Consensus Scoring Research Articles

Metapredict: a fast, accurate, and easy-to-use predictor of consensus disorder and structure

Intrinsically disordered proteins and protein regions make up a substantial fraction of many proteomes in which they play a wide variety of essential roles. A critical first step in understanding the role of disordered protein regions in biological function is to identify those disordered regions correctly. Computational methods for disorder prediction have emerged as a core set of tools to guide experiments, interpret results, and develop hypotheses. Given the multiple different predictors available, consensus scores have emerged as a popular approach to mitigate biases or limitations of any single method. Consensus scores integrate the outcome of multiple independent disorder predictors and provide a per-residue value that reflects the number of tools that predict a residue to be disordered. Although consensus scores help mitigate the inherent problems of using any single disorder predictor, they are computationally expensive to generate. They also necessitate the installation of multiple different software tools, which can be prohibitively difficult. To address this challenge, we developed a deep-learning-based predictor of consensus disorder scores. Our predictor, metapredict, utilizes a bidirectional recurrent neural network trained on the consensus disorder scores from 12 proteomes. By benchmarking metapredict using two orthogonal approaches, we found that metapredict is among the most accurate disorder predictors currently available. Metapredict is also remarkably fast, enabling proteome-scale disorder prediction in minutes. Importantly, metapredict is a fully open source and is distributed as a Python package, a collection of command-line tools, and a web server, maximizing the potential practical utility of the predictor. We believe metapredict offers a convenient, accessible, accurate, and high-performance predictor for single-proteins and proteomes alike.

Recommendations following a modified UK-Delphi consensus study on best practice for referral and management of severe asthma

IntroductionSevere asthma affects an estimated 3%–5% of people with asthma and is associated with frequent exacerbations, poor symptom control and significant morbidity from the disease itself, as well as high...

An automated ASPECTS method with atlas-based segmentation

Background and PurposeAs a simple and reliable systematic method to evaluate the early ischemic changes in the blood supply region of the middle cerebral artery of patients with ischemic stroke, the Alberta Stroke Program Early CT score (ASPECTS) can be used for rapid semi-quantitative evaluation of ischemic lesions, which is helpful to select potential candidates for intravenous and intra-arterial therapies, determine the thrombolytic effect and long-term prognosis. This method mainly relies on doctors’ visual observation. However, due to different levels of doctor's experience, the poor inter-reader agreement may result in errors in the final ASPECTS. The purpose of this work was to propose an automated semi-quantitative method for the diagnosis of acute ischemic stroke based on non-contrast computed tomography (NCCT), to provide a reference for doctors in the diagnosis and evaluation. MethodsNCCT data from a total of 90 patients were included for auto-ASPECTS training and testing. After preprocessing CT images, the regions of interest (ROI) for ASPECTS were labeled using atlas-based segmentation. The mean difference, mean ratio and brain density shifts (BDS) of the corresponding regions of the contralateral brain were used as the standard for quantitative analysis. The auto-ASPECTS method was developed and validated to predict early ischemic changes whose performance was evaluated by the agreement (accuracy) of predictions and consensus scores of two observers. ResultsA comparison was made among the results on mean difference, mean ratio, BDS and the combination of multiple parameters as the standard. The result of using BDS alone was relatively better than the result of using any other parameter alone or any combination of multiple parameters, and accuracy in the test set was 0.80. In the test set, accuracy with using different BDS thresholds increased by 6.67% compared with using the consistent BDS threshold. After dichotomy of auto-ASPECTS and consensus scores with the threshold of 7, the agreement of them was 83.3% and there was no significant difference between the two distributions (p = 0.344) in McNemar test. ConclusionsThe proposed auto-ASPECTS method for NCCT images can provide useful information for early diagnosis and evaluation of patients with acute ischemic stroke (AIS).

Drug repurposing for COVID-19 using computational screening: Is Fostamatinib/R406 a potential candidate?

