CoNIC Challenge: Pushing the frontiers of nuclear detection, segmentation, classification and counting.

Nuclear detection, segmentation and morphometric profiling are essential in helping us further understand the relationship between histology and patient outcome. To drive innovation in this area, we setup a community-wide challenge using the largest available dataset of its kind to assess nuclear segmentation and cellular composition. Our challenge, named CoNIC, stimulated the development of reproducible algorithms for cellular recognition with real-time result inspection on public leaderboards. We conducted an extensive post-challenge analysis based on the top-performing models using 1,658 whole-slide images of colon tissue. With around 700 million detected nuclei per model, associated features were used for dysplasia grading and survival analysis, where we demonstrated that the challenge's improvement over the previous state-of-the-art led to significant boosts in downstream performance. Our findings also suggest that eosinophils and neutrophils play an important role in the tumour microevironment. We release challenge models and WSI-level results to foster the development of further methods for biomarker discovery.

Screening of immune-related biological markers for aneurysmal subarachnoid hemorrhage based on machine learning approaches.

Background: Aneurysmal subarachnoid hemorrhage (aSAH) is a common hemorrhagic condition frequently encountered in the emergency department, which is characterized by high mortality and disability rates. However, the precise molecular mechanisms underlying the rupture of an aneurysm are still not fully understood. The primary objective of this study is to elucidate the fundamental molecular mechanisms underlying aSAH and provide novel therapeutic targets for the treatment of aSAH. Methods: The gene expression matrix of aSAH was downloaded from the Gene Expression Omnibus (GEO) database. In this study, we employed weighted gene co-expression network analysis (WGCNA) and differential gene expression analysis (DEGs) screening to identify crucial modules and genes associated with aSAH. Furthermore, the evaluation of immune cell infiltration was conducted through the utilization of the single-sample gene set enrichment analysis (ssGSEA) technique and the CIBERSORT algorithm. The study utilized Gene Set Variation Analysis (GSVA), Gene Ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) to investigate and comprehend the fundamental biological pathways and mechanisms. Results: Using WGCNA, six gene co-expression modules were constructed. Among the identified modules, the yellow module, which encompasses 184 genes, demonstrated the most significant correlation with aSAH. Consequently, it was determined to be the central module responsible for governing the pathogenesis of aSAH. Additionally, the application of WGCNA, LASSO regression, and multiple factor logistic regression analysis revealed ARHGAP26 and SLMAP as the key genes associated with aSAH. Furthermore, the diagnostic efficacy of these pivotal genes in aSAH was confirmed through the use of receiver operating characteristic (ROC) curve analysis, validating their discriminative potential. Moreover, the utilization of GO and KEGG pathway analysis revealed a significant enrichment of inflammation-related signaling in aSAH. Conclusion: The genes ARHGAP26 and SLMAP were identified as significant predictors of aSAH. Accordingly, these genes demonstrate significant potential to function as novel biological markers and therapeutic targets for aSAH.

Diagnostic value of plasma heparin-binding protein and the heparin-binding protein-to-albumin ratio in patients with community-acquired Pneumonia: a retrospective study.

