Post-Doc Position Open:
Exploratory processes in evolution and development

Project Title: How Exploratory and Selective Developmental Mechanisms Generate facilitated variation, confer agency on organisms and impose purpose on evolution

Kevin Laland (St Andrews) 

Richard A. Watson (Southampton) 

Applications are invited for a full-time, fixed-term (2-year) Postdoctoral Research Fellow to work with Professor Kevin Laland (Biology, University of St Andrews) and Professor Richard Watson (Computer Science, Southampton University). The successful candidate will develop computational models of adaptation and evolvability in biological systems that exhibit complex forms of adaptive plasticity (exploratory behavior). The successful candidate will have a good degree in biology or computer science and a relevant PhD, with a demonstrated ability: to build computational models, to reason about evolutionary process in a computational framework, to produce high-quality scientific publications and... for independent thinking.

​

The applicant will be based at the University of St Andrews (UK) with Kevin Laland, and also working with Richard Watson (University of Southampton, UK).

​

The post is available for 24 months starting January 2022 or as soon as possible thereafter.

Informal enquiries can be directed to Professor Kevin Laland, knl1@st-andrews.ac.uk or Linda Hall, lnh1@st-andrews.ac.uk. Or Richard Watson (DrRichardAWatson@gmail.com)

​

The project is funded by The John Templeton Foundation (The Science of Purpose funding initiative).

​

Working Hours: Full time

Grade/Salary Range: Grade 6/£34,304 per annum

Please quote ref: AR2609SB

Closing Date: 11 November 2021

Interview Date: Early December 2021

Further Particulars: AR2609SB FPs.doc

​

Links:

Jobs.ac.uk advert

StAndrews advert page   <== apply here

About the Project

Exploratory Process and their role in Evolution

Living organisms must produce functional responses to highly diverse, complex and constantly changing inputs (e.g. immunological responses to rapidly evolving bacteria or viruses, or learned adjustments to changing situations). Often organisms respond to such challenges through ‘exploratory mechanisms’ (Gerhart & Kirschner, 1997; West-Eberhard 2003; Kirschner & Gerhart, 2005), which are complex developmental systems that operate by generating variation (i.e. ‘exploring’ possibilities), largely at random, testing variants’ functionality, and selecting the best solutions for regeneration, in an iterative developmental process. The process resembles adaptation by natural selection (a.k.a. ‘somatic selection’), except that it allows for ontogenetic information gain within a lifetime rather than conventional genetic information gain across multiple generations.

​

Diverse biological processes function in this way. For instance, the adaptive immune system generates antibodies and T cells with initially random variation, then internal selection multiplies and refines those that bind successfully to antigens, with a memory of effective molecules retained (Klenerman, 2017). The vascular and tracheal systems, animal learning, much collective animal behavior (e.g. central-place foraging), and micro-tubular systems all operate on similar principles – exploiting exploratory and selective mechanisms to generate novel functional responses in development (Campbell 1960; Gerhart & Kirschner, 1997). Other biological processes, such as the remodeling of bone and soft tissue (muscles, tendons), are known to be responsive to functional demands (Hall, 2015), and these processes have also been characterized as reliant on somatic selection (West- Eberhard, 2003). The anatomical organization of brains exhibits similar adaptability. There are estimated to be over 100 trillion neural connections (synapses) in the human brain, several orders of magnitude more than could be specified by the c. 21,000 genes in our genome (Edelman, 1987; Kirschner & Gerhart, 2005). While the gross structure of the vertebrate nervous system is thought to be set up by the demarcation of pathways of nerve growth by genes (Kirschner & Gerhart 2005), experiments show that brains depend in part on exploratory mechanisms to establish their anatomical organization (Edelman 1987; Gerhart & Kirschner, 1997; Kirschner & Gerhart 2005). During development, the nervous system generates excess neurons, excess neuronal connections, and excessively distributed neuronal connections, through random exploration. It then prunes these, retaining solely those required. Much of the patterning of the brain depends on exploratory mechanisms’ use of functional interactions to sort out connectivity. The final anatomy of vertebrate brains thus depends heavily on experience.

 

Exploratory mechanisms are adaptive because rapid exploration of a large space of possibilities combined with feedback (e.g. reward/punishment) allows for information gain from the current environment. Crucially this occurs at timescales faster than genetic evolution (i.e. within individual’s lifetimes). Challenges arise from the internal environment too. Organisms must cope effectively with very large numbers of individual-specific ‘internal failures’ in somatic genome, epigenome and microbiome that are too numerous and/or unique to be anticipated by genetically coded plasticity, and the self-organization of random variation is critical to this form of ‘adaptive improvisation’ (Soen et al., 2015). As a result, across a very broad range of conditions, including unanticipated circumstances, organisms are often capable of producing highly functional responses. Experiments and theory both show that randomness in exploration is especially useful in confronting a variable or novel environment (Deneubourg et al 1983; Strickland et al 1995; Soen et al, 2015; Richerson 2019). Exploratory mechanisms have a major advantage in flexibility (Gerhart & Kirschner 1997), being self-correcting and adaptable to changes in other parts of the organism – e.g. resizing cortical areas to match sensory fields (Gerhart & Kirschner, 1997).

 

Exploratory mechanisms can be costly systems because they are wasteful – to find effective solutions they must generate a very large number of variants, only a fraction of which will be retained (Gerhart & Kirschner, 1997). Hence, natural selection should favor biases in the operation of exploratory mechanisms (e.g. allowing the immune system rapidly and cheaply to target a specific reliably present antigen). In principle, exploratory mechanisms can adjust to new challenges during ontogeny, and later these phenotypes can be stabilized by natural selection (e.g. generating probabilistic biases in exploration through shaping precursor cell numbers, nerve growth factor concentrations, or induction factor concentrations in particular regions). Yet there must be limits to the extent of such biases if the benefits of exploration are to be retained. That is, there is an inherent trade-off between being biased to produce particular responses efficiently and the cost of being able to explore novel responses when needed. Currently, the nature of those trade-offs is not understood.  

​

Program of Work

This project builds a novel interdisciplinary collaboration between a biologist (Laland, Project Leader) and computer scientist (Watson, Project Co-Leader), each with outstanding track records in this field. This ground-breaking study will pioneer computational models of adaptation in biological systems that exhibit exploratory behavior in interaction with genetic evolution. The novelty of our approach is to exploit the deep functional isomorphism between multiple adaptive timescales in learning systems, well-understood in (machine) learning theory (Watson & Szathmáry 2016). This has already been highly successful in modelling the interaction of variation and selection processes occurring on different timescales, namely, in the evolution of evolvability (Kouvaris et al, 2017; Kouvaris 2018; Watson et al 2014, 2016; Watson & Szathmáry 2016; see also Parter et al 2008). Here we build on this earlier work to study the interaction of variation and selection processes both within and between generations – the evolution of ‘explorability’. This enables (1) the formal evolutionary analysis of exploratory mechanisms, (2) the simultaneous modelling of adaptive processes across two timescales, developmental and evolutionary, without the latter removing the former, and (3) analyses of interactions between exploratory mechanisms and auxiliary processes. Each extension plausibly enables forms of adaptation impossible with simpler formulations.