With the gradual increase in the COVID-19 mortality rate, there is an urgent need for an effective drug/vaccine. Several drugs like Remdesivir, Azithromycin, Favirapir, Ritonavir, Darunavir, etc., are put under evaluation in more than 300 clinical trials to treat COVID-19. On the other hand, several vaccines like Pfizer-BioNTech, Moderna, Johnson & Johnson’s Janssen, Sputnik V, Covishield, Covaxin, etc., also evolved from the research study. While few of them already gets approved, others show encouraging results and are still under assessment. In parallel, there are also significant developments in new drug development. But, since the approval of new molecules takes substantial time, drug repurposing studies have also gained considerable momentum. The primary agent of the disease progression of COVID-19 is SARS-CoV2/nCoV, which is believed to have ~89% genetic resemblance with SARS-CoV, a coronavirus responsible for the massive outbreak in 2003. With this hypothesis, Human-SARS-CoV protein interactions are used to develop an in-silico Human-nCoV network by identifying potential COVID-19 human spreader proteins by applying the SIS model and fuzzy thresholding by a possible COVID-19 FDA drugs target-based validation. At first, the complete list of FDA drugs is identified for the level-1 and level-2 spreader proteins in this network, followed by applying a drug consensus scoring strategy. The same consensus strategy is involved in the second analysis but on a curated overlapping set of key genes/proteins identified from COVID-19 symptoms. Validation using subsequent docking study has also been performed on COVID-19 potential drugs with the available major COVID-19 crystal structures whose PDB IDs are: 6LU7, 6M2Q, 6W9C, 6M0J, 6M71 and 6VXX. Our computational study and docking results suggest that Fostamatinib (R406 as its active promoiety) may also be considered as one of the potential candidates for further clinical trials in pursuit to counter the spread of COVID-19.

A video-based situational judgement test of medical students’ communication competence in patient encounters: Development and first evaluation

ObjectiveWe developed and evaluated the Video-Based Assessment of Medical Communication Competence (VA-MeCo), a construct-driven situational judgement test measuring medical students’ communication competence in patient encounters. MethodsIn the construction phase, we conducted two expert studies (npanel1 = 6, npanel2 = 13) to ensure curricular and content validity and sufficient expert agreement on the answer key. In the evaluation phase, we conducted a cognitive pre-test (n = 12) and a pilot study (n = 117) with medical students to evaluate test usability and acceptance, item statistics and test reliability depending on the applied scoring method (raw consensus vs. pairwise comparison scoring). ResultsThe results of the expert interviews indicated good curricular and content validity. Expert agreement on the answer key was high (ICCs> .86). The pilot study showed favourable usability and acceptance by students. Irrespective of the scoring method, reliability for the complete test (Cronbach’s α >.93) and its subscales (α >.83) was high. ConclusionThere is promising evidence that medical communication competence can be validly and reliably measured using a construct-driven and video-based situational judgement test. Practice ImplicationsVideo-based SJTs allow efficient online assessment of medical communication competence and are well accepted by students and educators.

Using Ultrasound Color Doppler Twinkling to Identify Biopsy Markers in the Breast and Axilla

In breast radiology, ultrasound detection of biopsy markers or clips for localization purposes is often challenging, especially in the axilla. The purpose of this research was to test the hypothesis that the surface roughness of biopsy clips would elicit a twinkling signature on color Doppler, making them more readily identifiable by ultrasound. Ultrasound color Doppler imaging of 12 biopsy markers was performed and consensus scoring of the degree of twinkling (0 [no twinkling] to 4 [exuberant twinkling]) was obtained for each of the markers. The surface roughness characteristics of the markers were measured using 3-D coherence scanning interferometry. The 3 markers scoring at least 3 for twinkling in vitro were cork, Q and Vision. Of these 3 markers, only the cork marker scored a 4 ex vivo and in cadaveric tissue. Surface roughness metrics demonstrated a positive estimated correlation with the twinkling scores (rho = 0.33, 95% CI = [–0.48 to 0.84]). Of the 12 markers tested, the markers that twinkled corresponded to surface roughness measured with non-contact 3-D optical imaging. Qualitatively, lower color scales and color frequencies optimized twinkling, but the most specific qualitative predictor of confidence in twinkling was insensitivity to changes in color scale and color frequency values.

