Analysis of the Human Protein Atlas Image Classification competition

AbstractPinpointing subcellular protein localizations from microscopy images is easy to the trained eye, but challenging to automate. Based on the Human Protein Atlas image collection, we held a competition to identify deep learning solutions to solve this task. Challenges included training on highly imbalanced classes and predicting multiple labels per image. Over 3 months, 2,172 teams participated. Despite convergence on popular networks and training techniques, there was considerable variety among the solutions. Participants applied strategies for modifying neural networks and loss functions, augmenting data and using pretrained networks. The winning models far outperformed our previous effort at multi-label classification of protein localization patterns by ~20%. These models can be used as classifiers to annotate new images, feature extractors to measure pattern similarity or pretrained networks for a wide range of biological applications.

Download Full-text

Automated classification of immunostaining patterns in breast tissue from the human protein Atlas

Journal of Pathology Informatics ◽

10.4103/2153-3539.109881 ◽

2013 ◽

Vol 4 (2) ◽

pp. 14 ◽

Cited By ~ 14

Author(s):

IssacNiwas Swamidoss ◽

Palanisamy Ponnusamy ◽

Martin Simonsson ◽

Robin Strand ◽

Virginie Uhlmann ◽

...

Keyword(s):

Breast Tissue ◽

Human Protein ◽

Automated Classification ◽

Human Protein Atlas

Download Full-text

Modes of Interaction in Naturally Occurring Medical Encounters With General Practitioners: The “One in a Million” Study

Qualitative Health Research ◽

10.1177/1049732321993790 ◽

2021 ◽

pp. 104973232199379

Author(s):

Olaug S. Lian ◽

Sarah Nettleton ◽

Åge Wifstad ◽

Christopher Dowrick

Keyword(s):

General Practitioners ◽

Narrative Analysis ◽

Mode Competition ◽

General Applicability ◽

Naturally Occurring ◽

Wide Range ◽

The One ◽

Narrative Mode ◽

Modes Of Interaction

In this article, we qualitatively explore the manner and style in which medical encounters between patients and general practitioners (GPs) are mutually conducted, as exhibited in situ in 10 consultations sourced from the One in a Million: Primary Care Consultations Archive in England. Our main objectives are to identify interactional modes, to develop a classification of these modes, and to uncover how modes emerge and shift both within and between consultations. Deploying an interactional perspective and a thematic and narrative analysis of consultation transcripts, we identified five distinctive interactional modes: question and answer (Q&A) mode, lecture mode, probabilistic mode, competition mode, and narrative mode. Most modes are GP-led. Mode shifts within consultations generally map on to the chronology of the medical encounter. Patient-led narrative modes are initiated by patients themselves, which demonstrates agency. Our classification of modes derives from complete naturally occurring consultations, covering a wide range of symptoms, and may have general applicability.

Download Full-text

Uncertainty-Aware Deep Learning-Based Cardiac Arrhythmias Classification Model of Electrocardiogram Signals

Computers ◽

10.3390/computers10060082 ◽

2021 ◽

Vol 10 (6) ◽

pp. 82

Author(s):

Ahmad O. Aseeri

Keyword(s):

