Biology subjects

Bepler, T.

Publications and source records attributed to Bepler, T..

6 recordsLinked to original sources

bioRxiv2024

Accurate and fast segmentation of filaments and membranes in micrographs and tomograms with TARDIS

Segmentation of macromolecular structures is the primary bottleneck for studying biomolecules and their organization with electron microscopy in 2D/3D - requiring months of manual effort. Transformer-based Rapid Dimensionless Instance Segmentation (TARDIS) is a deep learning framework that automatically and accurately annotates membranes and filaments. Pre-trained TARDIS models can segment electron tomography (ET) reconstructions from both 3D and 2D electron micrographs of cryo and plastic-embedded samples. Furthermore, by implementing a novel geometric transformer architecture, TARDIS is the only method to provide accurate instance segmentations of these structures. Reducing the annotation time for ET data from months to minutes, we demonstrate segmentation of membranes and filaments in over 13,000 tomograms in the CZII Data Portal. TARDIS thus enables quantitative biophysical analysis at scale for the first time. We show this in application to kinetochore-microtubule attachment and viral-membrane interactions. TARDIS can be extended to new biomolecules and applications and open-source at https://github.com/SMLC-NYSBC/TARDIS.

bioinformatics↗

bioRxiv2022

Machine Learning Optimization of Candidate Antibodies Yields Highly Diverse Sub-nanomolar Affinity Antibody Libraries

Therapeutic antibodies are an important and rapidly growing drug modality. However, the design and discovery of early-stage antibody therapeutics remain a time and cost-intensive endeavor. In this work, we present an end-to-end Bayesian, language model-based method for designing large and diverse libraries of high-affinity single-chain variable fragments (scFvs). We integrate target-specific binding affinities with information from millions of natural protein sequences in a probabilistic machine learning framework to design thousands of scFvs that are then empirically measured. In a head-to-head comparison with a directed evolution approach, we show that the best scFv generated from our method represents a 28.8-fold improvement in binding over the best scFv from the directed evolution. Additionally, 99% of the designed scFvs in our most successful library are improvements over the initial candidate scFv. By comparing a librarys predicted success to actual measurements, we demonstrate our methods ability to explore tradeoffs between library success and diversity during the design phase and prior to experimental testing. The results of our work highlight the significant impact machine learning models can have on scFv development. We expect our end-to-end method to be broadly applicable and able to provide value to other protein engineering tasks.

bioinformatics↗

bioRxiv2022

Fully Automated Multi-Grid Cryo-EM Screening using Smart Leginon

Single particle cryo-electron microscopy (cryoEM) is a swiftly growing method for understanding protein structure. With increasing demand for high-throughput, high-resolution cryoEM services comes greater demand for rapid and automated cryoEM grid and sample screening. During screening, optimal grids and sample conditions are identified for subsequent high-resolution data collection. Screening is a major bottleneck for new cryoEM projects because grids must be optimized over several factors, including grid type, grid hole size, sample concentration, buffer conditions, ice thickness, and particle behaviors. Even for mature projects, multiple grids are commonly screened to select a subset for high-resolution data collection. Here, machine learning and novel, purpose-built image processing and microscope-handling algorithms are incorporated into the automated data collection software, Leginon, to provide an open-source solution for fully automated, high-throughput grid screening. This new version, broadly called Smart Leginon, emulates the actions of an operator in identifying areas on the grid to explore as potentially useful for data collection. Smart Leginon Autoscreen sequentially loads and examines grids from an automated specimen exchange system to provide completely unattended grid screening across a set of grids. Comparisons between a multi-grid Autoscreen session and conventional manual screening by five expert microscope operators are presented. On average, Autoscreen reduces operator time from [~]6 hours to <10 minutes and provides a comparable percentage of suitable images for evaluation as the best operator. Smart Leginons ability to target holes that are particularly difficult to identify is analyzed. Finally, Smart Leginons utility is illustrated with three real-world multi-grid user screening/collection sessions, demonstrating the efficiency and flexibility of the software package. Smart Leginons fully automated functionality significantly reduces the burden on operator screening time, improves the throughput of screening, and recovers idle microscope time, thereby improving availability of cryoEM services.

biophysics↗

bioRxiv2020

CryoDRGN: Reconstruction of heterogeneous structures from cryo-electron micrographs using neural networks

Cryo-EM single-particle analysis has proven powerful in determining the structures of rigid macromolecules. However, many protein complexes are flexible and can change conformation and composition as a result of functionally-associated dynamics. Such dynamics are poorly captured by current analysis methods. Here, we present cryoDRGN, an algorithm that for the first time leverages the representation power of deep neural networks to efficiently reconstruct highly heterogeneous complexes and continuous trajectories of protein motion. We apply this tool to two synthetic and three publicly available cryo-EM datasets, and we show that cryoDRGN provides an interpretable representation of structural heterogeneity that can be used to identify discrete states as well as continuous conformational changes. This ability enables cryoDRGN to discover previously overlooked structural states and to visualize molecules in motion.

molecular biology↗

bioRxiv2019

Topaz-Denoise: general deep denoising models for cryoEM

Cryo-electron microscopy (cryoEM) is becoming the preferred method for resolving protein structures. Low signal-to-noise (SNR) in cryoEM images reduces the confidence and throughput of structure determination during several steps of data processing, resulting in impediments such as missing particle orientations. Denoising cryoEM images can not only improve downstream analysis but also accelerate the time-consuming data collection process by allowing lower electron dose micrographs to be used for analysis. Here, we present Topaz-Denoise, a deep learning method for reliably and rapidly increasing the SNR of cryoEM images and cryoET tomograms. By training on a dataset composed of thousands of micrographs collected across a wide range of imaging conditions, we are able to learn models capturing the complexity of the cryoEM image formation process. The general model we present is able to denoise new datasets without additional training. Denoising with this model improves micrograph interpretability, enabling us to solve the first 3D single particle closed and partially open structures of clustered protocadherin, an elongated particle with previously-elusive views. We then show for the first time that low dose collection, enabled by Topaz-Denoise, improves downstream analysis in addition to reducing data collection time. We also present the first general 3D denoising model for cryoET. Topaz-Denoise and pre-trained general models are now included in Topaz. We expect that Topaz-Denoise will be of broad utility to the cryoEM community for improving micrograph and tomogram interpretability and accelerating analysis.

bioinformatics↗

bioRxiv2017

Intrinsic specificity differences between transcription factor paralogs partly explain their differential in vivo binding

Members of transcription factor (TF) families, i.e. paralogous TFs, are oftentimes reported to have identical DNA-binding motifs, despite the fact that they perform distinct regulatory functions in the cell. Differential genomic targeting by paralogous TFs is generally assumed to be due to interactions with protein cofactors or the chromatin environment. Contrary to previous assumptions, we find that paralogous TFs have different intrinsic preferences for DNA, not captured by current motif models, and these differences partly explain differential genomic binding and functional specificity. Our finding was possible due to a unique combination of carefully designed high-throughput assays and rigorous computation modeling, integrated into a unified framework called iMADS. We used iMADS to quantify, model, and analyze specificity differences between 11 paralogous TFs from 4 distinct human TF families. Our finding of differential specificity between closely related TFs has important implications for the interpretation of the regulatory effects of non-coding genetic variants.

genomics↗