An international multidisciplinary peer-driven consensus on the optimal use of hemostatic powders in surgical practice.

Bleeding is a consequence of insufficient hemostasis and excessive bleeding at a surgical site is associated with an increased risk of post-operative infection, transfusion and re-operation, in addition to increased hospital length of stay and costs. Surgeons employ a range of methods to achieve hemostasis, including topical hemostatic agents of differing composition and properties. Hemostatic powders are a sub-group of topical hemostats, which can be used in helping as adjuncts to manage troublesome bleeding in a variety of situations. As this technology is relatively new and potentially not well known by the broad surgical community, no specific guidelines or recommendations for the optimal use of hemostatic powders in surgery currently exist. A steering group throughout Europe of multidisciplinary surgeons, expert in hemostasis and hemostatics, identified from literature and from personal experience, five key topics. When to use hemostatic powder, the evidence for use, benefits of use, safety remarks and considerations in various surgical specialties. Thirty-seven statements were subsequently drawn from these five key topics. An online survey was sent to 128 high-volume surgeons working in breast surgery, gynaecological and obstetric surgery, general and emergency surgery, thoracic surgery and urological surgery in Europe to assess agreement (consensus) with these statements. Consensus was defined as high if ≥ 75% and very high if ≥ 90% of respondents agreed with a statement. A total of 79 responses were received and consensus among the surgical experts was very high in 27 (73%) statements, high in 8 (22%) statements and was not achieved in 2 (5%) statements. Based on the consensus scores, the steering group produced 16 key recommendations which they considered could improve patient outcomes by reducing post-operative bleeding and its associated complications using hemostatic powder.Supplementary InformationThe online version contains supplementary material available at 10.1007/s13304-021-01136-x.

Transplantation programs facing lack of empirical evidence on SARS-CoV-2 vaccination: A society recommendation consensus update.

BackgroundSince phase III trials for the most prominent vaccines excluded immunocompromised or immunosuppressed patients, data on safety and efficacy of SARS‐CoV‐2 vaccines for recipients of solid organ transplantations are scarce.AimsOur study offers a synthesis of expert opinions aligned with available data addressing key questions of the clinical management of SARS‐CoV‐2 vaccinations for transplant patients.MethodAn online research was performed retrieving available recommendations by national and international transplantation organizations and state institutions on SARS‐CoV2 vaccination management for transplant recipients.ResultsEleven key statements were identified from recommendations by 18 national and international societies, and consensus for the individual statements was evaluated by means of the Society Recommendation Consensus score. The highest consensus level (SRC A) was found for prioritized access to vaccination for transplant patients despite anticipation of a weakened immune response. All currently authorized vaccines can be considered safe for transplant patients (SRC A). The handling of immunosuppressive medication, the timely management of vaccines, and other aspects were aligned with available expert opinions.ConclusionExpert consensus can be determined for crucial aspects of the implementation of SARS‐CoV‐2 vaccination programs. We hereby offer a tool for immediate decision‐making until empirical data becomes available.

Strengths and Weaknesses of Docking Simulations in the SARS-CoV-2 Era: the Main Protease (Mpro) Case Study.

The scientific community is working against the clock to arrive at therapeutic interventions to treat patients with COVID-19. Among the strategies for drug discovery, virtual screening approaches have the capacity to search potential hits within millions of chemical structures in days, with the appropriate computing infrastructure. In this article, we first analyzed the published research targeting the inhibition of the main protease (Mpro), one of the most studied targets of SARS-CoV-2, by docking-based methods. An alarming finding was the lack of an adequate validation of the docking protocols (i.e., pose prediction and virtual screening accuracy) before applying them in virtual screening campaigns. The performance of the docking protocols was tested at some level in 57.7% of the 168 investigations analyzed. However, we found only three examples of a complete retrospective analysis of the scoring functions to quantify the virtual screening accuracy of the methods. Moreover, only two publications reported some experimental evaluation of the proposed hits until preparing this manuscript. All of these findings led us to carry out a retrospective performance validation of three different docking protocols, through the analysis of their pose prediction and screening accuracy. Surprisingly, we found that even though all tested docking protocols have a good pose prediction, their screening accuracy is quite limited as they fail to correctly rank a test set of compounds. These results highlight the importance of conducting an adequate validation of the docking protocols before carrying out virtual screening campaigns, and to experimentally confirm the predictions made by the models before drawing bold conclusions. Finally, successful structure-based drug discovery investigations published during the redaction of this manuscript allow us to propose the inclusion of target flexibility and consensus scoring as alternatives to improve the accuracy of the methods.

