Biology subjects

Sun, R.

Publications and source records attributed to Sun, R..

At least 127 records · Page 7Linked to original sources

bioRxiv2020

Improved subspecies identification in clinical Mycobacterium abscessus complex isolates using whole genome sequence

Mycobacterium abscessus complex, which is frequently reported causing a variety of skin and soft tissues diseases in humans, is composed of three subspecies, namely M. abscessus subsp. abscessus, M. abscessus subsp. massiliense and M. abscessus subsp. bolletii. Currently, the differentiation of these three subspecies in clinical isolates still largely depend on single gene identification methods like the genes namely hsp65, 16s with a limited accuracy. This study confirmed the limitations of the single gene based method in the subspecies identification. We performed a comprehensive analysis of MABC genomes in the NCBI database and tried to build an accurate and user-friendly identify method. Here, we describe an improved assay for Mycobacterium abscessus complex fast identification using WGS data, based on the identities of rpoB, erm(41) and rpls. Comprehensive analysis has been performed to compare our software results with the traditional method. The result showed that the method built-in this study could 100% identification the subspecies for the Mycobacterium abscessus complex in the public genome database (893 genomes from NCBI database and 6 clinical isolates from this study). Because this software can be easily integrated into a routine workflow to quickly and precisely provide subspecies-level identification and discrimination MABC different subspecies in clinical isolates by WGS. This assay will facilitate accurate molecular identification of species from the MABC complex in a variety of clinical specimens and diagnostic contexts.

microbiology↗

bioRxiv2019

Tuning antiviral CD8 T-cell response via proline-altered peptide ligand vaccination

Viral escape from CD8+ cytotoxic T lymphocyte responses correlates with disease progression and represents a significant challenge for vaccination. Here, we demonstrate that CD8+ T cell recognition of the naturally occurring MHC-I-restricted LCMV-associated immune escape variant Y4F is restored following vaccination with a proline-altered peptide ligand (APL). The APL increases MHC/peptide (pMHC) complex stability, rigidifies the peptide and facilitates T cell receptor (TCR) recognition through reduced entropy costs. Structural analyses of pMHC complexes before and after TCR binding, combined with biophysical analyses, revealed that although the TCR binds similarly to all complexes, the p3P modification alters the conformations of a very limited amount of specific MHC and peptide residues, facilitating efficient TCR recognition. This approach can be easily introduced in peptides restricted to other MHC alleles, and can be combined with currently available and future vaccination protocols in order to prevent viral immune escape. Author SummaryViral escape mutagenesis correlates often with disease progression and represents a major hurdle for vaccination-based therapies. Here, we have designed and developed a novel generation of altered epitopes that re-establish and enhance significantly CD8+ T cell recognition of a naturally occurring viral immune escape variant. Biophysical and structural analyses provide a clear understanding of the molecular mechanisms underlying this reestablished recognition. We believe that this approach can be implemented to currently available or novel vaccination approaches to efficiently restore T cell recognition of virus escape variants to control disease progression.

immunology↗

bioRxiv2019

Predominance of positive epistasis among drug resistance-associated mutations in HIV-1 protease

Drug-resistant mutations often have deleterious impacts on replication fitness, posing a fitness cost that can only be overcome by compensatory mutations. However, the role of fitness cost in the evolution of drug resistance has often been overlooked in clinical studies or in vitro selection experiments, as these observations only capture the outcome of drug selection. In this study, we systematically profile the fitness landscape of resistance-associated sites in HIV-1 protease using deep mutational scanning. We construct a mutant library covering combinations of mutations at 11 sites in HIV-1 protease, all of which are associated with resistance to protease inhibitors in clinic. Using deep sequencing, we quantify the fitness of thousands of HIV-1 protease mutants after multiple cycles of replication in human T cells. Although the majority of resistance-associated mutations have deleterious effects on viral replication, we find that epistasis among resistance-associated mutations is predominantly positive. Furthermore, our fitness data are consistent with genetic interactions inferred directly from HIV sequence data of patients. Fitness valleys formed by strong positive epistasis reduce the likelihood of reversal of drug resistance mutations. Overall, our results support the view that strong compensatory effects are involved in the emergence of clinically observed resistance mutations and provide insights to understanding fitness barriers in the evolution and reversion of drug resistance.

