Biology subjects

Mason, K.

Publications and source records attributed to Mason, K..

7 recordsLinked to original sources

bioRxiv2026

Versatile Learning without Synaptic Plasticity in a Spiking Neural Network

Learning in the brains cerebral cortex is widely attributed to synaptic plasticity among cortical neurons. However, a growing body of evidence suggests that alternative processes, such as modulation of intrinsic excitability or gating by subcortical inputs, may also serve as important learning mechanisms. We developed the Bias Adaptive Neural Firing Framework (BANFF) as a simplified model of such phenomena embodied by a learnable bias current for each neuron of a rate-based network. Here, we extend this framework to spiking neural networks. We show that learning such biases enables one recurrent spiking neural network with fixed and random synaptic weights to perform well on nine tasks spanning classification, regression, and closed-loop dynamical systems mimicry. The network learnt a unique bias set for each task, and unlike recurrent synapse-based learning, new learning did not interfere with previous learning. The network was robust to non-stationary F-I curves (spike frequency adaptation), and biases could be learned with a learning algorithm (e-prop) that is more biologically plausible than stock gradient descent. Overall, we show that the BANFF can be extended from rate-based to spiking neural networks, maintaining good multi-task performance with a single network of spiking neurons.

neuroscience↗

bioRxiv2025

Direct Training of Networks of Morris-Lecar Neurons withBackprop

Spiking Neural Networks (SNNs) have the potential to replicate the brains computational efficacy by explicitly incorporating action potentials or "spikes", which is not a feature of most artificial neural networks. However, training SNNs is difficult due to the non-differentiable nature of the most common spiking models: integrate-and-fire neurons. This study investigates if some of the difficulty in training SNNs arises from the use of integrate-and-fire neurons, rather than smoother alternatives, like conductance-based neurons. To that end, we considered networks of Morris-Lecar (ML) neurons, a conductance-based neuron model which is differentiable. Networks were built using kinetic synaptic models that smoothly link presynaptic voltage dynamics directly to postsynaptic conductance changes, ensuring that all components remain fully differentiable. Switching to biophysically detailed models of synapses and neurons enabled direct end-to-end training through Backpropagation Through Time (BPTT). Biophysically detailed networks were successfully trained on image classification, regression, and time series prediction tasks. These results demonstrate the feasibility of employing biophysically detailed differentiable point neuron models to create SNNs that function as more accurate paradigms for the study of neural computations and learning. Further, this work confirms that some aspects of the difficulty in translating gradient-based learning algorithms from machine learning may arise from model choice, rather than SNNs being intrinsically difficult to train. 1. Author summaryThe brains information-processing efficiency arises in part from neurons communicating via discrete spikes. Spiking Neural Networks (SNNs) mimic this process at the neuronal level but have been difficult to train as most machine learning algorithms are not directly applicable. Most SNNs use integrate-and-fire neurons, a modelling framework that simplifies spikes into non-differentiable, abrupt voltage changes, which makes them difficult to train with powerful, standard AI training methods that use derivatives to compute gradients (e.g. Backprop). In our work, we asked if this difficulty could be overcome by considering end-to-end differentiable spiking neural networks. We used completely differentiable SNNs using the Morris-Lecar neuron, a biophysically detailed neuron model that produces smooth spikes, along with differentiable kinetic synapses. With the entire network being mathematically differentiable, we found that we could train it directly using standard backpropagation through time on different tasks (regression, classification, and chaotic time series prediction). This work demonstrates that the use of integrate-and-fire models may be limiting applications of machine learning algorithms towards understanding how learning functions in the brain.

neuroscience↗

bioRxiv2025

Rapidly Reconfigurable Dynamic Computing in Neural Networks with Fixed Synaptic Connectivity

Learning and memory in the brains neocortex have long been hypothesised to be primarily mediated by synaptic plasticity. Extensive research in artificial neural networks has shown that training networks by adjusting connection weights faces computational challenges, including large parameter spaces and the tendency of new learning to interfere with previous learning (catastrophic forgetting). We propose that the brain, which is resistant to these challenges, can also learn by modulating the excitability of each neuron in a network rather than changing synaptic strengths. We show here that learning a task-specific set of bias currents enables a feedforward or recurrent network with fixed and randomly assigned connections to perform well on and switch between dozens of tasks, including regression, classification, autonomous time series generation, a game and robotic control. Bias-only learning also provides a novel mechanistic explanation for representational drift. It directly links the noise robustness of neuronal representations on short and long time scales to the ability of neural circuits to preserve learned information while remaining adaptable. We postulate that subcortical structures, such as the basal ganglia or cerebellum, may provide similar bias inputs to the neocortex for rapid task learning and robustness against interference.

