can develop alternative caricatures in order to survey a set of possibilities.

What we don't want is to:
  • pile up equations (doesn't give insight)
  • not take small numbers of molecules into account (cf Attofox)
  • only consider well-mixed situations (cf spatial pattern formation)

In this section we consider the following issues:
  1. Defining properties of biocomplexity?
    • Nothing in biology makes sense except in light of ...
  2. How to model a complex system (Major Transitions)?
    • Major transitions (Maynard Smith & Szathmary)
  3. Ecosystem complexity, selection and insights
    • Ecosystem complexity
    • Replicator versus influx
    • Ecosystem stability
    • Neutrality and speciation
    • Neutrality and evolvability
  4. Models of explanation (and end)
    • Waarom zijn de bananen krom?

Next: Defining properties of biocomplexity?


References















COURSE NOTES 2006-2007

Modelling (complex) biotic systems: what can / cannot be done (sensibly)
- methodological and biological results

- when form theory: is it possible to model complex biological systems
- can model a lot about them but can we tackle defining properties of biological system?

- model simple enough: no parameters, understanding
- but not simpler (cf Einstein, Ashby): if observables one uses are trivial enough then very easy!
- should keep this in mind: many observables are TOO general: i.e. is it an empty cross-section

- so should develop modeling appoaches such that bio complexity becomes feasible
- to develop alternative caricatures to survey set of possibilities


What we don't want
- pile up equations
- parameter unknown: lochness syndrome: cf lac operon using evolutoin to evaluate parameters
- small # of molecules: cf attofox?
- cell not well mixed: spatial pattern formation


Defining properties of biocomplexity?

- local interactions, many different entities in small numbers: spatial individual-based models

- leaky mutliple levels of organization (feedbacks): but not so that can be studied independently and plugged into other, but have feedbacks (leaks)
- self-organization (recognition): explicity effiort, e.g. if don't look at screen to id spirals can't understand reversal of selection
- predefined (mmm)(cpm): definition at more than one scale, nice tool to see leaky interactions

-interlocking-timescales: coupled to leaky levels
- timescale levels interact
- evo + eco dynamics: interacting, not too far apart. Eco co-existence depends on small scale evolution interactions (wiggle)
- we have seen that concept fitness (which is ill defined) is a very time dependent property
- regulation + evolution interlock: don't have a free space phenotype to choose from, but stays within sensible states: keep what before get something new (Kaufmann seminar paper)

-evolved / evolving systems:
- if we try to understand, not just function but also how we got there
- neutrality; non-functional (virtual cell seminar paper)
- not simplest implementation: dynamics of evo, + other things needs to do, i.e. relative to other things might be simplest
- evolutionary signatures: process make certain biases in how organisms do things

- make modification biocomplex systems very difficult!
- so lets make short cuts!
- lets focus on those phenomena which we can address without space, multi-level, interlocking, evolution
- but it is possible to target these defining properties as seen in the course


Nothing in biology makes sense....
- in light of evolution (Dobzhansky)
- in light of self-organization (not wow, but can be avoided)
- might be in the way! if want to be well mixed
- simple rule to complex behaviour
- loal interactions
- micro-macro transition
- non-linear dynamics etc
Much makes sense in light of both!


How model complex systems?
- how do biological systems become complex?
- certainly not due to survival of fittest mechanism: not all system evolve to complexity, many go to simpler!
- often ignored: wondering where complexity comes from!
Maynard Smith: all evolutionary theory doesn't help to understand how particular properties evolve .... elephant etc 80's.

1995: Major Transitions book
- try to identify real major changes in information processing, or prebiotic soup to now

They did
We did
- reconstruct: what did happen (at some times flawed)
- assume: Darwinian selection and local interactions (quite sure of these)
- in light of: Darwinian selection + chemistry

- infer intermediate steps + common princimples
- observe what happens due to those assumptions: what DOES happen
Bottomline for both:

multi-level evolution (entities first on own become part of larger system)


They:
- must be explained in terms of immediate selective advantage to individual replicators
why? if not constrained could make any story they want!
We:
- because of from of higher-level entities, both premises are NOT TRUE: a) fitness time dependent on different time scales, b) not in terms of individual replicators, but SPIRALS