evolutionary biology↗

bioRxiv2019

Genome dashboards: Framework and Examples

Genomics is a sequence-based informatics science and a 3D structure-based material science. Here we describe a framework for developing genome dashboards specifically designed to unify informatics with studies of chromatin structure and dynamics. The framework is based on the mathematical representation of geometrically exact rod models and the generalization of DNA base pair step parameters. A Model-View-Controller software design approach is proposed to implement genome dashboards as finite state machines either as desktop or web based applications. Two examples are demonstrated using our minimal genome dashboard called G-Dash-min. The data unification achieved with a genome dashboard supports the bi-directional exchange of data between informatics and structure. Thus any experimentally or theoretically determined sequence based informatics track can inform DNA, nucleosome or chromatin modeling (e.g. nucleosome positions) and structure features can be analyzed as informatics tracks in a genome browser (e.g. DNA base pair step parameters: Roll, Tilt, Twist). Here the framework is applied to chromatin, but genome dashboards are more broadly applicable. Genome dashboards are a novel means of investigating structure-function relationships for regions of the genome ranging from base pairs to entire chromosomes and for generating, validating, and testing mechanistic hypotheses.

genomics↗

bioRxiv2019

PulseDIA: in-depth data independent acquisition mass spectrometry using enhanced gas phase fractionation

An inherent bottleneck of data independent acquisition (DIA) analysis by Orbitrap-based mass spectrometers is the relatively large window width due to the relatively slow scanning rate compared to TOF. Here we present a novel gas phase separation and MS acquisition method called PulseDIA-MS, which improves the specificity and sensitivity of Orbitrap-based DIA analysis. This is achieved by dividing the ordinary DIA-MS analysis covering the entire mass range into multiple injections for DIA-MS analyses with complementary windows. Using standard HeLa digests, the PulseDIA method identified 69,530 peptide precursors from 9,337 protein groups with ten MS injections of 30 min LC gradient. The PulseDIA scheme containing two complementary windows led to the highest gain of peptide and protein identifications per time unit compared to the conventional 30 min DIA method. We further applied the method to profile the proteome of 18 cholangiocarcinoma (CCA) tissue samples (benign and malignant) from nine patients. PulseDIA identified 7,796 protein groups in these CCA samples, with 14% increase of protein identifications, compared to the conventional DIA method. The missing value for protein matrix dropped by 7% with PulseDIA acquisition. 681 proteins were significantly dysregulated in tumorous CCA samples. Together, we presented and benchmarked an alternative DIA method with higher sensitivity and lower missing rate.

cancer biology↗

bioRxiv2019

Streptococcus pyogenes Cas9 displays biased one-dimensional diffusion on dsDNA to search for a target

The RNA-guided Streptococcus pyogenes Cas9 (spCas9) is a sequence-specific DNA endonuclease that works as one of the most powerful genetic editing tools. However, how the Cas9 locates its target among huge amounts of dsDNAs remains controversial. Here, combining biochemical and single-molecule assays, we revealed that Cas9 uses both three-dimensional and one-dimensional diffusion to find its target. We further observed a surprising biased one-dimensional diffusion of Cas9 from 3 to 5 end of the non-target strand under physiological salt condition, whereas low ionic concentration or mutations on PAM recognition residues induce unbiased one-dimensional diffusion of Cas9 along dsDNA. We quantified the diffusion length of 27 bp, which accelerates the target search efficiency of Cas9 by [~] 10 folds. Our results reveal a unique searching mechanism of Cas9 at physiological salt conditions, and provide important guidance for both in-vitro and in-vivo applications of Cas9.

biochemistry↗

bioRxiv2019

Endosomal explosion induced by hypertonicity enhances chitosan transfection

Transferring DNA into cells to regulate cell function is a novel research field in recent decades. Chitosan is a gene vector with the properties of low-cost and safe, but high efficient delivery has remained challenging. We developed a strategy termed EEIH for endosomal explosion induced by hypertonicity, in which short-time exposure to hypertonic solutions triggers endosomes destabilization like explosions. EEIH can force chitosan/DNA polyplexes to break through the endosomal barriers to approach the nucleus, which results in boosting the transfection efficiency of chitosan in several cell lines. We demonstrate that EEIH is a significant and practical strategy in chitosan transfection system without sophisticated modification of chitosan.

biophysics↗

bioRxiv2019

Discovery and Multi-center Verification of Prostate Cancer Protein Biomarkers using Single-shot Short Gradient Microflow SWATH and MRMHR Mass Spectrometry

