Independent Test Set Research Articles

Discovery of potential antidiabetic peptides using deep learning

Antidiabetic peptides (ADPs), peptides with potential antidiabetic activity, hold significant importance in the treatment and control of diabetes. Despite their therapeutic potential, the discovery and prediction of ADPs remain challenging due to limited data, the complex nature of peptide functions, and the expensive and time-consuming nature of traditional wet lab experiments. This study aims to address these challenges by exploring methods for the discovery and prediction of ADPs using advanced deep learning techniques. Specifically, we developed two models: a single-channel CNN and a three-channel neural network (CNN + RNN + Bi-LSTM). ADPs were primarily gathered from the BioDADPep database, alongside thousands of non-ADPs sourced from anticancer, antibacterial, and antiviral peptide datasets. Subsequently, data preprocessing was performed with the evolutionary scale model (ESM-2), followed by model training and evaluation through 10-fold cross-validation. Furthermore, this work collected a series of newly published ADPs as an independent test set through literature review, and found that the CNN model achieved the highest accuracy (90.48 %) in predicting the independent test set, surpassing existing ADP prediction tools. Finally, the application of the model was considered. SeqGAN was used to generate new candidate ADPs, followed by screening with the constructed CNN model. Selected peptides were then evaluated using physicochemical property prediction and structural forecasts for pharmaceutical potential. In summary, this study not only established robust ADP prediction models but also employed these models to screen a batch of potential ADPs, addressing a critical need in the field of peptide-based antidiabetic research.

I-Biomarker CaDx: A circulating miRNA-based multi-cancer detection test with generative and explainable AI for colorectal cancer.

64 Background: Colorectal cancer (CRC), the second most deadly cancer, underscores the critical need for early detection to significantly improve treatment outcomes and survival rates. Colonoscopy, flexible sigmoidoscopy, and the fecal immunochemical tests (FIT), often fail to capture early disease, significantly decreasing the survival chance. Circulating miRNA is a promising non-invasive biomarker for various cancers, including CRC. The i-Biomarker CaDx (patent pending) is an innovative multi-cancer early detection and diagnosis platform, covering 32 cancer types with an impressive accuracy rate of 99-100%. This study focuses on the effectiveness of i-Biomarker CaDx for CRC, employing significant data collection and leveraging insights from Generative AI (GenAI) and Explainable Artificial Intelligence (XAI). Methods: We collected various available datasets (e.g. GSE106817, etc.) encompassing 305 colorectal cancer patients and a matching number of healthy controls. Microarray was employed to profile circulating miRNAs. We used various classification paradigms, e.g., decision trees, neural networks, etc. Hyperparameter optimization was conducted, and the best-performing classifiers were amalgamated in the final model, weighted by their respective efficacies. We assessed i-Biomarker CaDx's performance through cross-validation and independent test sets. The XAI provided detailed insights into miRNA variations associated with personalized diagnostic. GenAI integrated the test into the screening and early detection best practice workflows and performs Functional Analysis. Results: i-Biomarker CaDx shows exceptional diagnostic accuracy of 99-100%, surpassing conventional colorectal cancer diagnostic methods like colonoscopy, flexible sigmoidoscopy, and FIT. It can also be used for personalized treatment response monitoring. I-Biomarker effectively deciphers intricate miRNA relationships relevant to colorectal cancer using XAI. Thus, our analyses shed light on miRNA patterns and their associations with colorectal cancer, enhancing the understanding of the underlying molecular complexities of the disease. It goes further by explaining the personalized cellular and molecular mechanisms involved, and it will be integrated into the corresponding medical workflows. Conclusions: Our AI-powered multi-cancer early diagnostic platform demonstrates outstanding performance in detecting CRC, surpassing traditional diagnostic methods like colonoscopy, flexible sigmoidoscopy, and FIT. XAI allows for in-depth exploration of miRNA alterations and their impact on CRC, enriching our interpretations at both the population and individual levels. These findings underscore the substantial potential of i-Biomarker CaDx as a transformative, non-invasive diagnostic tool for CRC.

A novel deep learning model for obstructive sleep apnea diagnosis: hybrid CNN-Transformer approach for radar-based detection of apnea-hypopnea events.

