Biology subjects

Mehta, P.

Publications and source records attributed to Mehta, P..

At least 37 records · Page 2Linked to original sources

bioRxiv2023

Engineering synthetic phosphorylation signaling networks in human cells

Protein phosphorylation signaling networks play a central role in how cells sense and respond to their environment. Here, we describe the engineering of artificial phosphorylation networks in which "push-pull" motifs--reversible enzymatic phosphorylation cycles consisting of opposing kinase and phosphatase activities--are assembled from modular protein domain parts and then wired together to create synthetic phosphorylation circuits in human cells. We demonstrate that the composability of our design scheme enables model-guided tuning of circuit function and the ability to make diverse network connections; synthetic phosphorylation circuits can be coupled to upstream cell surface receptors to enable fast-timescale sensing of extracellular ligands, while downstream connections can regulate gene expression. We leverage these capabilities to engineer cell-based cytokine controllers that dynamically sense and suppress activated T cells. Our work introduces a generalizable approach for designing and building phosphorylation signaling circuits that enable user-defined sense-and- respond function for diverse biosensing and therapeutic applications.

synthetic biology↗

bioRxiv2023

Dysregulated actin dynamics and cofilin correlate with TDP-43 pathology in sporadic amyotrophic lateral sclerosis

Amyotrophic lateral sclerosis (ALS) is a fatal, rapidly progressive neurodegenerative disorder affecting motor neurons, that overlaps significantly with frontotemporal dementia (FTD). Most cases are sporadic (90%) with undefined aetiology, but pathological forms of TAR-binding protein 43 (TDP-43), involving its misfolding, aggregation and mislocalisation from the nucleus to the cytoplasm, are present in motor neurons in almost all cases (97%) and [~]45% FTD cases. Actin is the most abundant protein in eukaryotic cells, with structural roles in the cytoskeleton and diverse signalling functions. This includes neuronal-specific roles in dendritic spines, synapses, axonal growth cones, and plasticity. Actin is in constant dynamic equilibrium between two forms: free monomeric, globular actin (G-actin) and polymeric, filamentous actin (F-actin). Actin dynamics is regulated by several key actin-binding proteins, including tropomyosin 4.2 (Tpm4.2) and cofilin, which depolymerises actin filaments. Cofilin is activated by phosphorylation at Ser3 via LIM domain kinase1/2 (LIMK1/2), which is also regulated by phosphorylation via Rac1/cdc42. Here we demonstrate that actin dynamics is closely associated with pathological TDP-43 in ALS. More F-actin relative to G-actin was detected in lumbar spinal cords from both sporadic ALS patients and a mouse model displaying TDP-43 pathology (rNLS), and in neuronal cells expressing cytoplasmic TDP-43. Hence actin dynamics is dysregulated in sporadic ALS, resulting in more actin polymerization. We also detected increased levels of Tpm 4.2, Rac1/cdc42, and increased phosphorylation of both LIMK1/2 and cofilin, in sporadic ALS patients. TDP-43 also physically interacted with actin in vitro and in cell lysates, providing additional insights into actin dysregulation in ALS. rNLS mice display motor neuron loss and key ALS/MND behavioural phenotypes, and increased cofilin phosphorylation was also detected in these animals at symptom onset, implying that actin dynamics actively contributes to neurodegeneration. Moreover, pharmacological induction of actin polymerization produced features typical of pathological TDP-43 (cytoplasmic mis-localisation and formation of inclusions and stress granules) implying that actin dysregulation contributes to TDP-43 pathology in ALS. Importantly, we also detected more cofilin phosphorylation in spinal motor neurons from sporadic patients compared to healthy controls, revealing that our observations are clinically relevant and present in the relevant cell type. This study therefore identifies dysregulated actin dynamics as a novel disease mechanism associated with TDP-43 pathology and hence most ALS cases. It also implies that regulating cofilin or LIMK1/2 phosphorylation may be a novel therapeutic strategy in ALS, FTD and other diseases involving TDP-43 pathology.

neuroscience↗

bioRxiv2023

A transient protein folding response targets aggregation in the early phase of TDP-43-mediated disease

