Climate research priorities for policy-makers, practitioners, and scientists in Georgia, USA.

Climate change has far-reaching effects on human and ecological systems, requiring collaboration across sectors and disciplines to determine effective responses. To inform regional responses to climate change, decision-makers need credible and relevant information representing a wide swath of knowledge and perspectives. The southeastern U. S. State of Georgia is a valuable focal area for study because it contains multiple ecological zones that vary greatly in land use and economic activities, and it is vulnerable to diverse climate change impacts. We identified 40 important research questions that, if answered, could lay the groundwork for effective, science-based climate action in Georgia. Top research priorities were identified through a broad solicitation of candidate research questions (180 were received). A group of experts across sectors and disciplines gathered for a workshop to categorize, prioritize, and filter the candidate questions, identify missing topics, and rewrite questions. Participants then collectively chose the 40 most important questions. This cross-sectoral effort ensured the inclusion of a diversity of topics and questions (e.g., coastal hazards, agricultural production, ecosystem functioning, urban infrastructure, and human health) likely to be important to Georgia policy-makers, practitioners, and scientists. Several cross-cutting themes emerged, including the need for long-term data collection and consideration of at-risk Georgia citizens and communities. Workshop participants defined effective responses as those that take economic cost, environmental impacts, and social justice into consideration. Our research highlights the importance of collaborators across disciplines and sectors, and discussing challenges and opportunities that will require transdisciplinary solutions.

Dichotomy in the definition of prescriptive information suggests both prescribed data and prescribed algorithms: biosemiotics applications in genomic systems.

The fields of molecular biology and computer science have cooperated over recent years to create a synergy between the cybernetic and biosemiotic relationship found in cellular genomics to that of information and language found in computational systems. Biological information frequently manifests its "meaning" through instruction or actual production of formal bio-function. Such information is called prescriptive information (PI). PI programs organize and execute a prescribed set of choices. Closer examination of this term in cellular systems has led to a dichotomy in its definition suggesting both prescribed data and prescribed algorithms are constituents of PI. This paper looks at this dichotomy as expressed in both the genetic code and in the central dogma of protein synthesis. An example of a genetic algorithm is modeled after the ribosome, and an examination of the protein synthesis process is used to differentiate PI data from PI algorithms.

Training spiking neuronal networks to perform motor control using reinforcement and evolutionary learning.

Artificial neural networks (ANNs) have been successfully trained to perform a wide range of sensory-motor behaviors. In contrast, the performance of spiking neuronal network (SNN) models trained to perform similar behaviors remains relatively suboptimal. In this work, we aimed to push the field of SNNs forward by exploring the potential of different learning mechanisms to achieve optimal performance. We trained SNNs to solve the CartPole reinforcement learning (RL) control problem using two learning mechanisms operating at different timescales: (1) spike-timing-dependent reinforcement learning (STDP-RL) and (2) evolutionary strategy (EVOL). Though the role of STDP-RL in biological systems is well established, several other mechanisms, though not fully understood, work in concert during learning in vivo. Recreating accurate models that capture the interaction of STDP-RL with these diverse learning mechanisms is extremely difficult. EVOL is an alternative method and has been successfully used in many studies to fit model neural responsiveness to electrophysiological recordings and, in some cases, for classification problems. One advantage of EVOL is that it may not need to capture all interacting components of synaptic plasticity and thus provides a better alternative to STDP-RL. Here, we compared the performance of each algorithm after training, which revealed EVOL as a powerful method for training SNNs to perform sensory-motor behaviors. Our modeling opens up new capabilities for SNNs in RL and could serve as a testbed for neurobiologists aiming to understand multi-timescale learning mechanisms and dynamics in neuronal circuits.

Training a spiking neuronal network model of visual-motor cortex to play a virtual racket-ball game using reinforcement learning.