The demand for cost-effective and accessible alternatives to polysomnography (PSG), the conventional diagnostic method for obstructive sleep apnea (OSA), has surged. In this study, we have developed and validated a deep learning model for detecting apnea-hypopnea events using radar data. We conducted a single-center prospective cohort study, dividing participants with suspected sleep-disordered breathing into development and temporally independent test sets. Utilizing a hybrid CNN-Transformer architecture, we performed 5-fold cross-validation on the development set to develop and subsequently validate the model. Evaluation metrics included sensitivity for event detection, mean absolute error (MAE), intraclass correlation coefficient (ICC), and Pearson correlation coefficient (r) for apnea-hypopnea index (AHI) estimation. Linearly weighted kappa statistics (κ) assessed OSA severity. The development set comprised 54 participants (July 2021-May 2022), while the test set included 35 participants (June 2022-June 2023). In the test set, our model achieved an event detection sensitivity of 67.2% (95% CI: 65.8%, 68.5%) and demonstrated a MAE of 7.54 (95% CI: 5.36, 9.72), indicating good agreement (ICC = 0.889 [95% CI: 0.792, 0.942]) and a strong correlation (r = 0.892 [95% CI: 0.795, 0.945]) with the ground truth for AHI estimation. Furthermore, OSA severity estimation showed substantial agreement (κ = 0.780 [95% CI: 0.658, 0.903]). Our study highlights radar sensors and advanced AI models' potential to improve OSA diagnosis, paving the path for future radar-based diagnostic models in sleep medicine research.

AI-based automated segmentation for ovarian/adnexal masses and their internal components on ultrasound imaging.

Segmentation of ovarian/adnexal masses from surrounding tissue on ultrasound images is a challenging task. The separation of masses into different components may also be important for radiomic feature extraction. Our study aimed to develop an artificial intelligence-based automatic segmentation method for transvaginal ultrasound images that (1)outlines the exterior boundary of adnexal masses and (2)separates internal components. A retrospective ultrasound imaging database of adnexal masses was reviewed for exclusion criteria at the patient, mass, and image levels, with one image per mass. The resulting 54 adnexal masses (36 benign/18 malignant) from 53 patients were separated by patient into training (26 benign/12 malignant) and independent test (10 benign/6 malignant) sets. U-net segmentation performance on test images compared to expert detailed outlines was measured using the Dice similarity coefficient (DSC) and the ratio of the Hausdorff distance to the effective diameter of the outline ( ) for each mass. Subsequently, in discovery mode, a two-level fuzzy c-means (FCM) unsupervised clustering approach was used to separate the pixels within masses belonging to hypoechoic or hyperechoic components. The DSC (median [95% confidence interval]) was 0.91 [0.78, 0.96], and was 0.04 [0.01, 0.12], indicating strong agreement with expert outlines. Clinical review of the internal separation of masses into echogenic components demonstrated a strong association with mass characteristics. A combined U-net and FCM algorithm for automatic segmentation of adnexal masses and their internal components achieved excellent results compared with expert outlines and review, supporting future radiomic feature-based classification of the masses by components.

Abstract Mo125: Translating ECG responses to drugs across sexes: model development and validation in the clinic

Background: Sex-based differences in cardiac electrophysiology and their potential to impact susceptibility to cardiac arrhythmia and response to treatment are widely recognized. For example, female sex is an independent risk factor for drug-induced Long QT syndrome, Torsade de Pointes, and sudden cardiac death. However, arrhythmia studies have been predominantly performed in males, thus limiting our understanding of sex-specific arrhythmia mechanisms and the development of sex-specific metrics for arrhythmia risk. Computational tools are a powerful approach for integrating sex differences across scales and studying such multi-level phenomena, accelerating challenging and costly human and animal experiments. Aim: To develop a quantitative tool that predicts drug-induced responses in female electrocardiogram (ECG) features based on ECG data collected in males (and vice versa). Methods: Male and female 1D cable models of ventricular cardiomyocytes were constructed by incorporating transmural heterogeneity and sex-specific parameterizations. We constructed populations of male and female models by randomly perturbing baseline parameters, and derived pseudo-ECGs from computed extracellular potentials. We then utilized lasso regression to map ECG features (QRS duration, QT interval, T-peak-end duration, and T-wave amplitude) extracted from the male pseudo-ECG onto the female ECG features (and vice versa) ( Fig 1 ). Results: Our simulated data-driven translator accurately predicted female pseudo-ECG features from male data in an independent simulated test set (R 2 >0.95). This predictive capability was further assessed in an independent dataset simulating the effects of 95 drugs at different concentrations ( Fig 2A ), exhibiting an average translation error <1% for the QT interval ( Fig 2B ). Proof-of-concept application to ECG data sourced from publicly available FDA-approved datasets, in patients of both sexes under various anti-arrhythmic drugs and drug combinations, demonstrates high accuracy ( Fig 3 ) and validates the translator's applicability to the clinic. Conclusions: Our findings underscore the translator's potential in conducting sex-specific cardiac safety and efficacy assessments, enhancing the precision of evaluations in females. We plan to expand our translator capability to predict sex-specific effects of perturbations, including physiological and pathophysiological stressors.