Understanding the mechanisms that drive TDP-43 pathology is integral to combating neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD). To address this, we sought to determine the timeline of proteomic alterations across disease course in TDP-43 proteinopathy. Using longitudinal quantitative proteomics analysis of cortex samples from the cytoplasmic TDP-43 rNLS8 mouse model of ALS and FTLD, we identified several distinct protein subsets characterized by temporal alterations in protein abundance across diverse biological pathways, including protein folding, intracellular transport, myelination, and neuronal synaptic function. Remarkably, neurons in the rNLS8 cortex elicited a transitory response primarily comprising protein-folding factors prior to and in the earliest stages of disease progression. This response included increased levels of DnaJ homolog subfamily B member 5, DNAJB5, and proof-of-concept studies showed that DNAJB5 over-expression decreased TDP-43 aggregation in cell and cortical neuron cultures. Conversely, knockout of Dnajb5 exacerbated motor impairments caused by AAV-mediated cytoplasmic TDP-43 expression in the brains and spinal cords of mice. Lastly, the late disease proteomic signatures of rNLS8 mouse cortex strongly correlated with changes in human autopsy-derived TDP-43 proteinopathy tissues, indicating commonality of disease processes. Together, these findings reveal molecular mechanisms that regulate protein levels through distinct stages of ALS and FTLD progression, and suggest that protein folding factors that combat cytoplasmic TDP-43 protein aggregation could be protective in disease. HighlightsO_LIThe first longitudinal map of the cortex proteome throughout TDP-43-driven disease in a mouse model of cytoplasmic TDP-43 proteinopathy (rNLS8 mice). C_LIO_LICytoplasmic TDP-43 accumulation drives many dynamic changes to the cortex proteome, including increases in protein folding factors prior to disease onset. C_LIO_LIThe protein folding factor DNAJB5 decreases TDP-43 aggregation in HEK293 cells and primary cortical neurons and Dnajb5 knockout exacerbates cytoplasmic TDP-43-associated motor impairments in vivo. C_LIO_LIThe proteomic signature of the rNLS8 mouse cortex correlates strongly with postmortem brain tissue from human TDP-43 proteinopathies. C_LIO_LIA new webtool, TDP-map (https://shiny.rcc.uq.edu.au/TDP-map/), allows comparison of transcriptomic and proteomic datasets from mouse and human TDP-43 proteinopathy. C_LI

neuroscience↗

bioRxiv2023

Multiplexed fluorescence and scatter detection with single cell resolution using on-chip fiber optics for droplet microfluidic applications

Droplet microfluidics has emerged as a critical component of several high-throughput single cell analysis techniques in biomedical research and diagnostics. However, while there has been significant progress in the individual assays being developed, multi-parametric optical sensing of droplets and their encapsulated contents remains challenging. The current common approach of microscopy based high-speed imaging of droplets is technically complex and requires expensive instrumentation limiting their large-scale adoption. Here we have adapted the principles of the widely established technique of flow cytometry to a novel optofluidic setup - the OptiDrop platform, with on-chip detection of scatter and multiple fluorescence signals from microfluidic droplets and their contents using optical fibers. The highly customizable on-chip optical fiber-based signal detection system enables simplified, miniaturized, low-cost, multi-parametric sensing of optical signals with high sensitivity and single cell resolution within each droplet. Our work on differential expression analysis of the Major Histocompatibility Complex (MHC) protein in response to IFN{gamma} stimulation further demonstrates the OptiDrop platforms capabilities in sensitively detecting cell surface biomarkers using fluorescently labeled antibodies. The OptiDrop platform thus combines the versatility of flow cytometry with the power of droplet microfluidics to provide wide-ranging optical sensing solutions for research and diagnostics.

bioengineering↗

bioRxiv2023

Geometry of ecological coexistence and niche differentiation

A fundamental problem in ecology is to understand how competition shapes biodiversity and species coexistence. Historically, one important approach for addressing this question has been to analyze Consumer Resource Models (CRMs) using geometric arguments. This has led to broadly applicable principles such as Tilmans R* and species coexistence cones. Here, we extend these arguments by constructing a novel geometric framework for understanding species coexistence based on convex polytopes in the space of consumer preferences. We show how the geometry of consumer preferences can be used to predict species coexistence and enumerate ecologically-stable steady states and transitions between them. Collectively, these results constitute a qualitatively new way of understanding the role of species traits in shaping ecosystems within niche theory.