Deep Learning ◽

Cardiac Arrhythmias ◽

Large Scale ◽

Clinical Decision Making ◽

Probabilistic Approach ◽

Classification Model ◽

Gating Mechanism ◽

Uncertainty Estimates ◽

Wide Range

Deep Learning-based methods have emerged to be one of the most effective and practical solutions in a wide range of medical problems, including the diagnosis of cardiac arrhythmias. A critical step to a precocious diagnosis in many heart dysfunctions diseases starts with the accurate detection and classification of cardiac arrhythmias, which can be achieved via electrocardiograms (ECGs). Motivated by the desire to enhance conventional clinical methods in diagnosing cardiac arrhythmias, we introduce an uncertainty-aware deep learning-based predictive model design for accurate large-scale classification of cardiac arrhythmias successfully trained and evaluated using three benchmark medical datasets. In addition, considering that the quantification of uncertainty estimates is vital for clinical decision-making, our method incorporates a probabilistic approach to capture the model’s uncertainty using a Bayesian-based approximation method without introducing additional parameters or significant changes to the network’s architecture. Although many arrhythmias classification solutions with various ECG feature engineering techniques have been reported in the literature, the introduced AI-based probabilistic-enabled method in this paper outperforms the results of existing methods in outstanding multiclass classification results that manifest F1 scores of 98.62% and 96.73% with (MIT-BIH) dataset of 20 annotations, and 99.23% and 96.94% with (INCART) dataset of eight annotations, and 97.25% and 96.73% with (BIDMC) dataset of six annotations, for the deep ensemble and probabilistic mode, respectively. We demonstrate our method’s high-performing and statistical reliability results in numerical experiments on the language modeling using the gating mechanism of Recurrent Neural Networks.

Download Full-text

Classification of unlabeled online media

Scientific Reports ◽

10.1038/s41598-021-85608-5 ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Sakthi Kumar Arul Prakash ◽

Conrad Tucker

Keyword(s):

Social Media ◽

Real World ◽

Graphical Model ◽

Ground Truth ◽

Classification Problem ◽

Machine Learning Algorithms ◽

Social Media Networks ◽

Online Social Media ◽

Wide Range

AbstractThis work investigates the ability to classify misinformation in online social media networks in a manner that avoids the need for ground truth labels. Rather than approach the classification problem as a task for humans or machine learning algorithms, this work leverages user–user and user–media (i.e.,media likes) interactions to infer the type of information (fake vs. authentic) being spread, without needing to know the actual details of the information itself. To study the inception and evolution of user–user and user–media interactions over time, we create an experimental platform that mimics the functionality of real-world social media networks. We develop a graphical model that considers the evolution of this network topology to model the uncertainty (entropy) propagation when fake and authentic media disseminates across the network. The creation of a real-world social media network enables a wide range of hypotheses to be tested pertaining to users, their interactions with other users, and with media content. The discovery that the entropy of user–user and user–media interactions approximate fake and authentic media likes, enables us to classify fake media in an unsupervised learning manner.

Download Full-text

Sequence data from isolated lichen-associated melanized fungi enhance delimitation of two new lineages within Chaetothyriomycetidae

Mycological Progress ◽

10.1007/s11557-021-01706-8 ◽

2021 ◽

Vol 20 (7) ◽

pp. 911-927

Author(s):

Lucia Muggia ◽

Yu Quan ◽

Cécile Gueidan ◽

Abdullah M. S. Al-Hatmi ◽

Martin Grube ◽

...

Keyword(s):

Sequence Data ◽

Single Species ◽

Sister Group ◽

Asexual Propagation ◽

Dna Sequence Data ◽

Wide Range ◽

The Family ◽

Rock Inhabiting Fungi ◽

Stable Habitat

AbstractLichen thalli provide a long-lived and stable habitat for colonization by a wide range of microorganisms. Increased interest in these lichen-associated microbial communities has revealed an impressive diversity of fungi, including several novel lineages which still await formal taxonomic recognition. Among these, members of the Eurotiomycetes and Dothideomycetes usually occur asymptomatically in the lichen thalli, even if they share ancestry with fungi that may be parasitic on their host. Mycelia of the isolates are characterized by melanized cell walls and the fungi display exclusively asexual propagation. Their taxonomic placement requires, therefore, the use of DNA sequence data. Here, we consider recently published sequence data from lichen-associated fungi and characterize and formally describe two new, individually monophyletic lineages at family, genus, and species levels. The Pleostigmataceae fam. nov. and Melanina gen. nov. both comprise rock-inhabiting fungi that associate with epilithic, crust-forming lichens in subalpine habitats. The phylogenetic placement and the monophyly of Pleostigmataceae lack statistical support, but the family was resolved as sister to the order Verrucariales. This family comprises the species Pleostigma alpinum sp. nov., P. frigidum sp. nov., P. jungermannicola, and P. lichenophilum sp. nov. The placement of the genus Melanina is supported as a lineage within the Chaetothyriales. To date, this genus comprises the single species M. gunde-cimermaniae sp. nov. and forms a sister group to a large lineage including Herpotrichiellaceae, Chaetothyriaceae, Cyphellophoraceae, and Trichomeriaceae. The new phylogenetic analysis of the subclass Chaetothyiomycetidae provides new insight into genus and family level delimitation and classification of this ecologically diverse group of fungi.