MetalTrans: A Biological Language Model-Based Approach for Predicting Disease-Associated Mutations in Protein Metal-Binding Sites.

The critical importance of accurately predicting mutations in protein metal-binding sites for advancing drug discovery and enhancing disease diagnostic processes cannot be overstated. In response to this imperative, MetalTrans emerges as an accurate predictor for disease-associated mutations in protein metal-binding sites. The core innovation of MetalTrans lies in its seamless integration of multifeature splicing with the Transformer framework, a strategy that ensures exhaustive feature extraction. Central to MetalTrans's effectiveness is its deep feature combination strategy, which merges evolutionary-scale modeling amino acid embeddings with ProtTrans embeddings, thus shedding light on the biochemical properties of proteins. Employing the Transformer component, MetalTrans leverages the self-attention mechanism to delve into higher-level representations. Utilizing mutation site information for feature fusion not only enriches the feature set but also sidesteps the common pitfall of overestimation linked to protein sequence-based predictions. This nuanced approach to feature fusion is a key differentiator, enabling MetalTrans to outperform existing methods significantly, as evidenced by comparative analyses. Our evaluations across varied metal binding site data sets (specifically Zn, Ca, Mg, and Mix) underscore MetalTrans's superior performance, which achieved the average AUC values of 0.971, 0.965, 0.980, and 0.945 on multiple 5-fold cross-validation, respectively. Remarkably, against the multichannel convolutional neural network method on a benchmark independent test set, MetalTrans demonstrated unparalleled robustness and superiority, boasting the AUC score of 0.998 on multiple 5-fold cross-validation. Our comprehensive examination of the predicted outcomes further confirms the effectiveness of the model. The source codes, data sets, and prediction results for MetalTrans can be accessed for academic usage at https://github.com/EduardWang/MetalTrans.

Integrating MRI-based radiomics and clinicopathological features for preoperative prognostication of early-stage cervical adenocarcinoma patients: in comparison to deep learning approach

ObjectivesThe roles of magnetic resonance imaging (MRI) -based radiomics approach and deep learning approach in cervical adenocarcinoma (AC) have not been explored. Herein, we aim to develop prognosis-predictive models based on MRI-radiomics and clinical features for AC patients.MethodsClinical and pathological information from one hundred and ninety-seven patients with cervical AC was collected and analyzed. For each patient, 107 radiomics features were extracted from T2-weighted MRI images. Feature selection was performed using Spearman correlation and random forest (RF) algorithms, and predictive models were built using support vector machine (SVM) technique. Deep learning models were also trained with T2-weighted MRI images and clinicopathological features through Convolutional Neural Network (CNN). Kaplan-Meier curve was analyzed using significant features. In addition, information from another group of 56 AC patients was used for the independent validation.ResultsA total of 107 radiomics features and 6 clinicopathological features (age, FIGO stage, differentiation, invasion depth, lymphovascular space invasion (LVSI), and lymph node metastasis (LNM) were included in the analysis. When predicting the 3-year, 4-year, and 5-year DFS, the model trained solely on radiomics features achieved AUC values of 0.659 (95%CI: 0.620–0.716), 0.791 (95%CI: 0.603–0.922), and 0.853 (95%CI: 0.745–0.912), respectively. However, the combined model, incorporating both radiomics and clinicopathological features, outperformed the radiomics model with AUC values of 0.934 (95%CI: 0.885–0.981), 0.937 (95%CI: 0.867–0.995), and 0.916 (95%CI: 0.857–0.970), respectively. For deep learning models, the MRI-based models achieved an AUC of 0.857, 0.777 and 0.828 for 3-year DFS, 4-year DFS and 5-year DFS prediction, respectively. And the combined deep learning models got a improved performance, the AUCs were 0.903. 0.862 and 0.969. In the independent test set, the combined model achieved an AUC of 0.873, 0.858 and 0.914 for 3-year DFS, 4-year DFS and 5-year DFS prediction, respectively.ConclusionsWe demonstrated the prognostic value of integrating MRI-based radiomics and clinicopathological features in cervical adenocarcinoma. Both radiomics and deep learning models showed improved predictive performance when combined with clinical data, emphasizing the importance of a multimodal approach in patient management.

