Premathilaka, D. | Search bioRxiv

bioRxiv2025

Uncertainty-Gated Min-Cost Flows for In Vivo NanoScale Synaptic Plasticity Tracking

Synapses are the fundamental unit of neural connectivity, exhibiting dynamic functional and structural changes that enable the brain to learn, adapt, and form memories. Recent advances in fluorescent labeling of endogenous proteins offer an opportunity to image synaptic strength in vivo and study mechanisms underlying adaptive neural computation. Studying synaptic dynamics requires tracking signals of small, densely packed synapses over days as they change in size, position, and intensity between imaging sessions, and may even appear or disappear. Associating >50,000 dynamic, submicrometer particles across time is difficult, even for state-of-the-art algorithms. Moreover, most algorithms assign equal weight to the lateral (XY) and noisier axial (Z) dimensions, reducing performance. To address these challenges and accurately track synapses in vivo, we developed SynTrack. We formulate tracking as a Maximum A Posteriori estimation problem that identifies the K most likely disjoint paths in a Hidden Markov Model, solved using min-cost circulation optimization. An anisotropic uncertainty model accounts for poorer axial resolution, and a fully temporally connected spatio-temporal graph overcomes long-term occlusions. SynTrack achieves a mean displacement of 0.50 {micro}m with a Multiple Object Tracking Accuracy (MOTA) score of 89.8%, on par with expert annotators but with substantially increased speed and scalability. In a large-scale volume imaged over two weeks, SynTrack reconstructed 74,000 synapse trajectories detected in 4.9 out of 8 imaging sessions on average, with 18,000 synapses tracked in at least seven sessions. We present a state-of-the-art algorithm capable of high-fidelity longitudinal tracking of individual synapses in behaving mice at an unprecedented scale.

neuroscience↗

bioRxiv2024

Deep Geometric Framework to Predict Antibody-Antigen Binding Affinity

In drug development, the efficacy of an antibody depends on how the antibody interacts with the target antigen. The strength of these interactions gives an indication of how successful an antibody is in neutralizing an antigen. Therefore, the strength, measured by "binding affinity", is a critical aspect of antibody engineering. In theory, the higher the binding affinity, the higher the chances are that the antibody is successful against the target antigen. Currently, techniques such as molecular docking and molecular dynamics are utilized in quantifying the binding affinity. However, owing to the computational complexity of the aforementioned techniques, running simulations for large antibodies/antigens remains a daunting task. Despite the commendable improvements in deep learning-based binding affinity prediction, such approaches are highly dependent on the quality of the antibody-antigen structures and they tend to overlook the importance of capturing the evolutionary details of proteins upon mutation. Further, most of the existing datasets for the task only include antibody-antigen pairs related to one antigen variant and, thus, are not suitable for developing comprehensive data-driven approaches. To circumvent the said complexities, we first curate the largest and most generalized datasets for antibody-antigen binding affinity prediction, consisting of both protein sequences and structures. Subsequently, we propose a deep geometric neural network comprising a structure-based model and a sequence-based model that considers both atomistic and evolutionary details when predicting the binding affinity. The proposed framework exhibited a 10% improvement in mean absolute error compared to the state-of-the-art models while showing a strong correlation between the predictions and target values. We release the datasets and code publicly (https://drug-discovery-entc.github.io/p2pxml/) to support the development of antibody-antigen binding affinity prediction frameworks for the benefit of science and society.

bioinformatics↗