BACKGROUND: Patients presenting to the emergency department with community-acquired pneumonia (CAP) are characterized by advanced age, comorbidities, critical illness and less-than-typical symptoms, posing a diagnostic challenge. Plasma heparin-binding protein (HBP) and the heparin-binding protein-to-albumin ratio (HBP/Alb) have not been adequately studied in the early diagnosis of CAP. This study assessed the diagnostic value of plasma HBP, HBP/Alb, and conventional inflammatory markers in emergency department patients with CAP. METHODS: We enrolled 103 patients with CAP, retrospectively analyzed the patients' clinical data, and divided the CAP patients into antibiotic (n = 79) and non-antibiotic (n = 24) groups based on whether antibiotics were administered prior to blood sampling and laboratory tests. The control group was comprised of 52 non-infected patients admitted during the same period. Within 24 h of admission, plasma HBP, serum procalcitonin (PCT), white blood cell count (WBC), neutrophil-to-lymphocyte ratio (NLR) and HBP/Alb levels were collected separately and compared. The receiver operating characteristic (ROC) curve was plotted to assess the diagnostic value of each indicator for CAP patients. Utilizing the Kappa test, the consistency of each indicator used to evaluate CAP and clinical diagnosis was analyzed. Spearman correlation was used to analyze the correlation between plasma HBP and clinical indicators of CAP patients. RESULTS: Plasma HBP, serum PCT, WBC, NLR and HBP/Alb were all elevated in the CAP group in comparison to the control group (P < 0.001). Plasma HBP, serum PCT, WBC, NLR and HBP/Alb levels did not differ statistically between antibiotic and non-antibiotic groups (P > 0.05). Plasma HBP and HBP/Alb had the highest diagnostic accuracy for CAP, the area under the ROC curve (AUC) were 0.931 and 0.938 (P < 0.0001), and the best cut-off values were 35.40 ng/mL and 0.87, respectively. In evaluating the consistency between CAP and clinical diagnosis, the Kappa values for HBP, PCT, WBC, NLR and HBP/Alb were 0.749, 0.465, 0.439, 0.566 and 0.773, respectively. Spearman correlation analysis showed that plasma HBP was positively correlated with serum PCT, WBC, NLR and HBP/Alb in CAP patients (P < 0.001). CONCLUSIONS: Plasma HBP and HBP/Alb have a high clinical diagnostic value for CAP and can be used as good and reliable novel inflammatory markers in the emergency department for the early diagnosis of CAP patients.

Renal Fibrosis Is Alleviated through Targeted Inhibition of IL-11-Induced Renal Tubular Epithelial-to-Mesenchymal Transition.

Renal fibrosis is a pathologic process that leads to irreversible renal failure without effective treatment. Epithelial-to-mesenchymal transition (EMT) plays a key role in this process. The current study found that aberrant expression of IL-11 is critically involved in tubular EMT. IL-11 and its receptor subunit alpha-1 (IL-11Rα1) were significantly induced in renal tubular epithelial cells (RTECs) in unilateral ureteral obstruction (UUO) kidneys, co-localized with transforming growth factor-ß1. IL-11 knockdown ameliorated UUO-induced renal fibrosis in vivo and transforming growth factor-ß1-induced EMT in vitro. IL-11 intervention directly induced the transdifferentiation of RTECs to the mesenchymal phenotype and increased the synthesis of profibrotic mediators. The EMT response induced by IL-11 was dependent on the sequential activation of STAT3 and extracellular signal-regulated kinase 1/2 signaling pathways and the up-regulation of metadherin in RTECs. Micheliolide (MCL) competitively inhibited the binding of IL-11 with IL-11Rα1, suppressing the activation of STAT3 and extracellular signal-regulated kinase 1/2-metadherin pathways, ultimately inhibiting renal tubular EMT and interstitial fibrosis induced by IL-11. In addition, treatment with dimethylaminomicheliolide, a pro-drug of MCL for in vivo use, significantly ameliorated renal fibrosis exacerbated by IL-11 in the UUO model. These findings suggest that IL-11 is a promising target in renal fibrosis and that MCL/dimethylaminomicheliolide exerts its antifibrotic effect by suppressing IL-11/IL-11Rα1 interaction and blocking its downstream effects.

HEROHE Challenge: Predicting HER2 Status in Breast Cancer from Hematoxylin-Eosin Whole-Slide Imaging.

Breast cancer is the most common malignancy in women worldwide, and is responsible for more than half a million deaths each year. The appropriate therapy depends on the evaluation of the expression of various biomarkers, such as the human epidermal growth factor receptor 2 (HER2) transmembrane protein, through specialized techniques, such as immunohistochemistry or in situ hybridization. In this work, we present the HER2 on hematoxylin and eosin (HEROHE) challenge, a parallel event of the 16th European Congress on Digital Pathology, which aimed to predict the HER2 status in breast cancer based only on hematoxylin-eosin-stained tissue samples, thus avoiding specialized techniques. The challenge consisted of a large, annotated, whole-slide images dataset (509), specifically collected for the challenge. Models for predicting HER2 status were presented by 21 teams worldwide. The best-performing models are presented by detailing the network architectures and key parameters. Methods are compared and approaches, core methodologies, and software choices contrasted. Different evaluation metrics are discussed, as well as the performance of the presented models for each of these metrics. Potential differences in ranking that would result from different choices of evaluation metrics highlight the need for careful consideration at the time of their selection, as the results show that some metrics may misrepresent the true potential of a model to solve the problem for which it was developed. The HEROHE dataset remains publicly available to promote advances in the field of computational pathology.