Stacking-Kcr: A Stacking Model for Predicting the Crotonylation Sites of Lysine by Fusing Serial and Automatic Encoder

Background: Protein lysine crotonylation (Kcr), a newly discovered important posttranslational modification (PTM), is typically localized at the transcription start site and regulates gene expression, which is associated with a variety of pathological conditions such as developmental defects and malignant transformation. Objective: Identifying Kcr sites is advantageous for the discovery of its biological mechanism and the development of new drugs for related diseases. However, traditional experimental methods for identifying Kcr sites are expensive and inefficient, necessitating the development of new computational techniques. Methods: In this work, to accurately identify Kcr sites, we propose a model for ensemble learning called Stacking-Kcr. Firstly, extract features from sequence information, physicochemical properties, and sequence fragment similarity. Then, the two characteristics of sequence information and physicochemical properties are fused using automatic encoder and serial, respectively. Finally, the fused two features and sequence fragment similarity features are then respectively input into the four base classifiers, a meta classifier is constructed using the first level prediction results, and the final forecasting results are obtained. Results: The five-fold cross-validation of this model has achieved an accuracy of 0.828 and an AUC of 0.910. This shows that the Stacking-Kcr method has obvious advantages over traditional machine learning methods. On independent test sets, Stacking-Kcr achieved an accuracy of 84.89% and an AUC of 92.21%, which was higher than 1.7% and 0.8% of other state-of-the-art tools. Additionally, we trained Stacking-Kcr on the phosphorylation site, and the result is superior to the current model. Conclusion: These outcomes are additional evidence that Stacking-Kcr has strong application potential and generalization performance.

Artificial intelligence improves the diagnosis of human leukocyte antigen (HLA)-B27-negative axial spondyloarthritis based on multi-sequence magnetic resonance imaging and clinical features.

Axial spondyloarthritis (axSpA) is frequently diagnosed late, particularly in human leukocyte antigen (HLA)-B27-negative patients, resulting in a missed opportunity for optimal treatment. This study aimed to develop an artificial intelligence (AI) tool, termed NegSpA-AI, using sacroiliac joint (SIJ) magnetic resonance imaging (MRI) and clinical SpA features to improve the diagnosis of axSpA in HLA-B27-negative patients. We retrospectively included 454 HLA-B27-negative patients with rheumatologist-diagnosed axSpA or other diseases (non-axSpA) from the Third Affiliated Hospital of Southern Medical University and Nanhai Hospital between January 2010 and August 2021. They were divided into a training set (n=328) for 5-fold cross-validation, an internal test set (n=72), and an independent external test set (n=54). To construct a prospective test set, we further enrolled 87 patients between September 2021 and August 2023 from the Third Affiliated Hospital of Southern Medical University. MRI techniques employed included T1-weighted (T1W), T2-weighted (T2W), and fat-suppressed (FS) sequences. We developed NegSpA-AI using a deep learning (DL) network to differentiate between axSpA and non-axSpA at admission. Furthermore, we conducted a reader study involving 4 radiologists and 2 rheumatologists to evaluate and compare the performance of independent and AI-assisted clinicians. NegSpA-AI demonstrated superior performance compared to the independent junior rheumatologist (≤5 years of experience), achieving areas under the curve (AUCs) of 0.878 [95% confidence interval (CI): 0.786-0.971], 0.870 (95% CI: 0.771-0.970), and 0.815 (95% CI: 0.714-0.915) on the internal, external, and prospective test sets, respectively. The assistance of NegSpA-AI promoted discriminating accuracy, sensitivity, and specificity of independent junior radiologists by 7.4-11.5%, 1.0-13.3%, and 7.4-20.6% across the 3 test sets (all P<0.05). On the prospective test set, AI assistance also improved the diagnostic accuracy, sensitivity, and specificity of independent junior rheumatologists by 7.7%, 7.7%, and 6.9%, respectively (all P<0.01). The proposed NegSpA-AI effectively improves radiologists' interpretations of SIJ MRI and rheumatologists' diagnoses of HLA-B27-negative axSpA.