Recent models of spiking neuronal networks have been trained to perform behaviors in static environments using a variety of learning rules, with varying degrees of biological realism. Most of these models have not been tested in dynamic visual environments where models must make predictions on future states and adjust their behavior accordingly. The models using these learning rules are often treated as black boxes, with little analysis on circuit architectures and learning mechanisms supporting optimal performance. Here we developed visual/motor spiking neuronal network models and trained them to play a virtual racket-ball game using several reinforcement learning algorithms inspired by the dopaminergic reward system. We systematically investigated how different architectures and circuit-motifs (feed-forward, recurrent, feedback) contributed to learning and performance. We also developed a new biologically-inspired learning rule that significantly enhanced performance, while reducing training time. Our models included visual areas encoding game inputs and relaying the information to motor areas, which used this information to learn to move the racket to hit the ball. Neurons in the early visual area relayed information encoding object location and motion direction across the network. Neuronal association areas encoded spatial relationships between objects in the visual scene. Motor populations received inputs from visual and association areas representing the dorsal pathway. Two populations of motor neurons generated commands to move the racket up or down. Model-generated actions updated the environment and triggered reward or punishment signals that adjusted synaptic weights so that the models could learn which actions led to reward. Here we demonstrate that our biologically-plausible learning rules were effective in training spiking neuronal network models to solve problems in dynamic environments. We used our models to dissect the circuit architectures and learning rules most effective for learning. Our model shows that learning mechanisms involving different neural circuits produce similar performance in sensory-motor tasks. In biological networks, all learning mechanisms may complement one another, accelerating the learning capabilities of animals. Furthermore, this also highlights the resilience and redundancy in biological systems.

A comparative approach for the investigation of biological information processing: an examination of the structure and function of computer hard drives and DNA.

BACKGROUND: The robust storage, updating and utilization of information are necessary for the maintenance and perpetuation of dynamic systems. These systems can exist as constructs of metal-oxide semiconductors and silicon, as in a digital computer, or in the "wetware" of organic compounds, proteins and nucleic acids that make up biological organisms. We propose that there are essential functional properties of centralized information-processing systems; for digital computers these properties reside in the computer's hard drive, and for eukaryotic cells they are manifest in the DNA and associated structures. METHODS: Presented herein is a descriptive framework that compares DNA and its associated proteins and sub-nuclear structure with the structure and function of the computer hard drive. We identify four essential properties of information for a centralized storage and processing system: (1) orthogonal uniqueness, (2) low level formatting, (3) high level formatting and (4) translation of stored to usable form. The corresponding aspects of the DNA complex and a computer hard drive are categorized using this classification. This is intended to demonstrate a functional equivalence between the components of the two systems, and thus the systems themselves. RESULTS: Both the DNA complex and the computer hard drive contain components that fulfill the essential properties of a centralized information storage and processing system. The functional equivalence of these components provides insight into both the design process of engineered systems and the evolved solutions addressing similar system requirements. However, there are points where the comparison breaks down, particularly when there are externally imposed information-organizing structures on the computer hard drive. A specific example of this is the imposition of the File Allocation Table (FAT) during high level formatting of the computer hard drive and the subsequent loading of an operating system (OS). Biological systems do not have an external source for a map of their stored information or for an operational instruction set; rather, they must contain an organizational template conserved within their intra-nuclear architecture that "manipulates" the laws of chemistry and physics into a highly robust instruction set. We propose that the epigenetic structure of the intra-nuclear environment and the non-coding RNA may play the roles of a Biological File Allocation Table (BFAT) and biological operating system (Bio-OS) in eukaryotic cells. CONCLUSIONS: The comparison of functional and structural characteristics of the DNA complex and the computer hard drive leads to a new descriptive paradigm that identifies the DNA as a dynamic storage system of biological information. This system is embodied in an autonomous operating system that inductively follows organizational structures, data hierarchy and executable operations that are well understood in the computer science industry. Characterizing the "DNA hard drive" in this fashion can lead to insights arising from discrepancies in the descriptive framework, particularly with respect to positing the role of epigenetic processes in an information-processing context. Further expansions arising from this comparison include the view of cells as parallel computing machines and a new approach towards characterizing cellular control systems.

Redundancy of the genetic code enables translational pausing.

The codon redundancy ("degeneracy") found in protein-coding regions of mRNA also prescribes Translational Pausing (TP). When coupled with the appropriate interpreters, multiple meanings and functions are programmed into the same sequence of configurable switch-settings. This additional layer of Ontological Prescriptive Information (PIo) purposely slows or speeds up the translation-decoding process within the ribosome. Variable translation rates help prescribe functional folding of the nascent protein. Redundancy of the codon to amino acid mapping, therefore, is anything but superfluous or degenerate. Redundancy programming allows for simultaneous dual prescriptions of TP and amino acid assignments without cross-talk. This allows both functions to be coincident and realizable. We will demonstrate that the TP schema is a bona fide rule-based code, conforming to logical code-like properties. Second, we will demonstrate that this TP code is programmed into the supposedly degenerate redundancy of the codon table. We will show that algorithmic processes play a dominant role in the realization of this multi-dimensional code.