They:
- transitions are: entities capable of independent replication before transition can replicate only as part of larger whole afterwards
We;
- have observed this: in spirals, subfunctionalization duplicated genes: can't do it alone: losing ability due to being in environment where something done by other certain will do it!; speciation, division of labour + human / insects

They:
- If there are transitions to higher-levels: it is exception to rule: conflict selection at different levels
e.g. cooperation: fair meiosis, sex reproduction, division of somatic growth, non-reproductive castes
conflict: meitic drive, parthenogenesis, escape from growth control, egg-laying workers
We:
- can also see multi-level when no conflict! e.g. host-parastie wave-spirals
- if in chaotic region evolution such that chaotic region formed + vice versa: self-reinforcing spatial pattern

- Also ? of int here: population / individual-based diversity: plasmids -ind level same direction to higher level!
- Div labour: automatic arise due to TODO, not cooperation per se!


LOCAL INTERACTION DARWINIAN EVOLUTION: feedback emerging pattern to local rules self reinforce
LOCAL INTERACTIONS: intertwined mutliple space and time scales: comp depends on global int. structure
LOCAL INTERACTION important for evolution complex replicators: this is suggested needs to be pinpointed, change circumstance all kinds of scales
LOCAL INTERACTIONS: sparse fitness more efficient optimization
TODO + LOCAL INTERACTION: automatic adaptation: self-organization clock.


Multi-level modelling (Methodology)

Not glueing together model at multi-scale
predefined notion what each will do

BUT

exploiting self-organization that can take place

- simple interactions: complex behaviour: coherence of ?
- need multi-level to understand single level phenomena: alleviate uncertain parameters intensive problems

- need it to understand: complex to complex, evolutionary signitures, novelty


Major Transitions cont.
- limit levels? - unlimited (open-ended)
- examples: - autocatalytic set vs replicating molecules: complexity molecules increases
- regulatory states vs epigenetic inheritance
- limited cultural inheritance vs language

Hogeweg;
- attractor based vs storage-based inheritance: defined dynamics system but not part of dynamics of dynamic system (Q: Possible?)
(cf Interaction between gene regulation and evolution)


Maynard Smith: same phenomena at different levels
Example:
TODO + cell behaviour: decision making
TODO + grouping: in order to have coherent motion: do what neighbour does + align
- picks up signal from those who do something else
- prototype BOIDS
Compare to cells: persistency leads to coherences (via crowding)
Beltman & Maree: CPM + persistency
- have direction: initial random
- because direction movement particular direction more probably
- if moved another direction, adjust direction
- cell alone: more random direction
- more cells: all move same direction + move faster
Same mechanism as animals: but slight different feeling during interpretation! of mechanism because "need" less to interpret!


Ecosystem complexity, selection + insights

- equilibrium system? of what property?: easier that what will transient be! what is equilibrium, which species, how many speices!

- replicator vs influx systems
- entities self- replicating vs autocatalytic not self replication

- neutrality vs selection

- levels of complexity


Ecosystem complexity

- 60's: diversity leads to stability (emergent truth, misinterpretation of lotka-volterra models)
- 70's: Gardener + Ashby / May: arbitrary interation: diversity systems less likely to be stable: close to equilibrium, random jocobian matrices:
sigma*root(species*connectivity) < diagonal ? : almost all cases
if get bigger ?? terms hold, less chance of being stable
- very little biological knowledge: just equillibrium stability analysis

later: more bio structure; same trend but more stable
- equilibrium (true for ecosystems?): predifined non-changing species


Replicator vs influx: de Boer + Hogeweg 1985 (REF)

postulated- immune system: complexity different from ecosystem
similar analysis as above:
replicator: on diagonal -a + Ecijnj: to be stable, negative: harder with greater N
influx: on diag: -b - Ecijnj, always negative, the more stable

Bottomline: self or non-self replicator dynamics quite different
- problem is out competition (hard to stabilize diversity + stability): not problem in influx
- this is important for autocatalytic to self-replicator system!


Ecosystem stability
- interlocking timescales: evolution timescale important for stabilizing diversity! wiggles (adapting polymorphism)

- population-based diversity: co-evolution populations (virus + host): may blow up complexity of system!