We report and evaluated a microflow, single-shot, short gradient SWATH MS method intended to accelerate the discovery and verification of protein biomarkers in clinical specimens. The method uses 15-min gradient microflow-LC peptide separation, an optimized SWATH MS window configuration and OpenSWATH software for data analysis.\n\nWe applied the method to a cohort 204 of FFPE prostate tissue samples from 58 prostate cancer patients and 10 prostatic hyperplasia patients. Altogether we identified 27,976 proteotypic peptides and 4,043 SwissProt proteins from these 204 samples. Compared to a reference SWATH method with 2-hour gradient the accelerated method consumed only 27% instrument time, quantified 80% proteins and showed reduced batch effects. 3,800 proteins were quantified by both methods in two different instruments with relatively high consistency (r = 0.77). 75 proteins detected by the accelerated method with differential abundance between clinical groups were selected for further validation. A shortlist of 134 selected peptide precursors from the 75 proteins were analyzed using MRM-HR, exhibiting high quantitative consistency with the 15-min SWATH method (r = 0.89) in the same sample set. We further verified the capacity of these 75 proteins in separating benign and malignant tissues (AUC = 0.99) in an independent prostate cancer cohort (n=154).\n\nOverall our data show that the single-shot short gradient microflow-LC SWATH MS method achieved about 4-fold acceleration of data acquisition with reduced batch effect and a moderate level of protein attrition compared to a standard SWATH acquisition method. Finally, the results showed comparable ability to separate clinical groups.

cancer biology↗

bioRxiv2019

NeuroPAL: A Neuronal Polychromatic Atlas of Landmarks for Whole-Brain Imaging in C. elegans

Comprehensively resolving single neurons and their cellular identities from whole-brain fluorescent images is a major challenge. We achieve this in C. elegans through the engineering and use of a multicolor transgene called NeuroPAL (a Neuronal Polychromatic Atlas of Landmarks). NeuroPAL worms share a stereotypical multicolor fluorescence map for the entire hermaphrodite nervous system that allows comprehensive determination of neuronal identities. Neurons labeled with NeuroPAL do not exhibit fluorescence in the green, cyan, or yellow emission channels, allowing the transgene to be used with numerous reporters of gene expression or neuronal dynamics. Here we showcase three studies that leverage NeuroPAL for nervous-system-wide neuronal identification. First, we determine the brainwide expression patterns of all metabotropic receptors for acetylcholine, GABA, and glutamate, completing a map of this communication network. Second, we uncover novel changes in cell fate caused by transcription factor mutations. Third, we record brainwide activity in response to attractive and repulsive chemosensory cues, characterizing multimodal coding and novel neuronal asymmetries for these stimuli. We present a software package that enables semi-automated determination of all neuronal identities based on color and positional information. The NeuroPAL framework and software provide a means to design landmark atlases for other tissues and organisms. In conclusion, we expect NeuroPAL to serve as an invaluable tool for gene expression analysis, neuronal fate studies, and for mapping whole-brain activity patterns.

neuroscience↗

bioRxiv2019

High-throughput proteomic analysis of FFPE tissue samples facilitates tumor stratification

Formalin-fixed, paraffin-embedded (FFPE), biobanked tissue samples offer an invaluable resource for clinical and biomarker research. Here we developed a pressure cycling technology (PCT)-SWATH mass spectrometry workflow to analyze FFPE tissue proteomes and applied it to the stratification of prostate cancer (PCa) and diffuse large B-cell lymphoma (DLBCL) samples. We show that the proteome patterns of FFPE PCa tissue samples and their analogous fresh frozen (FF) counterparts have a high degree of similarity and we confirmed multiple proteins consistently regulated in PCa tissues in an independent sample cohort. We further demonstrate temporal stability of proteome patterns from FFPE samples that were stored between one to 15 years in a biobank and show a high degree of the proteome pattern similarity between two types histological region of small FFPE samples, i.e. punched tissue biopsies and thin tissue sections of micrometer thickness, despite the existence of certain degree of biological variations. Applying the method to two independent DLBCL cohorts we identified myeloperoxidase (MPO), a peroxidase enzyme, as a novel prognostic marker. In summary, this study presents a robust proteomic method to analyze bulk and biopsy FFPE tissues and reports the first systematic comparison of proteome maps generated from FFPE and FF samples. Our data demonstrate the practicality and superiority of FFPE over FF samples for proteome in biomarker discovery. Promising biomarker candidates for PCa and DLBCL have been discovered.

biochemistry↗

bioRxiv2019

Single amino acid changes decouple two opposite activities of a viral auxiliary replication protein