neuroscience↗

bioRxiv2024

Autonomous control of extrusion bioprinting using convolutional neural networks

Extrusion bioprinting technology suffers from reproducibility challenges due to the open-loop nature of current hardware systems. Here, we present a novel AI-powered extrusion bioprinting platform with integrated real-time quality monitoring and automated error correction capabilities. To achieve this, we engineered a custom bioprinting system with an integrated camera for continuous process monitoring and trained convolutional neural networks (CNNs) to classify the extrusion process in real-time. The CNN models, including Xception and ResNet, were trained on a combination of real and synthetic data to classify extrusion quality (good, over, or under) across various printing scenarios, including single-line and infill patterns. Notably, transfer learning, utilizing synthetic data for initial training followed by refinement with real-world data enhanced classification accuracy, with the Xception model displaying 90% accuracy for single-line extrusion and 75% for infill extrusion. This intelligent monitoring system was then coupled with a closed-loop control system that dynamically adjusted extrusion parameters on-the-fly to correct errors. The platform successfully corrected both over- and under-extrusion errors for alginate and collagen bioinks with varying rheological properties, demonstrating adaptability to unseen materials. Importantly, extrusion errors were corrected within [~]10 seconds. This novel closed-loop bioprinting platform represents a significant advance over traditional open-loop systems.

bioengineering↗

bioRxiv2023

Robust differential expression testing for single-cell CRISPR screens at low multiplicity of infection

Single-cell CRISPR screens (perturb-seq) link genetic perturbations to phenotypic changes in individual cells. The most fundamental task in perturb-seq analysis is to test for association between a perturbation and a count outcome, such as gene expression. We conduct the first-ever comprehensive benchmarking study of association testing methods for low multiplicity-of-infection (MOI) perturb-seq data, finding that existing methods produce excess false positives. We conduct an extensive empirical investigation of the data, identifying three core analysis challenges: sparsity, confounding, and model misspecification. Finally, we develop an association testing method -- SCEPTRE low-MOI -- that resolves these analysis challenges and demonstrates improved calibration and power.

genomics↗

bioRxiv2023

Niche differential gene expression analysis in spatial transcriptomics data identifies context-dependent cell-cell interactions

Single cells influence, and are shaped by, their local spatial niche. Technologies for in situ measurement of gene expression at the transcriptome scale have enabled the detailed profiling of the spatial distributions of cell types in tissue as well as the interrogation of local signaling patterns between cell types [1]. Towards these goals, we propose a new statistical procedure called niche-differential expression (niche-DE) analysis. Niche-DE identifies cell-type specific niche-associated genes, defined as genes whose expression within a specific cell type is significantly up- or down-regulated, in the context of specific spatial niches. We develop effective and interpretable measures for global false discovery control and show, through the analysis of data sets generated by myriad protocols, that the method is robust to technical issues such as over-dispersion and spot swapping. Niche-DE can be applied to low-resolution spot- and ROI-based spatial transcriptomics data as well as data that is single-cell or subcellular in resolution. Based on niche-DE, we also develop a procedure to reveal the ligand-receptor signaling mechanisms that underlie niche-differential gene expression patterns. When applied to 10x Visium data from liver metastases of colorectal cancer, niche-DE identifies marker genes for cancer-associated fibroblasts and macrophages and elucidates ligand-receptor crosstalk patterns between tumor cells, macrophages and fibroblasts. Co-detection by indexing (CODEX) was performed on the same patient samples, to corroborate the niche-DE results.

bioinformatics↗

bioRxiv2018

A non-canonical metal center drives activity of the Sediminispirochaeta smaragdinae metallo-β- lactamase SPS-1

In an effort to evaluate whether a recently reported putative metallo-{beta}-lactamase (M{beta}L) contains a novel M{beta}L active site, SPS-1 from Sediminispirochaeta smaragdinae was over-expressed, purified, and characterized using spectroscopic and crystallographic studies. Metal analyses demonstrate that recombinant SPS-1 binds nearly 2 equivalents of Zn (II), and steady-state kinetic studies show that the enzyme hydrolyzes carbapenems and certain cephalosporins but not {beta}-lactam substrates with bulky substituents in the 6-7 position. Spectroscopic studies on Co (II)-substituted SPS-1 suggest a novel metal center in SPS-1, with reduced spin coupling between the metal ions and a novel Zn1 metal binding site. This site was confirmed with a crystal structure of the enzyme. The structure shows a Zn2 site that is similar that that in NDM-1 and other subclass B1 M{beta}Ls; however, the Zn1 metal ion is coordinated by 2 histidine residues and a water molecule, which is held in position by a hydrogen bond network. The Zn1 metal is displaced nearly 1 [A] from the position reported in other M{beta}Ls. The structure also shows extended helices above the active site, which create a binding pocket that precludes the binding of substrates with large, bulky substituents in the 6/7 position of {beta}-lactam antibiotics. This study reveals a novel metal binding site in M{beta}Ls, and suggests that the targeting of metal binding sites in M{beta}Ls with inhibitors is now more challenging with the identification of this new M{beta}L.

biochemistry↗