Patch test informed consent form: position statement by European Academy of Dermatology and Venereology Task Force on Contact Dermatitis.

To our knowledge, an international consensus is lacking regarding the development of an adequate informed consent form for a patch test (PT) and the information that should be included in such document. The aim of the study was to reach a consensus on the specific points that need to be addressed in a PT consent form. A Delphi survey, comprising 2 rounds and 1 final discussion, was used to gather and analyse data, which was conducted over the Internet. Each statement that reached a consensus with the respondents (9 expert dermatologists from Europe) was defined as a median consensus score (MED) of ≥7 and agreement among panelists as an interquartile range (IQR) of ≤3. All study participants were members of the EADV task force on contact dermatitis. The expert panel addressed several topics that should be included in an informed consent form for a PT: introduction, preparation for PT, testing procedure, allowed activities, adverse events and additional authorizations. Our results assess recommendations regarding points to be contained in an informed consent form for a PR. Future actions towards standardization and harmonization of this specific consent form are needed.

Comparison of 4 commercial enzyme immunoassays for serology testing of human parvovirus B19 infection

BackgroundParvovirus B19 is a pathogenic virus often diagnosed by serology, yet little is known about analytical performance of commercial enzyme immunoassays (EIAs). ObjectiveTo investigate performance of 4 EIAs for parvovirus B19 IgM and IgG: Liaison, Euroimmun, Mikrogen and Virion/Serion. Study designTo compare 4 EIAs to Biotrin's ELISA on 168 samples and determine consensus score for discordant samples using Mikrogen's confirmatory line assay. ResultsTwo thirds of results for IgM/IgG were identical for all 4 EIAs and Biotrin. Liaison shows the highest IgM sensitivity, but has low specificity. Euroimmun lacks IgM sensitivity. Mikrogen had a good overall performance, but had the lowest IgG specificity. Virion/Serion had variable performance with a low IgM specificity and the most borderline and cross-reactive results. ConclusionsLiaison and Mikrogen have similar performance to Biotrin's ELISA. Euroimmun lacks sensitivity and Virion/Serion produced many borderline and cross-reactive results.

Predictors of missed infection control care: A tri-partite international study.

AimThis study aimed to quantify types and frequencies of missed infection control care and to develop a theoretical model for estimating nurses’ consensus scores about this form of missed care.DesignA non‐experimental research design using self‐audit data was selected to collect information about the types and frequencies of missed infection control care from nurses employed in hospitals located in three different countries. Data collection commenced mid‐year 2018.MethodsA multivariate approach was used to apply the consensus scores of 1.911 internationally based nurses in the missed opportunities for maintaining infection control.Results/findingsThirteen variables exert direct effects on the nurses’ total scores underpinning missed infection control care. These include the methods used to prevent hospital‐acquired infections, surveillance and hand hygiene practices. Significant nurses’ demographic factors also included their countries of origin, employment status, employer type, job retention intentions, work intensity, length of clinical experience and staff development attendance.ConclusionIn magnitude of importance and having the largest effect on missed infection control care is missed care related to reducing hospital‐acquired infections followed closely by surveillance. Missed infection control care can be quantified, and variances in its practices can be accounted by exploring the nurses’ differing demographic factors, including the nurses’ country of origin.ImpactVariations in missed infection control care can be accounted for across three countries. While ward hygiene is underestimated by staff as a mechanism to minimize nosocomial infections, infection control surveillance remains the key to reducing hospital‐acquired infections. The study's outcomes invite the use of an ongoing, whole‐of‐organization approach to infection control with scrutiny being needed for improved staff adherence particularly with hand hygiene.