We recently reported that the p28 auxiliary replication protein encoded by turnip crinkle virus (TCV) is also responsible for eliciting superinfection exclusion (SIE) against superinfecting TCV. However, it remains unresolved whether the replication function of p28 could be separated from its ability to elicit SIE. Here we report the identification of two single amino acid (aa) mutations that decouple these two functions. Using an Agrobacterium infiltration-based delivery system, we transiently expressed a series of p28 deletion and point mutants, and tested their ability to elicit SIE against a co-introduced TCV replicon. We found that substituting alanine (A) for valine (V) and phenylalanine (F) at p28 positions 181 and 182, respectively, modestly compromised SIE in transiently expressed p28 derivatives. Upon incorporation into TCV replicons, V181A and F182A decoupled TCV replication and SIE diametrically. While V181A impaired SIE without detectably compromising replication, F182A abolished TCV replication but had no effect on SIE once the replication of the defective replicon was restored through complementation. Both mutations diminished accumulation of p28 protein, suggesting that p28 must reach a concentration threshold in order to elicit a strong SIE. Importantly, the severe reduction of F182A protein levels correlated with a dramatic loss in the number of intracellular p28 foci formed by p28-p28 interactions. Together these findings not only decouples the replication and SIE functions of p28, but also unveils a concentration dependence for p28 coalescence and SIE elicitation. These data further highlight the role of p28 multimerization in driving the exclusion of secondary TCV infections.\n\nIMPORTANCESuperinfection exclusion (SIE) insulates virus-infected cells from subsequent invasion by the same or closely related viruses. SIE has been observed in both animal and plant virus-infected cells. Therefore, a thorough understanding of how SIE is achieved at the molecular level is expected to inspire novel strategies for combating virus infections in humans, animals, and plants. Our group has been using turnip crinkle virus (TCV) to elucidate the molecular interactions critical for SIE elicitation. The current study builds on the previous observation that TCV SIE is elicited by one single TCV-encoded protein (p28), and further identifies key regions and amino acids that are needed for SIE. We unravel key amino acid changes that decouple the replication and SIE functions of p28, and provides novel mechanistic insights of SIE.

molecular biology↗

bioRxiv2019

PRH1 mediates ARF7-LBD dependent auxin signaling to regulate lateral root development in Arabidopsis thaliana

The development of lateral roots in Arabidopsis thaliana is strongly dependent on signaling directed by the AUXIN RESPONSE FACTOR7 (ARF7), which in turn activates LATERAL ORGAN BOUNDARIES DOMAIN (LBD) transcription factors (LBD16, 18, 29 and 33). Here, the product of PRH1, a PR-1 homolog annotated previously as encoding a pathogen-responsive protein, was identified as a target of ARF7-mediated auxin signaling and also as participating in the development of lateral roots. PRH1 was shown to be strongly induced by auxin treatment, and plants lacking a functional copy of PRH1 formed fewer lateral roots. The transcription of PRH1 was controlled by the binding of both ARF7 and LBDs to its promoter region. An interaction was detected between PRH1 and GATA23, a protein which regulates cell identity in lateral root founder cells. Author SummaryIn Arabidopsis thaliana AUXIN RESPONSE FACTOR7 (ARF7)-mediated auxin signaling plays a key role in lateral roots (LRs) development. The LATERAL ORGAN BOUNDARIES DOMAIN (LBD) transcription factors (LBD16, 18, 29 and 33) act downstream of ARF7-mediated auxin signaling to control LRs formation. Here, the PR-1 homolog PRH1 was identified as a novel target of both ARF7 and LBDs (especially the LBD29) during auxin induced LRs formation, as both ARF7 and LBDs were able to bind to the PRH1 promoter. More interestingly, PRH1 has a physical interaction with GATA23, which has been also reported to be up-regulated by auxin and influences LR formation through its regulation of LR founder cell identity. Whether the interaction between GATA23 and PRH1 affects the stability and/or the activity of either (or both) of these proteins remains an issue to be explored. This study provides improves new insights about how auxin regulates lateral root development.

plant biology↗

bioRxiv2019

Complete Chloroplast Genome of Castanopsis sclerophylla (Lindl.) Schott: Genome Structures, Comparative and Phylogenetic Analysis