ecology↗

bioRxiv2023

Ultra-high throughput mapping of genetic design space

Massively parallel genetic screens have been used to map sequence-to-function relationships for a variety of genetic elements. However, because these approaches only interrogate short sequences, it remains challenging to perform high throughput (HT) assays on constructs containing combinations of multiple sequence elements arranged across multi-kb length scales. Overcoming this barrier could accelerate synthetic biology; by screening diverse gene circuit designs and learning "composition-to-function" mappings that reveal genetic part composability rules and enable rapid identification of behavior-optimized variants. Here, we introduce CLASSIC, a genetic screening platform that combines long- and short-read next-generation sequencing (NGS) modalities to quantitatively assess pools of constructs of arbitrary length containing diverse part compositions. We show that CLAS-SIC can measure expression profiles of >105 gene circuit designs (from 5-20 kb) in a single experiment in human cells. The resulting datasets can be used to train ML models that accurately predict circuit behavior across expansive circuit design landscapes, revealing part composability rules that govern circuit performance. Our work shows that by expanding the throughput of each design-build-test-learn (DBTL) cycle, CLASSIC enhances the pace and scale of synthetic biology and establishes an experimental basis for data-driven design of complex genetic systems.

synthetic biology↗

bioRxiv2023

scTOP: physics-inspired order parameters for cellular identification and visualization

Advances in single-cell RNA-sequencing (scRNA-seq) provide an unprecedented window into cellular identity. The increasing abundance of data requires new theoretical and computational frameworks for understanding cell fate determination, accurately classifying cell fates from expression data, and integrating knowledge from cell atlases. Here, we present single-cell Type Order Parameters (scTOP): a statistical-physics-inspired approach for constructing "order parameters" for cell fate given a reference basis of cell types. scTOP can quickly and accurately classify cells at a single-cell resolution, generate interpretable visualizations of developmental trajectories, and assess the fidelity of engineered cells. Importantly, scTOP does this without using feature selection, statistical fitting, or dimensional reduction (e.g., UMAP, PCA, etc.). We illustrate the power of scTOP utilizing a wide variety of human and mouse datasets (both in vivo and in vitro). By reanalyzing mouse lung alveolar development data, we characterize a transient perinatal hybrid alveolar type 1/alveolar type 2 (AT1/AT2) cell population that disappears by 15 days post-birth and show that it is transcriptionally distinct from previously identified adult AT2-to-AT1 transitional cell types. Visualizations of lineage tracing data on hematopoiesis using scTOP confirm that a single clone can give rise to as many as three distinct differentiated cell types. We also show how scTOP can quantitatively assess the transcriptional similarity between endogenous and transplanted cells in the context of murine pulmonary cell transplantation. Finally, we provide an easy-to-use Python implementation of scTOP. Our results suggest that physics-inspired order parameters can be an important tool for understanding development and characterizing engineered cells.

biophysics↗

bioRxiv2023

The emergence of goblet inflammatory or ITGB6hi nasal progenitor cells determines age-associated SARS-CoV-2 pathogenesis

Children infected with SARS-CoV-2 rarely progress to respiratory failure, but the risk of mortality in infected people over 85 years of age remains high, despite vaccination and improving treatment options. Here, we take a comprehensive, multidisciplinary approach to investigate differences in the cellular landscape and function of paediatric (<11y), adult (30- 50y) and elderly (>70y) nasal epithelial cells experimentally infected with SARS-CoV-2. Our data reveal that nasal epithelial cell subtypes show different tropism to SARS-CoV-2, correlating with age, ACE2 and TMPRSS2 expression. Ciliated cells are a viral replication centre across all age groups, but a distinct goblet inflammatory subtype emerges in infected paediatric cultures, identifiable by high expression of interferon stimulated genes and truncated viral genomes. In contrast, infected elderly cultures show a proportional increase in ITGB6hi progenitors, which facilitate viral spread and are associated with dysfunctional epithelial repair pathways. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=155 SRC="FIGDIR/small/524211v2_ufig1.gif" ALT="Figure 1"> View larger version (61K): org.highwire.dtl.DTLVardef@dab12aorg.highwire.dtl.DTLVardef@1a57334org.highwire.dtl.DTLVardef@12e7983org.highwire.dtl.DTLVardef@2bbe6e_HPS_FORMAT_FIGEXP M_FIG C_FIG

cell biology↗

bioRxiv2022

Root Walker: an automated pipeline for large scale quantification of early root growth responses at high spatial and temporal resolution