Accelerating breast MRI acquisition with generative AI models.

To investigate the use of the score-based diffusion model to accelerate breast MRI reconstruction. We trained a score-based model on 9549 MRI examinations of the female breast and employed it to reconstruct undersampled MRI images with undersampling factors of 2, 5, and 20. Images were evaluated by two experienced radiologists who rated the images based on their overall quality and diagnostic value on an independent test set of 100 additional MRI examinations. The score-based model produces MRI images of high quality and diagnostic value. Both T1- and T2-weighted MRI images could be reconstructed to a high degree of accuracy. Two radiologists rated the images as almost indistinguishable from the original images (rating 4 or 5 on a scale of 5) in 100% (radiologist 1) and 99% (radiologist 2) of cases when the acceleration factor was 2. This fraction dropped to 88% and 70% for an acceleration factor of 5 and to 5% and 21% with an extreme acceleration factor of 20. Score-based models can reconstruct MRI images at high fidelity, even at comparatively high acceleration factors, but further work on a larger scale of images is needed to ensure that diagnostic quality holds. The number of MRI examinations of the breast is expected to rise with MRI screening recommended for women with dense breasts. Accelerated image acquisition methods can help in making this examination more accessible. Accelerating breast MRI reconstruction remains a significant challenge in clinical settings. Score-based diffusion models can achieve near-perfect reconstruction for moderate undersampling factors. Faster breast MRI scans with maintained image quality could revolutionize clinic workflows and patient experience.

Utilizing active learning strategies in machine-assisted annotation for clinical named entity recognition: a comprehensive analysis considering annotation costs and target effectiveness.

Active learning (AL) has rarely integrated diversity-based and uncertainty-based strategies into a dynamic sampling framework for clinical named entity recognition (NER). Machine-assisted annotation is becoming popular for creating gold-standard labels. This study investigated the effectiveness of dynamic AL strategies under simulated machine-assisted annotation scenarios for clinical NER. We proposed 3 new AL strategies: a diversity-based strategy (CLUSTER) based on Sentence-BERT and 2 dynamic strategies (CLC and CNBSE) capable of switching from diversity-based to uncertainty-based strategies. Using BioClinicalBERT as the foundational NER model, we conducted simulation experiments on 3 medication-related clinical NER datasets independently: i2b2 2009, n2c2 2018 (Track 2), and MADE 1.0. We compared the proposed strategies with uncertainty-based (LC and NBSE) and passive-learning (RANDOM) strategies. Performance was primarily measured by the number of edits made by the annotators to achieve a desired target effectiveness evaluated on independent test sets. When aiming for 98% overall target effectiveness, on average, CLUSTER required the fewest edits. When aiming for 99% overall target effectiveness, CNBSE required 20.4% fewer edits than NBSE did. CLUSTER and RANDOM could not achieve such a high target under the pool-based simulation experiment. For high-difficulty entities, CNBSE required 22.5% fewer edits than NBSE to achieve 99% target effectiveness, whereas neither CLUSTER nor RANDOM achieved 93% target effectiveness. When the target effectiveness was set high, the proposed dynamic strategy CNBSE exhibited both strong learning capabilities and low annotation costs in machine-assisted annotation. CLUSTER required the fewest edits when the target effectiveness was set low.

SAFER: sub-hypergraph attention-based neural network for predicting effective responses to dose combinations