Castanopsis sclerophylla (Lindl.) Schott is an important species of evergreen broad-leaved forest in subtropical area and has important ecological and economic value. However, there are little studies on its chloroplast genome. In this study, the complete chloroplast genome sequences of C. sclerophylla was reported based on the Illumina Hiseq 2500 platform. The complete chloroplast genome of C. sclerophylla was 160,497bp, including a pair of inverted repeated (IRs) regions (25,675bp) that were separated by a large single copy (LSC) region of 90,255bp, and a small single copy (SSC) region of 18,892bp. The overall GC content of chloroplast genome was 36.82%. A total of 131 genes were found, of these 111 genes were unique and annotated, including 79 protein-coding genes, 27 transfer RNA genes (tRNAs), and four ribosomal RNA genes (rRNAs). Twenty-one genes were found to be duplicated in the IR regions. Comparative analysis indicated that IR contraction might be the reason for the relatively smaller chloroplast genome size of C. sclerophylla compared with other three congeneric species. Sequence analysis detected that the LSC and SSC regions were more divergent than the IR regions within the Castanopsis, furthermore, a higher divergence was found in non-coding regions than in coding regions. The maximum likelihood (ML) phylogenetic analysis showed that these four species of the genus Castanopsis formed a monophyletic clade and that C. sclerophylla is closely related to Castanopsis hainanensis with strong bootstrap values. These results not only provide basic knowledge about characteristics of C. sclerophylla and also enhance our understanding of Castanopsis species evolution within the Fagaceae family. Meanwhile, these findings will contribute to the exploration, utilization and conservation genetics of C. sclerophylla.

genomics↗

bioRxiv2018

G-Dash: A Genome Dashboard Integrating Modeling and Informatics

Genomics is a sequence based informatics science and a structure based molecular material science. There are few tools available that unite these approaches in a scientifically robust manner. Here we describe G-Dash, a web based prototype of a genomics dashboard, specifically designed to integrate informatics and 3D material studies of chromatin. G-Dash unites our Interactive Chromatin Modeling(ICM) tools with the Biodalliance genome browser and the JSMol molecular viewer to rapidly fold any DNA sequence into atomic or coarse-grained models of DNA, nucleosomes or chromatin. As a chromatin modeling tool, G-Dash enables users to specify nucleosome positions from various experimental or theoretical sources, interactively manipulate nucleosomes, and assign different conformational states to each nucleosome. As an informatics tool, data associated with 3D structures are displayed as tracks in a genome browser. The exchange of data between informatics and structure is bi-directional so any informatics track can inform a molecular structure (e.g. color by function) and structure features can be displayed as informatics tracks in a genome browser(e.g. Roll, Slide, or Twist). As a sample application, models of the CHA1 promoter based on experimentally determined nucleosome positions are explored with G-Dash. Steric clashes and DNA knotting are observed but can be resolved with G-Dashs minimal coarse-grained model without significant variation in structure. Results raise questions about the interpretation of nucleosome positioning data and promoter structures. In this regard, G-Dash is a novel tool for investigating structure-function relationships for regions of the genome ranging from base pairs to chromosomes and for generating, validating and testing mechanistic hypotheses. http://dna.engr.latech.edu/[~]gdash/GDash-landing-page

genomics↗

bioRxiv2018

Mechanism of Mycolactone Toxin Membrane Permeation: Atomistic vs Coarse-Grained MARTINI Simulations

Mycolactone, a cytotoxic and immunosuppressive macrolide produced by Mycobacterium ulcerans, is the central virulent factor in the skin disease Buruli ulcer. This multifunctional cytotoxin affects fundamental cellular processes such as cell adhesion, immune response and cell death by targeting various cellular structures. Developing effective diagnostics that target mycolactone has been challenging, potentially due to suspected interactions with lipophilic architectures, including membranes. To better understand the pathogenesis of Buruli ulcer disease, aid in the development of diagnostics, and learn how amphiphiles in general use lipid trafficking to navigate the host environment, we seek to understand the nature of mycolactone-membrane interactions. Herein we characterize how the two dominant isomers of mycolactone (A and B) interact with and permeate DPPC membranes with all-atom molecular dynamics simulations employing transition tempered metadynamics, and compare these results with those obtained by MARTINI coarse-grained simulations. Our all-atom simulations reveal that both isomers have a strong preference to associate with the membrane, although their mechanisms and energetics of membrane permeation differ slightly. Water molecules are found to play an important role in the permeation process. Although the MARTINI coarse-grained simulations give the correct free energy of membrane association, they fail to capture the mechanism of permeation and role of water during permeation as seen in all-atom simulations.

