The ability to simultaneously record from large numbers of neurons in behaving animals has ushered in a new era for the study of the neural circuit mechanisms underlying cognitive functions. One ...promising approach to uncovering the dynamical and computational principles governing population responses is to analyze model recurrent neural networks (RNNs) that have been optimized to perform the same tasks as behaving animals. Because the optimization of network parameters specifies the desired output but not the manner in which to achieve this output, "trained" networks serve as a source of mechanistic hypotheses and a testing ground for data analyses that link neural computation to behavior. Complete access to the activity and connectivity of the circuit, and the ability to manipulate them arbitrarily, make trained networks a convenient proxy for biological circuits and a valuable platform for theoretical investigation. However, existing RNNs lack basic biological features such as the distinction between excitatory and inhibitory units (Dale's principle), which are essential if RNNs are to provide insights into the operation of biological circuits. Moreover, trained networks can achieve the same behavioral performance but differ substantially in their structure and dynamics, highlighting the need for a simple and flexible framework for the exploratory training of RNNs. Here, we describe a framework for gradient descent-based training of excitatory-inhibitory RNNs that can incorporate a variety of biological knowledge. We provide an implementation based on the machine learning library Theano, whose automatic differentiation capabilities facilitate modifications and extensions. We validate this framework by applying it to well-known experimental paradigms such as perceptual decision-making, context-dependent integration, multisensory integration, parametric working memory, and motor sequence generation. Our results demonstrate the wide range of neural activity patterns and behavior that can be modeled, and suggest a unified setting in which diverse cognitive computations and mechanisms can be studied.
Abstract
To explore any relationship between the ABO blood group and the coronavirus disease 2019 (COVID-19) susceptibility, we compared ABO blood group distributions in 2173 COVID-19 patients with ...local control populations, and found that blood group A was associated with an increased risk of infection, whereas group O was associated with a decreased risk.
Recently it has been proposed that information in working memory (WM) may not always be stored in persistent neuronal activity but can be maintained in 'activity-silent' hidden states, such as ...synaptic efficacies endowed with short-term synaptic plasticity. To test this idea computationally, we investigated recurrent neural network models trained to perform several WM-dependent tasks, in which WM representation emerges from learning and is not a priori assumed to depend on self-sustained persistent activity. We found that short-term synaptic plasticity can support the short-term maintenance of information, provided that the memory delay period is sufficiently short. However, in tasks that require actively manipulating information, persistent activity naturally emerges from learning, and the amount of persistent activity scales with the degree of manipulation required. These results shed insight into the current debate on WM encoding and suggest that persistent activity can vary markedly between short-term memory tasks with different cognitive demands.
Recurrent neural networks (RNNs) trained with machine learning techniques on cognitive tasks have become a widely accepted tool for neuroscientists. In this short opinion piece, we discuss ...fundamental challenges faced by the early work of this approach and recent steps to overcome such challenges and build next-generation RNN models for cognition. We propose several essential questions that practitioners of this approach should address to continue to build future generations of RNN models.
•The standard approach of training one RNN with one learning algorithm on one cognitive task needs to be revised.•Multi-area RNNs should be embraced to model different brain regions that process information hierarchically and in parallel.•RNNs should be trained with multiple algorithms so they can learn at disparate timescales as animals do.•A set of canonical tasks is needed to pre-train RNNs to guide them towards more naturalistic solutions.
Enzyme stability is an important issue for protein engineers. Understanding how rigidity in the active site affects protein kinetic stability will provide new insight into enzyme stabilization. In ...this study, we demonstrated enhanced kinetic stability of Candida antarctica lipase B (CalB) by mutating the structurally flexible residues within the active site. Six residues within 10 Å of the catalytic Ser105 residue with a high B factor were selected for iterative saturation mutagenesis. After screening 2200 colonies, we obtained the D223G/L278M mutant, which exhibited a 13-fold increase in half-life at 48 °C and a 12 °C higher T5015, the temperature at which enzyme activity is reduced to 50% after a 15-min heat treatment. Further characterization showed that global unfolding resistance against both thermal and chemical denaturation also improved. Analysis of the crystal structures of wild-type CalB and the D223G/L278M mutant revealed that the latter formed an extra main chain hydrogen bond network with seven structurally coupled residues within the flexible α10 helix that are primarily involved in forming the active site. Further investigation of the relative B factor profile and molecular dynamics simulation confirmed that the enhanced rigidity decreased fluctuation of the active site residues at high temperature. These results indicate that enhancing the rigidity of the flexible segment within the active site may provide an efficient method for improving enzyme kinetic stability.