- ecosystem-based information accumulation: (based-problem solving, hypercycles), all sorts of side-effects create new niches: based on selection

- neutral theory (Hubble 2001): simply due to randomness birth death and dispersal, explain observed diversity: species abundant patterns in "equilibrium", but what is there is changing
- so non-functional niches? no, but should take into account 'baseline' expectation
- does function niche disturb species abundance as we would expect from random process?
(now combined with Tilman niche differentiation ...)
cf Cordero: random process genes: number of observation in gene regulation networks, conclusion not that FFL cannot have function ...


Neutrality + speciation
- quasi-species variation: -mutatnts close in genotype but not in phenotype, - depend on G-P mapping, variants quite different phenotype (cf Critters) "shapes in shadow" on neutral path

- on flat landscape (finite): (diff) clustering due to gyneology, simply random births + death Higgs + Derrida: cf neutral path

- standard speciation: allopatric (could be selection) + mating, incompatability after neutral drift (e.g. via duplications + network attractors)
- population / individual based diversity (bistability)
- alternative: nich differentiation + new interactions: ecosystem based probelm solving
- selection based, competition avoidance

About the elephant:
- individual based diversity

vs
- information threshold: not arbitrary amount information
- side-effect of evolution (fixed target) KOZA
- side-effect of small populations (Lynch): contradict information threshold
"constructive neutral evolution": effective population size is much smaller of complex organisms
might be over information threshold: selected genome, population size shrinks, suboptimal genotypes will arise, longer in population because mechanism by subfunctionalization
- e.g. chaperones because there, proteins loose independent protein, more freedom to evolve
chaperones: with stress work less well, more variation proteins, way to rescue system?

- genetic operators: duplication, GCR, transposons

- structuring gene regulation networks

- evolutionary signatures (only information local ia)
- as alternative to population based diversity
- as alternative to mutation priming red Q
- study how solve different tasks? qualitative insights what can be done
- study as mutli-level system:
critters: genes + system with dynamics of own, multiple side effects
- automatic orchestration of mutliple features
- interface with self-organization
- robustness


Neutrality + Evolvability
- important for adaptation in evolution
- long netural paths + networks
- short way to go to better solution (many solutions 1% of all protein sequences do have enzyme function)
- phenotype first evolution (prepattern): Kimura Waddington
- mutational priming: needs nothing other than previous evolutionary history
- evolutionary robustness (vs evolvability): fixed vs variable target, do not hinder each other, explore genotype space. Kauffmann: robust attractors can remain while making novelty.


Modes of explanation

what is an explanation?
"waarom waarom zijn de bananen krom?"

why why: if not bent no banana!
- trivial tautology: not bent not banana! (making form reasonable question and unreasonable)
NOT INT FORM OF EXPLANATION?
cf. survival of fitness: as tautology can give very interesting insight: not necessarily the case , fitness time dependent

what is real, banana / culture / fitness

- e.g. definitions 'culture'
- cf ALL model results: tautology follow from inputs, doesn't mean not interesting, (re)define concepts such as most informative interesting insights
- define sets of concepts such that they hang together in a meaningful way

ALMOST ALL CASES -
BEND: is generic, many ways
- straight: only one explanation
- if local interaction: default is spatial pattern formation (cf lymph nodes)
- however straight simplest (linear) case: easiest solvable vs generic (so explain why bent)


Bending is OPTIMAL: against gravity?
- cf walking in circles behaviour: very special?
- for what is optimal: minimum distance to patrol target area
-or-
- robot: one motor: one step forward + turn angle
- dangerous to have explanation in terms of optimal because generic property of motor is such that more circle

- optimal for whom? (in multi-level selection)
- cost / benefit trade-off
- dangerous, not much info + don't see other interacting pattern than those would expect


NOT TRUE ONLY ONE EXPLANATION WORTHWHILE TO DO
- ALL ARE: LOOK AT DIFFERENT ASPECTS

MULTI-LEVEL SYSTEMS IN BIOLOGICAL SYSTEMS REQUIRE MULTI-LEVEL EXPLANATIONS