biophysics↗

bioRxiv2018

Fall Risk Prediction in Multiple Sclerosis Using Postural Sway Measures, A Machine Learning Approach

BackgroundBalance impairment affects over 75% of individuals with multiple sclerosis (MS), and leads to an increased risk of falling. Numerous postural sway metrics have been shown to be sensitive to balance impairment and fall risk in individuals with MS. Yet, there are no guidelines concerning the most appropriate postural sway metrics to monitor impairment. This investigation implemented a machine learning approach to assess the accuracy and feature importance of various postural sway metrics to differentiate individuals with MS from healthy controls as a function of physiological fall risk.\n\nMethodsThis secondary data analysis included 153 participants (50 controls and 103 individuals with MS) who underwent posturography based balance assessment (30s eyes open standing on a force platform) and physiological fall risk assessment (Physiological Profile Assessment - PPA). Participants were further classified into four subgroups based on fall risk: controls (n=50, 64.9 {+/-} 4.9 years old, PPA < 1); low-risk MS (n=34, 54.0 {+/-} 13.1 years old, PPA < 1); moderate-risk MS (n=27, 58.3 {+/-} 8.3 years old, 1 [≤] PPA < 2); high-risk MS (n=42, 56.8 {+/-} 9.7 years old, PPA [≥] 2). Twenty common sway metrics were derived following standard procedures, and subsequently used to train a machine learning algorithm (random forest - RF, with 10-fold cross validation) to predict individuals fall risk grouping. The feature importance from the RF algorithms was used to select the strongest sway metric for fall risk prediction.\n\nResults and DiscussionThe sway-metric based RF classifier had high classification accuracy in discriminating controls from MS individuals (> 86%). Sway sample entropy, a sway regularity metric, was identified as the strongest feature for classification of low-risk MS individuals from healthy controls. Whereas for all other comparisons, mediolateral sway amplitude was identified as the strongest predictor for fall risk groupings. These findings may set the foundation for the development of guidelines for reporting balance impairment in individuals with MS.

bioengineering↗

bioRxiv2018

Powerful gene set analysis in GWAS with the Generalized Berk-Jones statistic

A common complementary strategy in Genome-Wide Association Studies (GWAS) is to perform Gene Set Analysis (GSA), which tests for the association between one phenotype of interest and an entire set of Single Nucleotide Polymorphisms (SNPs) residing in selected genes. While there exist many tools for performing GSA, popular methods often include a number of ad-hoc steps that are difficult to justify statistically, provide complicated interpretations based on permutation inference, and demonstrate poor operating characteristics. Additionally, the lack of gold standard gene set lists can produce misleading results and create difficulties in comparing analyses even across the same phenotype. We introduce the Generalized Berk-Jones (GBJ) statistic for GSA, a permutation-free parametric framework that offers asymptotic power guarantees in certain set-based testing settings. To adjust for confounding introduced by different gene set lists, we further develop a GBJ step-down inference technique that can discriminate between gene sets driven to significance by single genes and those demonstrating group-level effects. We compare GBJ to popular alternatives through simulation and re-analysis of summary statistics from a large breast cancer GWAS, and we show how GBJ can increase power by incorporating information from multiple signals in the same gene. In addition, we illustrate how breast cancer pathway analysis can be confounded by the frequency of FGFR2 in pathway lists. Our approach is further validated on two other datasets of summary statistics generated from GWAS of height and schizophrenia.

genetics↗

bioRxiv2018

Scalable approximate Bayesian inference for particle tracking data

Many important datasets in physics, chemistry, and biology consist of noisy sequences of images of multiple moving overlapping particles. In many cases, the observed particles are indistinguishable, leading to unavoidable uncertainty about nearby particles identities. Exact Bayesian inference is intractable in this setting, and previous approximate Bayesian methods scale poorly. Non-Bayesian approaches that output a single \"best\" estimate of the particle tracks (thus discarding important uncertainty information) are therefore dominant in practice. Here we propose a flexible and scalable amortized approach for Bayesian inference on this task. We introduce a novel neural network method to approximate the (intractable) filter-backward-sample-forward algorithm for Bayesian inference in this setting. By varying the simulated training data for the network, we can perform inference on a wide variety of data types. This approach is therefore highly flexible and improves on the state of the art in terms of accuracy; provides uncertainty estimates about the particle locations and identities; and has a test run-time that scales linearly as a function of the data length and number of particles, thus enabling Bayesian inference in arbitrarily large particle tracking datasets.

bioinformatics↗