BackgroundThe potential benefits of drug combination synergy in cancer medicine are significant, yet the risks must be carefully managed due to the possibility of increased toxicity. Although artificial intelligence applications have demonstrated notable success in predicting drug combination synergy, several key challenges persist: (1) Existing models often predict average synergy values across a restricted range of testing dosages, neglecting crucial dose amounts and the mechanisms of action of the drugs involved. (2) Many graph-based models rely on static protein–protein interactions, failing to adapt to dynamic and higher-order relationships. These limitations constrain the applicability of current methods.ResultsWe introduce SAFER, a Sub-hypergraph Attention-based graph model, addressing these issues by incorporating complex relationships among biological knowledge networks and considering dosing effects on subject-specific networks. SAFER outperformed previous models on the benchmark and the independent test set. The analysis of subgraph attention weight for the lung cancer cell line highlighted JAK-STAT signaling pathway, PRDM12, ZNF781, and CDC5L that have been implicated in lung fibrosis.ConclusionsSAFER presents an interpretable framework designed to identify drug-responsive signals. Tailored for comprehending dose effects on subject-specific molecular contexts, our model uniquely captures dose-level drug combination responses. This capability unlocks previously inaccessible avenues of investigation compared to earlier models. Furthermore, the SAFER framework can be leveraged by future inquiries to investigate molecular networks that uniquely characterize individual patients and can be applied to prioritize personalized effective treatment based on safe dose combinations.

Prediction of Protein Allosteric Sites with Transfer Entropy and Spatial Neighbor-Based Evolutionary Information Learned by an Ensemble Model.

Allostery is one of the most direct and efficient ways to regulate protein functions. The diverse allosteric sites make it possible to design allosteric modulators of differential selectivity and improved safety compared with those of orthosteric drugs targeting conserved orthosteric sites. Here, we develop an ensemble machine learning method AllosES to predict protein allosteric sites in which the new and effective features are utilized, including the entropy transfer-based dynamic property, secondary structure features, and our previously proposed spatial neighbor-based evolutionary information besides the traditional physicochemical properties. To overcome the class imbalance problem, the multiple grouping strategy is proposed, which is applied to feature selection and model construction. The ensemble model is constructed where multiple submodels are trained on multiple training subsets, respectively, and their results are then integrated to be the final output. AllosES achieves a prediction performance of 0.556 MCC on the independent test set D24, and additionally, AllosES can rank the real allosteric sites in the top three for 83.3/89.3% of allosteric proteins from the test set D24/D28, outperforming the state-of-the-art peer methods. The comprehensive results demonstrate that AllosES is a promising method for protein allosteric site prediction. The source code is available at https://github.com/ChunhuaLab/AllosES.

Comparison between 16S rRNA and shotgun sequencing in colorectal cancer, advanced colorectal lesions, and healthy human gut microbiota

BackgroundGut dysbiosis has been associated with colorectal cancer (CRC), the third most prevalent cancer in the world. This study compares microbiota taxonomic and abundance results obtained by 16S rRNA gene sequencing (16S) and whole shotgun metagenomic sequencing to investigate their reliability for bacteria profiling. The experimental design included 156 human stool samples from healthy controls, advanced (high-risk) colorectal lesion patients (HRL), and CRC cases, with each sample sequenced using both 16S and shotgun methods. We thoroughly compared both sequencing technologies at the species, genus, and family annotation levels, the abundance differences in these taxa, sparsity, alpha and beta diversities, ability to train prediction models, and the similarity of the microbial signature derived from these models.ResultsAs expected, the results showed that 16S detects only part of the gut microbiota community revealed by shotgun, although some genera were only profiled by 16S. The 16S abundance data was sparser and exhibited lower alpha diversity. In lower taxonomic ranks, shotgun and 16S highly differed, partially due to a disagreement in reference databases. When considering only shared taxa, the abundance was positively correlated between the two strategies. We also found a moderate correlation between the shotgun and 16S alpha-diversity measures, as well as their PCoAs. Regarding the machine learning models, only some of the shotgun models showed some degree of predictive power in an independent test set, but we could not demonstrate a clear superiority of one technology over the other. Microbial signatures from both sequencing techniques revealed taxa previously associated with CRC development, e.g., Parvimonas micra.ConclusionsShotgun and 16S sequencing provide two different lenses to examine microbial communities. While we have demonstrated that they can unravel common patterns (including microbial signatures), shotgun often gives a more detailed snapshot than 16S, both in depth and breadth. Instead, 16S will tend to show only part of the picture, giving greater weight to dominant bacteria in a sample. Therefore, we recommend choosing one or another sequencing technique before launching a study. Specifically, shotgun sequencing is preferred for stool microbiome samples and in-depth analyses, while 16S is more suitable for tissue samples and studies with targeted aims.

An artificial intelligence platform for the screening and managing of strabismus.