Background: Improving the kinetic stability of enzymes is a key issue for protein engineers.
Results: Mutagenesis of residues with a high B factor located within 10 Å of the catalytic Ser105 residue enhances kinetic stability dramatically.
Conclusion: Increasing the rigidity of the flexible segment within the active site improves enzymatic kinetic stability.
Significance: Optimization of the active site may an alternative, efficient approach for enhancing protein stabilization.
Transcript assembly from RNA-seq reads is a critical step in gene expression and subsequent functional analyses. Here we present PsiCLASS, an accurate and efficient transcript assembler based on an ...approach that simultaneously analyzes multiple RNA-seq samples. PsiCLASS combines mixture statistical models for exonic feature selection across multiple samples with splice graph based dynamic programming algorithms and a weighted voting scheme for transcript selection. PsiCLASS achieves significantly better sensitivity-precision tradeoff, and renders precision up to 2-3 fold higher than the StringTie system and Scallop plus TACO, the two best current approaches. PsiCLASS is efficient and scalable, assembling 667 GEUVADIS samples in 9 h, and has robust accuracy with large numbers of samples.
Trained neural network models, which exhibit features of neural activity recorded from behaving animals, may provide insights into the circuit mechanisms of cognitive functions through systematic ...analysis of network activity and connectivity. However, in contrast to the graded error signals commonly used to train networks through supervised learning, animals learn from reward feedback on definite actions through reinforcement learning. Reward maximization is particularly relevant when optimal behavior depends on an animal's internal judgment of confidence or subjective preferences. Here, we implement reward-based training of recurrent neural networks in which a value network guides learning by using the activity of the decision network to predict future reward. We show that such models capture behavioral and electrophysiological findings from well-known experimental paradigms. Our work provides a unified framework for investigating diverse cognitive and value-based computations, and predicts a role for value representation that is essential for learning, but not executing, a task.
Rockfalls are an important factor affecting underground engineering safety. However, there has been limited progress in understanding and predicting these disasters in the past few years. Therefore, ...a large-scale three-dimensional experimental simulation apparatus to study failure mechanisms of rockfalls occurring during underground engineering was developed. This apparatus, measuring 4 m × 4 m × 3.3 m in size, can achieve vertical and horizontal symmetric loading. It not only simulates the structure and stress environment of a rock mass but also simulates the stepwise excavation processes involved in underground engineering. A complete simulation experiment of rockfalls in an underground engineering context was performed using this apparatus. Dynamic evolution characteristics of block displacement, temperature, natural vibration frequency, and acoustic emissions occurring during rockfalls were studied during the simulation. These data indicate there are several indicators that could be used to predict rockfalls in underground engineering contexts, leading to better prevention and control.
Tools for differential splicing detection have failed to provide a comprehensive and consistent view of splicing variation. We present MntJULiP, a novel method for comprehensive and accurate ...quantification of splicing differences between two or more conditions. MntJULiP detects both changes in intron splicing ratios and changes in absolute splicing levels with high accuracy, and can find classes of variation overlooked by other tools. MntJULiP identifies over 29,000 differentially spliced introns in 1398 GTEx brain samples, including 11,242 novel introns discovered in this dataset. Highly scalable, MntJULiP can process thousands of samples within hours to reveal splicing constituents of phenotypic differentiation.
Pure tungsten samples were prepared by the selective electron beam melting (SEBM) process. The effect of the SEBM process parameters on the density, microstructure and compression strength of pure ...tungsten was studied. In addition, the influence of substrate preheating temperature during SEBM was studied. A processing window for additive manufacturing of pure tungsten by SEBM was preliminarily determined. Pure tungsten samples with relative density of 99.5% and without obvious pores and microcracks were successfully fabricated. The as-built pure tungsten samples showed strong columnar grain structures. Compression strength along the columnar grains in the build direction was measured to be 1560 MPa. Fracture occurred predominantly along the columnar grain boundaries by decohesion, in addition to brittle transgranular fracture. Refinement and strengthening of the columnar grain boundaries are expected to improve the compression strength of the SEBM-fabricated pure tungsten.
•Process window of pure tungsten fabricated by SEBM was determined.•Pure tungsten samples without obvious pores and cracks were successfully fabricated.•SEBM pure tungsten exhibited columnar crystal structure by epitaxial growth.•The more stable the molten pool, the more complete the columnar crystals.