S-Pred: protein structural property prediction using MSA transformer.

Predicting the local structural features of a protein from its amino acid sequence helps its function prediction to be revealed and assists in three-dimensional structural modeling. As the sequence-structure gap increases, prediction methods have been developed to bridge this gap. Additionally, as the size of the structural database and computing power increase, the performance of these methods have also significantly improved. Herein, we present a powerful new tool called S-Pred, which can predict eight-state secondary structures (SS8), accessible surface areas (ASAs), and intrinsically disordered regions (IDRs) from a given sequence. For feature prediction, S-Pred uses multiple sequence alignment (MSA) of a query sequence as an input. The MSA input is converted to features by the MSA Transformer, which is a protein language model that uses an attention mechanism. A long short-term memory (LSTM) was employed to produce the final prediction. The performance of S-Pred was evaluated on several test sets, and the program consistently provided accurate predictions. The accuracy of the SS8 prediction was approximately 76%, and the Pearson's correlation between the experimental and predicted ASAs was 0.84. Additionally, an IDR could be accurately predicted with an F1-score of 0.514. The program is freely available at https://github.com/arontier/S_Pred_Paper and https://ad3.io as a code and a web server.

Utility of artificial intelligence with deep learning of hematoxylin and eosin-stained whole slide images to predict lymph node metastasis in T1 colorectal cancer using endoscopically resected specimens; prediction of lymph node metastasis in T1 colorectal cancer.

BACKGROUND: When endoscopically resected specimens of early colorectal cancer (CRC) show high-risk features, surgery should be performed based on current guidelines because of the high-risk of lymph node metastasis (LNM). The aim of this study was to determine the utility of an artificial intelligence (AI) with deep learning (DL) of hematoxylin and eosin (H&E)-stained endoscopic resection specimens without manual-pixel-level annotation for predicting LNM in T1 CRC. In addition, we assessed AI performance for patients with only submucosal (SM) invasion depth of 1000 to 2000 µm known to be difficult to predict LNM in clinical practice. METHODS: H&E-stained whole slide images (WSIs) were scanned for endoscopic resection specimens of 400 patients who underwent endoscopic treatment for newly diagnosed T1 CRC with additional surgery. The area under the curve (AUC) of the receiver operating characteristic curve was used to determine the accuracy of AI for predicting LNM with a fivefold cross-validation in the training set and in a held-out test set. RESULTS: We developed an AI model using a two-step attention-based DL approach without clinical features (AUC, 0.764). Incorporating clinical features into the model did not improve its prediction accuracy for LNM. Our model reduced unnecessary additional surgery by 15.1% more than using the current guidelines (67.4% vs. 82.5%). In patients with SM invasion depth of 1000 to 2000 µm, the AI avoided 16.1% of unnecessary additional surgery than using the JSCCR guidelines. CONCLUSIONS: Our study is the first to show that AI trained with DL of H&E-stained WSIs has the potential to predict LNM in T1 CRC using only endoscopically resected specimens with conventional histologic risk factors.

A-Prot: protein structure modeling using MSA transformer.

