Biology subjects

Lee, A. Y.

Publications and source records attributed to Lee, A. Y..

9 recordsLinked to original sources

bioRxiv2018

DNA replication-mediated error correction of ectopic CENP-A deposition maintains centromere identity

Chromatin assembled with the histone H3 variant CENP-A is the heritable epigenetic determinant of human centromere identity. Using genome-wide mapping and reference models for 23 human centromeres, CENP-A is shown in early G1 to be assembled into nucleosomes within megabase, repetitive -satellite DNAs at each centromere and onto 11,390 transcriptionally active sites on the chromosome arms. Here we identify that DNA replication acts as an error correction mechanism to sustain centromere identity through the removal of the sites of CENP-A loading on the chromosome arms, while maintaining centromere-bound CENP-A with the same DNA sequence preferences as in its initial loading.

cell biology↗

bioRxiv2018

Multilabel multiclass classification of OCT images augmented with age, gender and visual acuity data

Optical Coherence Tomography (OCT) imaging of the retina is in widespread clinical use to diagnose a wide range of retinal pathologies and several previous studies have used deep learning to create systems that can accurately classify retinal OCT as indicative of one of these pathologies. However, patients often exhibit multiple pathologies concurrently. Here, we designed a novel neural network algorithm that performs multiclass and multilabel classification of retinal images from OCT images in four common retinal pathologies: epiretinal membrane, diabetic macular edema, dry age-related macular degeneration and neovascular age-related macular degeneration. Furthermore, clinicians often also use additional information about the patient for diagnosis. Second contribution of this study is improvement of multiclass, multilabel classification augmented with information about the patient: age, visual acuity and gender. We compared two training strategies: a network pre-trained with ImageNet was used for transfer learning, or the network was trained from randomly initialized weights. Transfer learning does not perform better in this case, because many of the low-level filters are tuned to colors, and the OCT images are monochromatic. Finally, we provide a transparent and interpretable diagnosis by highlighting the regions recognized by the neural network.

physiology↗

bioRxiv2018

Forecasting Future Humphrey Visual Fields Using Deep Learning

PurposeTo determine if deep learning networks could be trained to forecast a future 24-2 Humphrey Visual Field (HVF).\n\nDesignRetrospective database study.\n\nParticipantsAll patients who obtained a HVF 24-2 at the University of Washington.\n\nMethodsAll datapoints from consecutive 24-2 HVFs from 1998 to 2018 were extracted from a University of Washington database. Ten-fold cross validation with a held out test set was used to develop the three main phases of model development: model architecture selection, dataset combination selection, and time-interval model training with transfer learning, to train a deep learning artificial neural network capable of generating a point-wise visual field prediction.\n\nMain outcome measuresMean absolute error (MAE) and difference in Mean Deviation (MD) between predicted and actual future HVF.\n\nResultsMore than 1.7 million perimetry points were extracted to the hundredth decibel from 32,443 24-2 HVFs. The best performing model with 20 million trainable parameters, CascadeNet-5, was selected. The overall MAE for the test set was 2.47 dB (95% CI: 2.45 dB to 2.48 dB). The 100 fully trained models were able to successfully predict progressive field loss in glaucomatous eyes up to 5.5 years in the future with a correlation of 0.92 between the MD of predicted and actual future HVF (p < 2.2 = 10-16) and an average difference of 0.41 dB.\n\nConclusionsUsing unfiltered real-world datasets, deep learning networks show an impressive ability to not only learn spatio-temporal HVF changes but also to generate predictions for future HVFs up to 5.5 years, given only a single HVF.

bioinformatics↗

bioRxiv2018

MRI2MRI: A deep convolutional network that accurately transforms between brain MRI contrasts

Different brain MRI contrasts represent different tissue properties and are sensitive to different artifacts. The relationship between different contrasts is therefore complex and nonlinear. We developed a deep convolutional network that learns the mapping between different MRI contrasts. Using a publicly available dataset, we demonstrate that this algorithm accurately transforms between T1- and T2-weighted images, proton density images, time-of-flight angiograms, and diffusion MRI images. We demonstrate that these transformed images can be used to improve spatial registration between MR images of different contrasts.

neuroscience↗

bioRxiv2017

Assessment of the need for separate test set and number of medical images necessary for deep learning: a sub-sampling study

Deep learning algorithms have tremendous potential utility in the classification of biomedical images. For example, images acquired with retinal optical coherence tomography (OCT) can be used to accurately classify patients with adult macular degeneration (AMD), and distinguish them from healthy control patients. However, previous research has suggested that large amounts of data are required in order to train deep learning algorithms, because of the large number of parameters that need to be fit. Here, we show that a moderate amount of data (data from approximately 1,800 patients) may be enough to reach close-to-maximal performance in the classification of AMD patients from OCT images. These results suggest that deep learning algorithms can be trained on moderate amounts of data, provided that images are relatively homogenous, and the effective number of parameters is sufficiently small. Furthermore, we demonstrate that in this application, cross-validation with a separate test set that is not used in any part of the training does not differ substantially from cross-validation with a validation data-set used to determine the optimal stopping point for training.