Limited added value of whole spine MRI in spondyloarthritis for disease activity assessment.

To evaluate the added value of whole spine magnetic resonance imaging (MRI) for disease activity assessment in ankylosing spondylitis (AS) and psoriatic arthritis (PsA). Spine and sacroiliac joint (SIJ) MRI scans requested by rheumatologists between 2012 and 2018 were screened retrospectively, and patients who had known diagnosis of AS or PsA were included, if the MRI was done for disease activity assessment. All MRI scans were reviewed by two experienced musculoskeletal radiologists independently, blinded to patients' diagnosis and to the other MRI. Comparisons were done for the presence of active and structural lesions. In addition, radiologists were asked to rate for "confidence level for active inflammation related to SpA." Analysis was done using the consensus scores. Ninety patients with known diagnosis of AS (n = 55) or PsA (n = 35) were included. The frequency of active and structural lesions was not significantly different both in AS vs PsA, neither in the cervical/thoracic/lumbar spine or the SIJ. The percentage of people only with any inflammatory changes on the spine MRI without any inflammation in the SIJ MRI was 24% in AS and 23% in PsA. However, considering the confidence level of the radiologists on active inflammation, only one patient's spine MRI was scored as active, while SIJ MRI being negative for inflammation. The spinal MRI had limited added value to the SIJ MRI in SpA, when performed to assess disease activity, limiting its value in routine practice unless clinically indicated. Key Points • Spine MRI adds limited value to SIJs in SpA, when performed for disease activity assessment. • SpA disease activity assessment may be restricted to sacroiliac joint MRI, unless clinically indicated.

Effect of the acyl-group length on the chemoselectivity of the lipase-catalyzed acylation of propranolol-a computational study.

The selective N-acylation of 1,2-amino alcohols has been proposed to occur through the proton shuttle mechanism. However, the O-acetylation of propranolol catalyzed by Candida antarctica lipase B is an exception. We investigated the relation between the chemoselectivity of this reaction and the acyl group length. For this purpose, we compared the acyl groups: ethanoyl, butanoyl, octanoyl, and hexadecanoyl. We studied the Michaelis complexes between serine-acylated Candida antarctica lipase B and propranolol, employing a computational approach that involved sampling Michaelis complex conformations through ensemble docking plus consensus scoring and molecular dynamics simulations. The conformations were then classified as near attack conformations for acylation of the amino or hydroxy group. The relative populations of these two classes of conformations were found to be consistent with the experimentally observed chemoselective O-acetylation. We predict that increasing the length of the hydrocarbon chain of the acyl group will cause O-acylation to be unfavorable with respect to N-acylation. The nucleophilic attack of propranolol to the acylated lipase was found to be more favorable through the classical mechanism when compared with the proton shuttle mechanism.

Establishing expert consensus for the optimal approach to holistic risk-management in pulmonary arterial hypertension: a Delphi process and narrative review

ABSTRACT Background: Pulmonary arterial hypertension (PAH) is associated with significant morbidity and reduced life expectancy. Various medical therapies, together with non-medical therapies such as exercise training, have been shown to improve outcomes for patients. We performed a Delphi consensus process to establish optimal approaches to optimizing patient care. Methods: A steering group of PAH experts formulated 38 statements grouped into 6 themes: burden of PAH, risk-stratification, the role of clinical phenotyping in the management of PAH, assessing clinical response to treatment, maximizing the medical treatment pathway and the role of other management options. An online survey was sent to PAH health-care professionals throughout the UK to assess consensus with these statements. Consensus was defined as high if ≥70% and very high if ≥90% of the respondents agreed with a statement. A narrative review for each theme was then performed Results: Consensus was very high in 27 (71%) statements, high in 7 (18%) statements and was not achieved in 4 (11%) statements. Conclusions: Based on the consensus scores, the steering group derived 13 recommendations which, if implemented, should result in improved holistic care of patients with PAH.