Plants are sessile organisms that constantly adapt to their changing environment. The root is exposed to numerous environmental signals ranging from nutrients and water to microbial molecular patterns. These signals can trigger distinct responses including the rapid increase or decrease of root growth. Consequently, using root growth as a readout for signal perception can help decipher which external cues are perceived by roots, and how these signals are integrated. To date, studies measuring root growth responses using large numbers of roots have been limited by a lack of high-throughput image acquisition, poor scalability of analytical methods, or low spatiotemporal resolution. Here, we developed the Root Walker pipeline, which uses automated microscopes to acquire time-series images of many roots exposed to controlled treatments with high-spatiotemporal-resolution, in conjunction with fast and automated image analysis software. We demonstrate the power of Root Walker by quantifying root growth rate responses at different time and throughput scales upon treatments with natural auxin, and upon treatment with two mitogen-associated protein kinase cascade inhibitors. We find a concentration-dependent root growth response to auxin and reveal the specificity of one MAPK inhibitor. We further demonstrate the ability of Root Walker for conducting genetic screens by performing a genome wide association study on 260 accessions under 2 weeks. This revealed known and unknown root growth regulators. Root Walker promises to be a useful toolkit for the plant science community, allowing large-scale screening of root growth dynamics for a variety of purposes, including genetic screens for root sensing and root growth response mechanisms.

plant biology↗

bioRxiv2022

Durable alveolar engraftment of PSC-derived lung epithelial cells into immunocompetent mice

Durable reconstitution of the injured distal lung epithelium with pluripotent stem cell (PSC) derivatives, if realized, would represent a promising potential therapy for diseases that result from alveolar damage. Here we differentiate murine PSCs in vitro into self-renewing lung epithelial progenitors able to engraft into the injured distal lung epithelium of immunocompetent, syngeneic mouse recipients. Emulating the roadmap of the developing embryo, we generate transplantable PSC-derived Nkx2-1+/Sox9+ lung epithelial progenitors that are highly similar to cultured primary embryonic distal lung bud tip progenitors. These cells display a stable phenotype after frozen archiving or extensive expansion in culture, providing a nearly inexhaustible source of cells that can be engrafted into syngeneic injured mouse lungs without the need for immunosuppression. After transplantation PSC-derived tip-like progenitors downregulate Sox9 and mature in the distal lung, upregulating alveolar type 2 cell markers or assuming the flat morphology and molecular phenotype of terminally differentiated alveolar type 1 cells. After months in vivo, donor-derived cells retain their alveolar epithelial type 2-like and type 1-like phenotypes, as characterized by single cell RNA sequencing, ultrastructural analyses, in vivo histologic profiling, and ex vivo organoid assays that demonstrate continued capacity of the engrafted cells to proliferate and differentiate. These results indicate durable reconstitution of the distal lungs facultative progenitor and differentiated epithelial cell compartments in vivo with PSC-derived cells, thus establishing a novel model for pulmonary cell therapy which can be utilized to better understand the mechanisms and utility of engraftment prior to future clinical studies.

cell biology↗

bioRxiv2021

Re-assessing the phylogenetic status and evolutionary relationship of Forest Owlet using genomic data

Phylogenetic relationships are often challenging to resolve in recent/younger lineage when only a few loci are used. Ultra Conserved Elements (UCE) are highly conserved regions across taxa that help resolve shallow and deep divergences. We utilized UCEs harvested from whole genomes to assess the phylogenetic position and taxonomic affiliation of an endangered endemic owlet in the family Strigidae - the Forest Owlet Athene blewitti. The taxonomic placement of this species has been revised multiple times. A multigene study attempted to address the question but showed a discrepancy across datasets in its placement of the species within genus Athene. We assembled a dataset of 5018 nuclear UCE loci with increased taxon sampling. Forest Owlet was found to be an early split from the Athene clade but sister to other Athene; and consistent across three approaches - maximum likelihood, bayesian, and the multispecies coalescence. Divergence dating using fossil calibrations suggest that the Athene lineage split from its ancestor about 7.6Mya, and the Forest Owlet diverged about 5.2Mya, consistent with previous multigene approaches. Despite osteological differences from other Athene, we suggest the placement of the Forest Owlet as a member of the Athene to emphasize its evolutionary relationship. Graphical Abstract O_FIG_DISPLAY_L [Figure 1] M_FIG_DISPLAY C_FIG_DISPLAY HIGHLIGHTSO_LIPhylogenomics using genome-wide nuclear markers yielded a well-supported topology for Athene and Glaucidium lineages. C_LIO_LIThree different methods of phylogenetic tree construction showed that Forest Owlet is an early split from all other Athene species. C_LIO_LIDivergence dating in the bayesian framework puts the Forest Owlet age between 5.0my to 5.5my. C_LI