Considering the escalating incidence of strabismus and its consequential jeopardy to binocular vision, there is an imperative demand for expeditious and precise screening methods. This study was to develop an artificial intelligence (AI) platform in the form of an applet that facilitates the screening and management of strabismus on any mobile device. The Visual Transformer (VIT_16_224) was developed using primary gaze photos from two datasets covering different ages. The AI model was evaluated by 5-fold cross-validation set and tested on an independent test set. The diagnostic performance of the AI model was assessed by calculating the Accuracy, Precision, Specificity, Sensitivity, F1-Score and Area Under the Curve (AUC). A total of 6194 photos with corneal light-reflection (with 2938 Exotropia, 1415 Esotropia, 739 Vertical Deviation and 1562 Orthotropy) were included. In the internal validation set, the AI model achieved an Accuracy of 0.980, Precision of 0.941, Specificity of 0.979, Sensitivity of 0.958, F1-Score of 0.951 and AUC of 0.994. In the independent test set, the AI model achieved an Accuracy of 0.967, Precision of 0.980, Specificity of 0.970, Sensitivity of 0.960, F1-Score of 0.975 and AUC of 0.993. Our study presents an advanced AI model for strabismus screening which integrates electronic archives for comprehensive patient histories. Additionally, it includes a patient-physician interaction module for streamlined communication. This innovative platform offers a complete solution for strabismus care, from screening to long-term follow-up, advancing ophthalmology through AI technology for improved patient outcomes and eye care quality.

Development of a deep learning model for cancer diagnosis by inspecting cell-free DNA end-motifs

Accurate discrimination between patients with and without cancer from cfDNA is crucial for early cancer diagnosis. Herein, we develop and validate a deep-learning-based model entitled end-motif inspection via transformer (EMIT) for discriminating individuals with and without cancer by learning feature representations from cfDNA end-motifs. EMIT is a self-supervised learning approach that models rankings of cfDNA end-motifs. We include 4606 samples subjected to different types of cfDNA sequencing to develop EIMIT, and subsequently evaluate classification performance of linear projections of EMIT on six datasets and an additional inhouse testing set encopassing whole-genome, whole-genome bisulfite and 5-hydroxymethylcytosine sequencing. The linear projection of representations from EMIT achieved area under the receiver operating curve (AUROC) values ranged from 0.895 (0.835–0.955) to 0.996 (0.994–0.997) across these six datasets, outperforming its baseline by significant margins. Additionally, we showed that linear projection of EMIT representations can achieve an AUROC of 0.962 (0.914–1.0) in identification of lung cancer on an independent testing set subjected to whole-exome sequencing. The findings of this study indicate that a transformer-based deep learning model can learn cancer-discrimative representations from cfDNA end-motifs. The representations of this deep learning model can be exploited for discriminating patients with and without cancer.

Leveraging protein language model embeddings and logistic regression for efficient and accurate in-silico acidophilic proteins classification

The increasing demand for eco-friendly technologies in biotechnology necessitates effective and sustainable catalysts. Acidophilic proteins, functioning optimally in highly acidic environments, hold immense promise for various applications, including food production, biofuels, and bioremediation. However, limited knowledge about these proteins hinders their exploration. This study addresses this gap by employing in silico methods utilizing computational tools and machine learning. We propose a novel approach to predict acidophilic proteins using protein language models (PLMs), accelerating discovery without extensive lab work. Our investigation highlights the potential of PLMs in understanding and harnessing acidophilic proteins for scientific and industrial advancements. We introduce the ACE model, which combines a simple Logistic Regression model with embeddings derived from protein sequences processed by the ProtT5 PLM. This model achieves high performance on an independent test set, with accuracy (0.91), F1-score (0.93), and Matthew's correlation coefficient (0.76). To our knowledge, this is the first application of pre-trained PLM embeddings for acidophilic protein classification. The ACE model serves as a powerful tool for exploring protein acidophilicity, paving the way for future advancements in protein design and engineering.

PLM_Sol: predicting protein solubility by benchmarking multiple protein language models with the updated Escherichia coli protein solubility dataset.