Oligomeric Formulas in Surgery: A Delphi and Consensus Study.

Nutritional management of patients with intestinal failure often includes the use of oligomeric formulas. Implementing the use of oligomeric formulas in surgical patients with maldigestion or malabsorption could be a nutritional strategy to be included in clinical protocols. We aim to generate knowledge from a survey focused on the effectiveness of nutritional therapy with oligomeric formulas with Delphi methodology. Each statement that reached an agreement consensus among participants was defined as a median consensus score ≥7 and as an interquartile range ≤3. The use of oligomeric formulas in surgical patients, starting enteral nutrition in the post-operative phase in short bowel syndrome and in nonspecific diarrhea after surgical procedures, could improve nutritional therapy implementation. Stakeholders agreed that early jejunal enteral nutrition with oligomeric formula is more effective compared to intravenous fluid therapy and it is useful in patients undergoing upper gastro-intestinal tract major surgery when malabsorption or maldigestion is suspected. Finally, oligomeric formulas may be useful when a feeding tube is placed distally to the duodenum. This study shows a practical approach to the use of oligomeric formulas in surgical patients with intestinal disorders and malabsorption, and it helps clinicians in the decision-making process.

Diagnostic contribution of contrast-enhanced 3D MR imaging of peripheral nerve pathology.

To assess the diagnostic contribution of contrast-enhanced 3D STIR (ce3D-SS) high-resolution magnetic resonance (MR) imaging of peripheral nerve pathology relative to conventional 2D sequences. In this IRB-approved retrospective study, two radiologists reviewed 60 MR neurography studies with nerve pathology findings. The diagnostic contribution of ce3D-SS imaging was scored on a 4-point Likert scale (1 = no additional information, 2 = supports interpretation, 3 = moderate additional information, and 4 = diagnosis not possible without ce3D-SS). Image quality, nerve visualization, and detection of nerve pathology were also assessed for both standard 2D neurography and ce3D-SS sequences utilizing a 3-point Likert scale. Descriptive statistics are reported. The diagnostic contribution score for ce3D-SS imaging was 2.25 for the brachial plexus, 1.50 for extremities, and 1.75 for the lumbosacral plexus. For brachial plexus, the mean consensus scores for image quality, nerve visualization, and detection of nerve pathology were 2.55, 2.5, and 2.55 for 2D and 2.35, 2.45, and 2.45 for 3D. For extremities, the mean consensus scores for image quality, nerve visualization, and detection of nerve pathology were 2.60, 2.80, and 2.70 for 2D and 1.8, 2.20, and 2.10 for 3D. For lumbosacral plexus, the mean consensus scores for image quality, nerve visualization, and detection of nerve pathology were 2.45, 2.75, and 2.65 for 2D and 2.0, 2.45, and 2.25 for 3D. Overall, our study supports the potential application of ce3D-SS imaging for MRN of the brachial plexus but suggests that 2D MRN protocols are sufficient for MRN of the extremities and lumbosacral plexus.

MTR3D: identifying regions within protein tertiary structures under purifying selection.

The identification of disease-causal variants is non-trivial. By mapping population variation from over 448,000 exome and genome sequences to over 81,000 experimental structures and homology models of the human proteome, we have calculated both regional intolerance to missense variation (Missense Tolerance Ratio, MTR), using a sliding window of 21–41 codons, and introduce a new 3D spatial intolerance to missense variation score (3D Missense Tolerance Ratio, MTR3D), using spheres of 5–8 Å. We show that the MTR3D is less biased by regions with limited data and more accurately identifies regions under purifying selection than estimates relying on the sequence alone. Intolerant regions were highly enriched for both ClinVar pathogenic and COSMIC somatic missense variants (Mann–Whitney U test P < 2.2 × 10−16). Further, we combine sequence- and spatial-based scores to generate a consensus score, MTRX, which distinguishes pathogenic from benign variants more accurately than either score separately (AUC = 0.85). The MTR3D server enables easy visualisation of population variation, MTR, MTR3D and MTRX scores across the entire gene and protein structure for >17,000 human genes and >42,000 alternative alternate transcripts, including both Ensembl and RefSeq transcripts. MTR3D is freely available by user-friendly web-interface and API at http://biosig.unimelb.edu.au/mtr3d/.