evolutionary biology↗

bioRxiv2021

Cross-feeding shapes both competition and cooperation in microbial ecosystems

Recent work suggests that cross-feeding - the secretion and consumption of metabolic biproducts by microbes - is essential for understanding microbial ecology. Yet how cross-feeding and competition combine to give rise to ecosystem-level properties remains poorly understood. To address this question, we analytically analyze the Microbial Consumer Resource Model (MiCRM), a prominent ecological model commonly used to study microbial communities. Our mean-field solution exploits the fact that unlike replicas, the cavity method does not require the existence of a Lyapunov function. We use our solution to derive new species-packing bounds for diverse ecosystems in the presence of cross-feeding, as well as simple expressions for species richness and the abundance of secreted resources as a function of cross-feeding (metabolic leakage) and competition. Our results show how a complex interplay between competition for resources and cooperation resulting from metabolic exchange combine to shape the properties of microbial ecosystems.

ecology↗

bioRxiv2021

Robust coordination of collective oscillatory signaling requires single-cell excitability and fold-change detection

Complex multicellular behaviors are coordinated at the level of biochemical signaling networks, yet how this decentralized mechanism enables robust control in variable environments and over many orders of magnitude of spatiotemporal scales remains an open question. A stunning example of these behaviors is found in the microbe Dictyostelium discoideum, which uses the small molecule cyclic AMP (cAMP) to drive the propagation of collective signaling oscillations leading to multicellular development. The critical design features of the Dictyostelium signaling network remain unclear despite decades of mathematical modeling and experimental interrogation because each model makes different assumptions about the network architecture and in general, normalizing models for direct comparison presents a major challenge. We overcome this challenge by using recent experimental data to normalize the time and response scales of five major signal relay network models to one another and assess their ability to recapitulate experimentally-observed population and single-cell dynamics. We find that to successfully reproduce the full range of observed dynamical behaviors, single cells must be excitable and respond to the relative fold-change of environmental signals. This suggests these features represent robust principles for coordinating cellular populations through oscillatory signaling and that single-cell excitable dynamics are a generalizable route for controlling population behaviors.

systems biology↗

bioRxiv2021

Scalable Clustering with Supervised Linkage Methods

Data clustering plays a significant role in biomedical sciences, particularly in single-cell data analysis. Researchers use clustering algorithms to group individual cells into populations that can be evaluated across different levels of disease progression, drug response, and other clinical statuses. In many cases, multiple sets of clusters must be generated to assess varying levels of cluster specificity. For example, there are many subtypes of leukocytes (e.g. T cells), whose individual preponderance and phenotype must be assessed for statistical/functional significance. In this report, we introduce a novel hierarchical density clustering algorithm (HAL-x) that uses supervised linkage methods to build a cluster hierarchy on raw single-cell data. With this new approach, HAL-x can quickly predict multiple sets of labels for immense datasets, achieving a considerable improvement in computational efficiency on large datasets compared to existing methods. We also show that cell clusters generated by HAL-x yield near-perfect F1-scores when classifying different clinical statuses based on single-cell profiles. Our hierarchical density clustering algorithm achieves high accuracy in single cell classification in a scalable, tunable and rapid manner. We make HAL-x publicly available at: https://pypi.org/project/hal-x/

bioinformatics↗

bioRxiv2021

Phenotyping the tumor-immune microenvironment (TiME) in vivo by cellular-level optical imaging reveals unique combination phenotypes with variable inflammation and endothelial anergy

Immunotherapies have shown unprecedented clinical benefits in several malignancies1-3. However, clinical responses remain variable and unpredictable, indicating the need to develop predictive platforms that can improve patient stratification4. Phenotyping of tumors into hot, altered, or cold5 based on assessment of only T-cell infiltration in static tumor biopsies provides suboptimal prediction of immunotherapy response6,7. In vivo dynamic mechanisms within the tumor microenvironment such as tumor angiogenesis and leukocyte trafficking5,8,9 also play a central role in modulating anti-tumor immunity and therefore immunotherapy response. Here, we report novel tumor immune microenvironment (TiME) phenotyping in vivo in patients with non-invasive spatially-resolved cellular-level imaging based on endogenous contrast. Investigating skin cancers as a model, with reflectance confocal microscopy (RCM) imaging10, we determined four major phenotypes with variable prevalence of vasculature (Vasc) and inflammation (Inf) features: VaschiInfhi, VaschiInflo, VascloInfhi and Vascmed/hiInflo. The VaschiInfhi phenotype correlates with high immune activation, exhaustion, and vascular signatures while VaschiInflo with endothelial anergy and immune exclusion. Automated quantification of TiME features demonstrates moderate-high accuracy and correlation with corresponding gene expression. Prospectively analyzed response to topical immunotherapy show highest response in VascloInfhi, and reveals the added value of vascular features in predicting treatment response. Our novel in vivo cellular-level imaging and phenotyping approach can potentially advance our fundamental understanding of TiME, develop robust predictors for immunotherapy outcomes and identify novel targetable pathways in future.