Download Full-text

Convolutional Neural Network-Based Artificial Intelligence for Classification of Protein Localization Patterns

Biomolecules ◽

10.3390/biom11020264 ◽

2021 ◽

Vol 11 (2) ◽

pp. 264

Author(s):

Kaisa Liimatainen ◽

Riku Huttunen ◽

Leena Latonen ◽

Pekka Ruusuvuori

Keyword(s):

Neural Network ◽

Artificial Intelligence ◽

Convolutional Neural Network ◽

High Throughput ◽

Protein Function ◽

Protein Localization ◽

Convolutional Network ◽

High Throughput Analysis ◽

Fully Convolutional Network

Identifying localization of proteins and their specific subpopulations associated with certain cellular compartments is crucial for understanding protein function and interactions with other macromolecules. Fluorescence microscopy is a powerful method to assess protein localizations, with increasing demand of automated high throughput analysis methods to supplement the technical advancements in high throughput imaging. Here, we study the applicability of deep neural network-based artificial intelligence in classification of protein localization in 13 cellular subcompartments. We use deep learning-based on convolutional neural network and fully convolutional network with similar architectures for the classification task, aiming at achieving accurate classification, but importantly, also comparison of the networks. Our results show that both types of convolutional neural networks perform well in protein localization classification tasks for major cellular organelles. Yet, in this study, the fully convolutional network outperforms the convolutional neural network in classification of images with multiple simultaneous protein localizations. We find that the fully convolutional network, using output visualizing the identified localizations, is a very useful tool for systematic protein localization assessment.

Download Full-text

Investigating the Applicability of Antibodies Generated within the Human Protein Atlas as Capture Agents in Immunoenrichment Coupled to Mass Spectrometry

Journal of Proteome Research ◽

10.1021/pr500691a ◽

2014 ◽

Vol 13 (10) ◽

pp. 4424-4435 ◽

Cited By ~ 5

Author(s):

Tove Boström ◽

Henrik J. Johansson ◽

Janne Lehtiö ◽

Mathias Uhlén ◽

Sophia Hober

Keyword(s):

Mass Spectrometry ◽

Human Protein ◽

Human Protein Atlas

Download Full-text

Antibodies for profiling the human proteome-The Human Protein Atlas as a resource for cancer research

PROTEOMICS ◽

10.1002/pmic.201100504 ◽

2012 ◽

Vol 12 (13) ◽

pp. 2067-2077 ◽

Cited By ~ 49

Author(s):

Anna Asplund ◽

Per-Henrik D. Edqvist ◽

Jochen M. Schwenk ◽

Fredrik Pontén

Keyword(s):

Cancer Research ◽

Human Proteome ◽

Human Protein ◽

Human Protein Atlas

Download Full-text

Introduction of a new AAC symbol classification system: the multidimensional quaternary symbol continuum (MQSC)

Journal of Enabling Technologies ◽

10.1108/jet-04-2021-0024 ◽

2021 ◽

Vol ahead-of-print (ahead-of-print) ◽

Author(s):

Eliada Pampoulou ◽

Donald R. Fuller

Keyword(s):