Machine learning models to quantify HER2 for real-time tissue image analysis in prospective clinical trials.

3061 Background: Patient eligibility for HER2-targeting treatments is commonly informed by testing tumor HER2 expression using immunohistochemistry. As HER2 expression is visually assessed by pathologists, inter- and intra-rater variability might affect treatment decisions. Here, we report the development of an automated machine learning (ML)-based algorithm to quantify HER2 cell membrane expression across a diversity of breast cancer phenotypes as a clinical tool for monitoring HER2 testing quality. Methods: A total of 689 breast cancer tissue samples were either procured (Avaden Biosciences) or were anonymized samples from the AstraZeneca biobank comprising tissues from primary and metastatic tumors, core needle biopsies and surgical resections, lobular and ductal carcinomas, across tumor grades and HER2 expression levels. Samples were stained for HER2 detection (Ventana HER2 (4B5) Assay) and digitized (Leica Biosystems) across 5 laboratories in the US. Whole-slide images (WSIs) were stratified into training (n = 407), validation (n = 110), and test sets (n = 172). Multiple convolutional neural network based ML models (PathAI, Boston, MA) were trained using 190,000 manual annotations provided by 30 board-certified pathologists to identify artifacts, invasive tumor, identify individual cancer cells and measure tumor cell membrane HER2 expression as partial or complete, and negative, weak-or-moderate, or intense. Cell-level scores were validated against a consensus of manual cell counts from 5 independent pathologists in 320 representative regions of test set WSIs. HER2 scores were generated by automatically applying rules derived from 2018 ASCO/CAP guidelines and then compared in the test set with consensus scores from 3 independent pathologists. Results: Cell counts provided by the ML model were strongly consistent with cell counts obtained by pathologist consensus in all cell-types except for faintly positive HER2 cells where ML-based quantification identified more cells on average. Automatically generated ML-ASCO/CAP HER2 scores using WSI showed substantial consistency across IHC categories with the consensus of pathologists (ICC 0.88, 95%CI 0.82-0.92) in the test set and improved further when ML models were trained to agree with pathologists by adjusting cut offs (ICC 0.91, 95%CI 0.89-0.94). The ML-based model was deployed through the PathAI cloud platform to calculate HER2 testing quality control metrics in real-time in multicentric clinical trials. Conclusions: Automated image analysis of HER2-stained breast cancer tissues using ML-based models is consistent with pathologist consensus across breast cancer tissue types. The results support evidence that ML-based algorithms can help pathologists assess HER2 testing reproducibility in clinical trials.

Neighborhood Consensus Networks for Unsupervised Multi-view Outlier Detection

Multi-view outlier detection recently attracted rapidly growing attention with the development of multi-view learning. Although promising performance demonstrated, we observe that identifying outliers in multi-view data is still a challenging task due to the complicated characteristics of multi-view data. Specifically, an effective multi-view outlier detection method should be able to handle (1) different types of outliers; (2) two or more views; (3) samples without clusters; (4) high dimensional data. Unfortunately, little is known about how these four issues can be handled simultaneously. In this paper, we propose an unsupervised multi-view outlier detection method to address these issues. Our method is based on the proposed novel neighborhood consensus networks termed NC-Nets, which automatically encodes intrinsic information into a comprehensive latent space for each view (for issue (4)) and uniforms the neighborhood structures among different views (for issue (2)). Accordingly, we propose an outlier score measurement which consists of two parts: the within-view reconstruction score and the cross-view neighborhood consensus score. The measurement is designed based on the characteristics of the different outlier types (for issue (1)) and no cluster assumption is needed (for issue (3)). Experimental results show that our method significantly outperforms state-of-the-art methods. On average, our method achieves 11.2% ~ 96.2% improvement in term of AUC and 33.5% ~ 352.7% improvement in term of F1-Score.