cancer biology↗

bioRxiv2021

Severe Acute Respiratory Syndrome Coronavirus-2 genome sequence variations relate to morbidity and mortality in Coronavirus Disease-19

Outcome of infection with Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2) may depend on the host, virus or the host-virus interaction-related factors. Complete SARS-CoV-2 genome was sequenced using Illumina and Nanopore platforms from naso-/oro-pharyngeal ribonucleic acid (RNA) specimens from COVID-19 patients of varying severity and outcomes, including patients with mild upper respiratory symptoms (n=35), severe disease ad-mitted to intensive care with respiratory and gastrointestinal symptoms (n=21), fatal COVID-19 outcome (n=17) and asymptomatic (n=42). Of a number of genome variants observed, p.16L>L (Nsp1), p.39C>C (Nsp3), p.57Q>H (ORF3a), p.71Y>Y (Membrane glycoprotein), p.194S>L (Nucleocapsid protein) were observed in similar frequencies in different patient subgroups. However, seventeen other variants were observed only in symptomatic patients with severe and fatal COVID-19. Out of the latter, one was in the 5UTR (g.241C>T), eight were synonymous (p.14V>V and p.92L>L in Nsp1 protein, p.226D>D, p.253V>V, and p.305N>N in Nsp3, p.34G>G and p.79C>C in Nsp10 protein, p.789Y>Y in Spike protein), and eight were non-synonymous (p.106P>S, p.157V>F and p.159A>V in Nsp2, p.1197S>R and p.1198T>K in Nsp3, p.97A>V in RdRp, p.614D>G in Spike protein, p.13P>L in nucleocapsid). These were completely absent in the asymptomatic group. SARS-CoV-2 genome variations have a significant impact on COVID-19 presentation, severity and outcome.

genomics↗

bioRxiv2020

Signatures of processing complexity during global cognitive states in ventromedial prefrontal cortex

Behavioral neuroscience almost exclusively studies behavior during tasks and ignores the unstructured inter-trial interval (ITI). However, it is unlikely that the ITI is simply an idling or paused mode; instead, it is a likely time for globally focused cognition, in which attention is disengaged from the task at hand and oriented more broadly. To gain insight into the computational underpinnings of globally focused cognition, we recorded from neurons in a core decision-making region, area 14 of ventromedial prefrontal cortex (vmPFC), as macaques performed a foraging search task with long inter-trial intervals (ITIs). We find that during the ITI, ensemble firing is associated with increased discriminability of a key mnemonic variable, recent reward rate, which in turn predicts upcoming search strategy. ITI activity is also associated with increased ensemble dimensionality and faster subspace reorganization, presumed markers of processing complexity. These results demonstrate the flexible nature of mnemonic processing and support the idea that the brain makes use of ostensible downtime to engage in complex processing.

animal behavior and cognition↗

bioRxiv2020

Automated detection of glaucoma with interpretable machine learning using clinical data and multi-modal retinal images

Glaucoma, the leading cause of irreversible blindness worldwide, is a disease that damages the optic nerve. Current machine learning (ML) approaches for glaucoma detection rely on features such as retinal thickness maps; however, the high rate of segmentation errors when creating these maps increase the likelihood of faulty diagnoses. This paper proposes a new, comprehensive, and more accurate ML-based approach for population-level glaucoma screening. Our contributions include: (1) a multi-modal model built upon a large data set that includes demographic, systemic and ocular data as well as raw image data taken from color fundus photos (CFPs) and macular Optical Coherence Tomography (OCT) scans, (2) model interpretation to identify and explain data features that lead to accurate model performance, and (3) model validation via comparison of model output with clinician interpretation of CFPs. We also validated the model on a cohort that was not diagnosed with glaucoma at the time of imaging but eventually received a glaucoma diagnosis. Results show that our model is highly accurate (AUC 0.97) and interpretable. It validated biological features known to be related to the disease, such as age, intraocular pressure and optic disc morphology. Our model also points to previously unknown or disputed features, such as pulmonary capacity and retinal outer layers.

bioengineering↗