Protein solubility plays a crucial role in various biotechnological, industrial, and biomedical applications. With the reduction in sequencing and gene synthesis costs, the adoption of high-throughput experimental screening coupled with tailored bioinformatic prediction has witnessed a rapidly growing trend for the development of novel functional enzymes of interest (EOI). High protein solubility rates are essential in this process and accurate prediction of solubility is a challenging task. As deep learning technology continues to evolve, attention-based protein language models (PLMs) can extract intrinsic information from protein sequences to a greater extent. Leveraging these models along with the increasing availability of protein solubility data inferred from structural database like the Protein Data Bank holds great potential to enhance the prediction of protein solubility. In this study, we curated an Updated Escherichia coli protein Solubility DataSet (UESolDS) and employed a combination of multiple PLMs and classification layers to predict protein solubility. The resulting best-performing model, named Protein Language Model-based protein Solubility prediction model (PLM_Sol), demonstrated significant improvements over previous reported models, achieving a notable 6.4% increase in accuracy, 9.0% increase in F1_score, and 11.1% increase in Matthews correlation coefficient score on the independent test set. Moreover, additional evaluation utilizing our in-house synthesized protein resource as test data, encompassing diverse types of enzymes, also showcased the good performance of PLM_Sol. Overall, PLM_Sol exhibited consistent and promising performance across both independent test set and experimental set, thereby making it well suited for facilitating large-scale EOI studies. PLM_Sol is available as a standalone program and as an easy-to-use model at https://zenodo.org/doi/10.5281/zenodo.10675340.

Prognostic Utility of the Flow Cytometry and Clonality Analysis Results for Feline Lymphomas.

Feline lymphoma, a prevalent cancer in cats, exhibits varied prognoses influenced by anatomical site and cellular characteristics. In this study, we investigated the utility of flow cytometry and clonality analysis via PCR for antigen receptor rearrangement (PARR) with respect to characterizing the disease and predicting prognosis. For this purpose, we received fine needle aspirates and/or blood from 438 feline patients, which were subjected to flow cytometry analysis and PARR. We used a subset of the results from patients with confirmed B- or T-cell lymphomas for comparison to cytological or histological evaluation (n = 53). Using them as a training set, we identified the optimal set of flow cytometry parameters, namely forward scatter thresholds, for cell size categorization by correlating with cytology-defined sizes. Concordance with cytological sizing among this training set was 82%. Furthermore, 90% concordance was observed when the proposed cell sizing was tested on an independent test set (n = 24), underscoring the reliability of the proposed approach. Additionally, lymphoma subtypes defined by flow cytometry and PARR demonstrated significant survival differences, validating the prognostic utility of these methods. The proposed methodology achieves high concordance with cytological evaluations and provides an additional tool for the characterization and management of feline lymphoproliferative diseases.

Interpretable Machine Learning-Based Risk Score for Predicting Ten-Year Corneal Graft Survival After Penetrating Keratoplasty and Deep Anterior Lamellar Keratoplasty in Asian Eyes.

To predict 10-year graft survival after deep anterior lamellar keratoplasty (DALK) and penetrating keratoplasty (PK) using a machine learning (ML)-based interpretable risk score. Singapore Corneal Transplant Registry patients (n = 1687) who underwent DALK (n = 524) or PK (n = 1163) for optical indications (excluding endothelial diseases) were followed up for 10 years. Variable importance scores from random survival forests were used to identify variables associated with graft survival. Parsimonious analysis using nested Cox models selected the top factors. An ML-based clinical score generator (AutoScore) converted identified variables into an interpretable risk score. Predictive performance was evaluated using Kaplan-Meier (KM) curves and time-integrated AUC (iAUC) on an independent testing set. Mean recipient age was 51.8 years, 54.1% were male, and majority were Chinese (60.0%). Surgical indications included corneal scar (46.5%), keratoconus (18.3%), and regraft (16.2%). Five-year and ten-year KM survival was 93.4% and 92.3% for DALK, compared with 67.6% and 56.6% for PK (log-rank P < 0.001). Five factors were identified by ML algorithm as predictors of 10-year graft survival: recipient sex, preoperative visual acuity, choice of procedure, surgical indication, and active inflammation. AutoScore stratified participants into low-risk and high-risk groups-with KM survival of 73.6% and 39.0%, respectively (log-rank P < 0.001). ML analysis outperformed traditional Cox regression in predicting graft survival beyond 5 years (iAUC 0.75 vs. 0.69). A combination of ML and traditional techniques identified factors associated with graft failure to derive a clinically interpretable risk score to stratify PK and DALK patients-a technique that may be replicated in other corneal transplant programs.