The consortium has been developed for complementary expertise, resources and approaches. The partners have had previous bilateral collaborations (Webb & Louis; Gerber & Louis; Webb & Armstrong; Armstrong & BWD; Louis & BWD). The consortium’s joint expertise in experimental biology and computational modelling, coupled with a commercial sector high throughput facility for Drosophila assays (BWD), represents a key asset in the delineation of abstract principles governing memory-based decision making and the organization of behaviour.

The strength of the consortium is best illustrated by its direct alignment to the project objectives:

• To analyse at a fine scale how larval olfactory behaviour is controlled and altered by associative conditioning with attractive and aversive gustatory stimuli, and how memory-based decision making is organised according to motivational and contextual states. This will be tightly linked to agent-based computational models of the larval behaviour that allow us to test crucial computational principles required to support the observed behavioural phenomena.

Some of our core questions have arisen from two different directions: LIN has demonstrated experimentally that memory-based larval behaviour is dependent for its expression on contextual assessment of the advantage of acting towards the conditioned stimulus, opening the way to address deep theoretical questions in this simple system. UEDIN has been engaged in modelling the circuits of insect learning and aligning these with the current theoretical, experimental and neuroscience literature in vertebrate learning which has revealed the significant gap in understanding of Pavlovian action that motivates our proposals. BWD has extensively used learning assays in validating its biomedical
models of human disease. Equally important is that CRG has a long-standing interest in understanding the mechanisms of innate olfactory behaviour. CRG has established highly controlled methods for investigating larval sensorimotor control, applied to date to understand innate attraction to odour. The extension of these methods to conditioned behaviour, both attractive and aversive, depends directly on
the learning paradigms that were pioneered by LIN and continue to be developed by them. This requires advance in the ability to track individual larvae, including in group testing situations, and to describe their behaviour quantitatively, which is directly the expertise available from CRG and BWD; UEDIN will also contribute technical support for the future development of these methods,particularly with regard to interpretation of the behavioural results and the integration of the results into models.

• To build and validate integrative models at the level of neural circuits and link them to the characterisation of neural function in larvae through controlled loss- of-function and gain-of-function experiments, functional imaging and functional perturbations through optogenetics. This will make use of unique and novel tools to gain unprecedented correspondence between model and system manipulations.

UEDIN has a substantial track record in building models at multiple levels to understand insect behaviour, including testing sets of alternative mathematical models against behavioural data, testing agent simulations to reveal key algorithms of control, and testing neural algorithms on real robots under identical stimulus conditions to the organism; CRG has already developed preliminary algorithmic descriptions of the larval chemotactic behaviour and will be able to supply on-going data and insights throughout the project; the tracking software and data analysis tools developed in the consortium will take full advantage of the behavioural analyses to be contributed by LIN, UEDIN; and
CRG will provide the basis for rigorous model comparison, and to sharpen our experimental designs to test novel predictions of the developed models. LIN and CRG have substantial experience and facilities for carrying out the genetic analyses: for example, CRG carried out a large-scale behavioural screen of the DGRC collection to identify neural circuits contributing to innate odour-search behaviours; the loss-of-function phenotype of more than 1300 Gal4 lines was assessed in less than 9 months. Furthermore, thanks to their access to the HHMI Janelia Farm as well as publically available DGRC collection of Drosophila strains we are in an ideal position to take advantage of these resources to understand memory-based decision making. CRG has experience in using GCaMP to monitor sensory neurons and antennal lobe neurons, using confocal and 2-photon microscopy and these methods can be extended to image from other neuronal classes. BWD has experience in developing bioluminescent assays for neuronal activity in Kenyon cells. Both CRG and LIN have used optogenetic methods to manipulate, respectively, olfactory sensory neuron activity and dopaminergic and octopaminergic (reward) neural activity in behaving animals. Technical and informatics expertise from UEDIN will help to integrate these methods with online tracking systems, enabling simultaneous observation and modulation of neuronal activity and behaviour, providing an unprecedented degree of experimental control. Importantly, UEDIN can also provide database support to organize the anatomical and functional results.

To abstract dynamical principles and derive novel, generalisable algorithms and architectures that can be used to enhance the learning and anticipatory capabilities of machines, and apply them in extended scenarios. This will involve a proof of concept that
such autonomous learning capacities in real environments can be achieved with relatively minimal computational power.

UEDIN has a strong record in applying abstract and mathematical methods to generalise findings from neuroscience and understand their computational significance. Indeed, UEDIN is a European leading centre for machine learning and robotics. CRG has pioneered the application of systems-biology approaches to characterize innate orientation behaviours in the Drosophila larva. The work of the group illustrates the importance of combining quantitative methods to describe dynamical processes, particularly sensorimotor integration, and to perturb these processes in a predictable way. This necessitates combining a wide range of experimental techniques with mathematical modelling. By combining the expertise of both groups, the consortium will achieve a unique expertise in systems and computational neuroscience. In addition UEDIN and CRG are connected to the neuromorphic engineering community, which would relay the testing of the findings of MINIMAL in VLSI devices.
Thus, the consortium’s composition echoes one of our key objectives, namely to understand how memory-based behaviour is integrated with innate behaviour – and to incorporate the rules of these interactions in theoretical models.

Leave a Reply

Your email address will not be published. Required fields are marked *