BACKGROUND: The accuracy of protein 3D structure prediction has been dramatically improved with the help of advances in deep learning. In the recent CASP14, Deepmind demonstrated that their new version of AlphaFold (AF) produces highly accurate 3D models almost close to experimental structures. The success of AF shows that the multiple sequence alignment of a sequence contains rich evolutionary information, leading to accurate 3D models. Despite the success of AF, only the prediction code is open, and training a similar model requires a vast amount of computational resources. Thus, developing a lighter prediction model is still necessary. RESULTS: In this study, we propose a new protein 3D structure modeling method, A-Prot, using MSA Transformer, one of the state-of-the-art protein language models. An MSA feature tensor and row attention maps are extracted and converted into 2D residue-residue distance and dihedral angle predictions for a given MSA. We demonstrated that A-Prot predicts long-range contacts better than the existing methods. Additionally, we modeled the 3D structures of the free modeling and hard template-based modeling targets of CASP14. The assessment shows that the A-Prot models are more accurate than most top server groups of CASP14. CONCLUSION: These results imply that A-Prot accurately captures the evolutionary and structural information of proteins with relatively low computational cost. Thus, A-Prot can provide a clue for the development of other protein property prediction methods.

Deep learning-based virtual cytokeratin staining of gastric carcinomas to measure tumor-stroma ratio.

The tumor-stroma ratio (TSR) determined by pathologists is subject to intra- and inter-observer variability. We aimed to develop a computational quantification method of TSR using deep learning-based virtual cytokeratin staining algorithms. Patients with 373 advanced (stage III [n = 171] and IV [n = 202]) gastric cancers were analyzed for TSR. Moderate agreement was observed, with a kappa value of 0.623, between deep learning metrics (dTSR) and visual measurement by pathologists (vTSR) and the area under the curve of receiver operating characteristic of 0.907. Moreover, dTSR was significantly associated with the overall survival of the patients (P = 0.0024). In conclusion, we developed a virtual cytokeratin staining and deep learning-based TSR measurement, which may aid in the diagnosis of TSR in gastric cancer.

Deep Learning Methods for Lung Cancer Segmentation in Whole-Slide Histopathology Images-The ACDC@LungHP Challenge 2019.

Accurate segmentation of lung cancer in pathology slides is a critical step in improving patient care. We proposed the ACDC@LungHP (Automatic Cancer Detection and Classification in Whole-slide Lung Histopathology) challenge for evaluating different computer-aided diagnosis (CADs) methods on the automatic diagnosis of lung cancer. The ACDC@LungHP 2019 focused on segmentation (pixel-wise detection) of cancer tissue in whole slide imaging (WSI), using an annotated dataset of 150 training images and 50 test images from 200 patients. This paper reviews this challenge and summarizes the top 10 submitted methods for lung cancer segmentation. All methods were evaluated using metrics using the precision, accuracy, sensitivity, specificity, and DICE coefficient (DC). The DC ranged from 0.7354 ±0.1149 to 0.8372 ±0.0858. The DC of the best method was close to the inter-observer agreement (0.8398 ±0.0890). All methods were based on deep learning and categorized into two groups: multi-model method and single model method. In general, multi-model methods were significantly better (p 0.01) than single model methods, with mean DC of 0.7966 and 0.7544, respectively. Deep learning based methods could potentially help pathologists find suspicious regions for further analysis of lung cancer in WSI.

Deep learning method for comet segmentation and comet assay image analysis.

Comet assay is a widely used method, especially in the field of genotoxicity, to quantify and measure DNA damage visually at the level of individual cells with high sensitivity and efficiency. Generally, computer programs are used to analyze comet assay output images following two main steps. First, each comet region must be located and segmented, and next, it is scored using common metrics (e.g., tail length and tail moment). Currently, most studies on comet assay image analysis have adopted hand-crafted features rather than the recent and effective deep learning (DL) methods. In this paper, however, we propose a DL-based baseline method, called DeepComet, for comet segmentation. Furthermore, we created a trainable and testable comet assay image dataset that contains 1037 comet assay images with 8271 manually annotated comet objects. From the comet segmentation test results with the proposed dataset, the DeepComet achieves high average precision (AP), which is an essential metric in image segmentation and detection tasks. A comparative analysis was performed between the DeepComet and the state-of-the-arts automatic comet segmentation programs on the dataset. Besides, we found that the DeepComet records high correlations with a commercial comet analysis tool, which suggests that the DeepComet is suitable for practical application.