Peer Review ◽

Classification System ◽

Augmentative And Alternative Communication ◽

Social Issues ◽

Content Type ◽

Design Attributes ◽

Wide Range ◽

Current Classification ◽

Symbol Classification

PurposeWhen the augmentative and alternative communication (ACC) model (Lloyd et al., 1990) was proposed, these components of symbols were not considered, nor were they contemplated when superordinate (Lloyd and Fuller, 1986) and subordinate levels (Fuller et al., 1992) of AAC symbol taxonomy were developed. The purpose of this paper is to revisit the ACC model and propose a new symbol classification system called multidimensional quaternary symbol continuum (MQSC)Design/methodology/approachThe field of AAC is evolving at a rapid rate in terms of its clinical, social, research and theoretical underpinnings. Advances in assessment and intervention methods, technology and social issues are all responsible to some degree for the significant changes that have occurred in the field of AAC over the last 30 years. For example, the number of aided symbol collections has increased almost exponentially over the past couple of decades. The proliferation of such a large variety of symbol collections represents a wide range of design attributes, physical attributes and linguistic characteristics for aided symbols and design attributes and linguistic characteristics for unaided symbols.FindingsTherefore, it may be time to revisit the AAC model and more specifically, one of its transmission processes referred to as the means to represent.Originality/valueThe focus of this theoretical paper then, is on the current classification of symbols, issues with respect to the current classification of symbols in terms of ambiguity of terminology and the evolution of symbols, and a proposal for a new means of classifying the means to represent.Peer reviewThe peer review history for this article is available at: https://publons.com/publon10.1108/JET-04-2021-0024

Download Full-text

Abstract TP132: Selection and Activation of Sites in a Large Multi-Center Randomized Clinical Trial

Stroke ◽

10.1161/str.48.suppl_1.tp132 ◽

2017 ◽

Vol 48 (suppl_1) ◽

Author(s):

Bart M Demaerschalk ◽

Robert D Brown ◽

Virginia J Howard ◽

MeeLee Tom ◽

Mary E Longbottom ◽

...

Keyword(s):

Site Selection ◽

Performance Metrics ◽

Interventional Neuroradiology ◽

Interquartile Range ◽

Carotid Revascularization ◽

Training Requirements ◽

Medical Centers ◽

Wide Range ◽

The Times ◽

And Training

Introduction: Careful selection and timely activation of clinical sites in multicenter clinical trials is critical for successful enrollment, subject safety, and generalizability of results. Methods: In the Carotid Revascularization and Medical Management for Asymptomatic Carotid Stenosis Trial (CREST-2), a multidisciplinary Site Selection Committee evaluated applicants referred via participation in CREST, CREST principal investigators (PIs) and other investigators, StrokeNet and industry partners. Data for consideration included performance metrics in CREST and other carotid trials and a site selection questionnaire containing information on the investigators as well as quantitative data on carotid procedures performed. Any FDA warning letters were reviewed. Results: The Committee met bi-weekly for 36 months (n=64 meetings). Applications from 176 sites between March 2014 and July 2016 were evaluated: 153 were approved, 7 are under Committee review, 5 were approved but withdrew, 5 were placed on a waiting list, and 6 were rejected. One-hundred-four sites have completed the regulatory and training requirements to randomize: 51 (49%) academic medical centers, 31 (30%) private hospital-based centers, 16 (15%) private office-based practices, and 6 (6%) Veterans Administration medical centers. The mean times from application-to- approval was 5.2 weeks (interquartile range, 1.9, 6.2), and from approval-to-randomization status was 46.7 weeks (interquartile range, 35.4, 51.7). Specialties of the 104 site PIs are vascular surgery for 35 (33.7%), cardiology for 30 (28.8%), neurology for 25 (24%), neurosurgery for 8 (7.7%), interventional radiology for 4 (3.8%), and interventional neuroradiology for 2 (1.9%). Conclusions: Careful site selection is time-consuming for prospective sites and for trial leadership. Times from application-to-site-approval were modest (mean = 5.2 weeks), in contrast to the times for completing regulatory and training requirements (mean = 46.7 weeks). However, subject enrollment by teams from a wide range of medical centers led by a multi-disciplinary cohort of PIs will promote the generalizability of trial results.

Download Full-text