neuroscience↗

bioRxiv2017

Systematic mapping of chromatin state landscapes during mouse development

Embryogenesis requires epigenetic information that allows each cell to respond appropriately to developmental cues. Histone modifications are core components of a cells epigenome, giving rise to chromatin states that modulate genome function. Here, we systematically profile histone modifications in a diverse panel of mouse tissues at 8 developmental stages from 10.5 days post conception until birth, performing a total of 1,128 ChIP-seq assays across 72 distinct tissue-stages. We combine these histone modification profiles into a unified set of chromatin state annotations, and track their activity across developmental time and space. Through integrative analysis we identify dynamic enhancers, reveal key transcriptional regulators, and characterize the role of chromatin-based repression in developmental gene regulation. We also leverage these data to link enhancers to putative target genes, revealing connections between coding and non-coding sequence variation in disease etiology. Our study provides a compendium of resources for biomedical researchers, and achieves the most comprehensive view of embryonic chromatin states to date.

genomics↗

bioRxiv2017

Spatiotemporal DNA Methylome Dynamics of the Developing Mammalian Fetus

Genetic studies have revealed an essential role for cytosine DNA methylation in mammalian development. However, its spatiotemporal distribution in the developing embryo remains obscure. Here, we profiled the methylome landscapes of 12 mouse tissues/organs at 8 developmental stages spanning from early embryogenesis to birth. Indepth analysis of these spatiotemporal epigenome maps systematically delineated ~2 million methylation variant regions and uncovered widespread methylation dynamics at nearly one-half million tissue-specific enhancers, whose human counterparts were enriched for variants involved in genetic diseases. Strikingly, these predicted regulatory elements predominantly lose CG methylation during fetal development, whereas the trend is reversed after birth. Accumulation of non-CG methylation within gene bodies of key developmental transcription factors coincided with their transcriptional repression during later stages of fetal development. These spatiotemporal epigenomic maps provide a valuable resource for studying gene regulation during mammalian tissue/organ progression and for pinpointing regulatory elements involved in human developmental diseases.

genomics↗

bioRxiv2017

Deep-Learning Based, Automated Segmentation Of Macular Edema In Optical Coherence Tomography

Evaluation of clinical images is essential for diagnosis in many specialties and the development of computer vision algorithms to analyze biomedical images will be important. In ophthalmology, optical coherence tomography (OCT) is critical for managing retinal conditions. We developed a convolutional neural network (CNN) that detects intraretinal fluid (IRF) on OCT in a manner indistinguishable from clinicians. Using 1,289 OCT images, the CNN segmented images with a 0.911 cross-validated Dice coefficient, compared with segmentations by experts. Additionally, the agreement between experts and between experts and CNN were similar. Our results reveal that CNN can be trained to perform automated segmentations.

bioinformatics↗

bioRxiv2016

Deep learning is effective for the classification of OCT images of normal versus Age-related Macular Degeneration

ObjectiveThe advent of Electronic Medical Records (EMR) with large electronic imaging databases along with advances in deep neural networks with machine learning has provided a unique opportunity to achieve milestones in automated image analysis. Optical coherence tomography (OCT) is the most commonly obtained imaging modality in ophthalmology and represents a dense and rich dataset when combined with labels derived from the EMR. We sought to determine if deep learning could be utilized to distinguish normal OCT images from images from patients with Age-related Macular Degeneration (AMD).\n\nDesignEMR and OCT database study\n\nSubjectsNormal and AMD patients who had a macular OCT.\n\nMethodsAutomated extraction of an OCT imaging database was performed and linked to clinical endpoints from the EMR. OCT macula scans were obtained by Heidelberg Spectralis, and each OCT scan was linked to EMR clinical endpoints extracted from EPIC. The central 11 images were selected from each OCT scan of two cohorts of patients: normal and AMD. Cross-validation was performed using a random subset of patients. Receiver operator curves (ROC) were constructed at an independent image level, macular OCT level, and patient level.\n\nMain outcome measureArea under the ROC.\n\nResultsOf a recent extraction of 2.6 million OCT images linked to clinical datapoints from the EMR, 52,690 normal macular OCT images and 48,312 AMD macular OCT images were selected. A deep neural network was trained to categorize images as either normal or AMD. At the image level, we achieved an area under the ROC of 92.78% with an accuracy of 87.63%. At the macula level, we achieved an area under the ROC of 93.83% with an accuracy of 88.98%. At a patient level, we achieved an area under the ROC of 97.45% with an accuracy of 93.45%. Peak sensitivity and specificity with optimal cutoffs were 92.64% and 93.69% respectively.\n\nConclusionsDeep learning techniques achieve high accuracy and is effective as a new image classification technique. These findings have important implications in utilizing OCT in automated screening and the development of computer aided diagnosis tools in the future.

bioinformatics↗