Independent Validation Dataset Research Articles

CBIL-VHPLI: a model for predicting viral-host protein-lncRNA interactions based on machine learning and transfer learning

Virus‒host protein‒lncRNA interaction (VHPLI) predictions are critical for decoding the molecular mechanisms of viral pathogens and host immune processes. Although VHPLI interactions have been predicted in both plants and animals, they have not been extensively studied in viruses. For the first time, we propose a new deep learning-based approach that consists mainly of a convolutional neural network and bidirectional long and short-term memory network modules in combination with transfer learning named CBIL‒VHPLI to predict viral–host protein‒lncRNA interactions. The models were first trained on large and diverse datasets (including plants, animals, etc.). Protein sequence features were extracted using a k-mer method combined with the one-hot encoding and composition–transition–distribution (CTD) methods, and lncRNA sequence features were extracted using a k-mer method combined with the one-hot encoding and Z curve methods. The results obtained on three independent external validation datasets showed that the pre-trained CBIL‒VHPLI model performed the best with an accuracy of approximately 0.9. Pretraining was followed by conducting transfer learning on a viral protein–human lncRNA dataset, and the fine-tuning results showed that the accuracy of CBIL‒VHPLI was 0.946, which was significantly greater than that of the previous models. The final case study results showed that CBIL‒VHPLI achieved a prediction reproducibility rate of 91.6% for the RIP-Seq experimental screening results. This model was then used to predict the interactions between human lncRNA PIK3CD-AS2 and the nonstructural protein 1 (NS1) of the H5N1 virus, and RNA pull-down experiments were used to prove the prediction readiness of the model in terms of prediction. The source code of CBIL‒VHPLI and the datasets used in this work are available at https://github.com/Liu-Lab-Lnu/CBIL-VHPLI for academic usage.

Unraveling pathogenesis, biomarkers and potential therapeutic agents for endometriosis associated with disulfidptosis based on bioinformatics analysis, machine learning and experiment validation

BackgroundEndometriosis (EMs) is an enigmatic disease of yet-unknown pathogenesis. Disulfidptosis, a novel identified form of programmed cell death resulting from disulfide stress, stands a chance of treating diverse ailments. However, the potential roles of disulfidptosis-related genes (DRGs) in EMs remain elusive. This study aims to thoroughly explore the key disulfidptosis genes involved in EMs, and probe novel diagnostic markers and candidate therapeutic compounds from the aspect of disulfidptosis based on bioinformatics analysis, machine learning, and animal experiments.ResultsEnrichment analysis on key module genes and differentially expressed genes (DEGs) of eutopic and ectopic endometrial tissues in EMs suggested that EMs was closely related to disulfidptosis. And then, we obtained 20 and 16 disulfidptosis-related DEGs in eutopic and ectopic endometrial tissue, respectively. The protein-protein interaction (PPI) network revealed complex interactions between genes, and screened nine and ten hub genes in eutopic and ectopic endometrial tissue, respectively. Furthermore, immune infiltration analysis uncovered distinct differences in the immunocyte, human leukocyte antigen (HLA) gene set, and immune checkpoints in the eutopic and ectopic endometrial tissues when compared with health control. Besides, the hub genes mentioned above showed a close correlation with the immune microenvironment of EMs. Furthermore, four machine learning algorithms were applied to screen signature genes in eutopic and ectopic endometrial tissue, including the binary logistic regression (BLR), the least absolute shrinkage and selection operator (LASSO), the support vector machine-recursive feature elimination (SVM-RFE), and the extreme gradient boosting (XGBoost). Model training and hyperparameter tuning were implemented on 80% of the data using a ten-fold cross-validation method, and tested in the testing sets which determined the excellent diagnostic performance of these models by six indicators (Sensitivity, Specificity, Positive Predictive Value, Negative Predictive Value, Accuracy, and Area Under Curve). And seven eutopic signature genes (ACTB, GYS1, IQGAP1, MYH10, NUBPL, SLC7A11, TLN1) and five ectopic signature genes (CAPZB, CD2AP, MYH10, OXSM, PDLIM1) were finally identified based on machine learning. The independent validation dataset also showed high accuracy of the signature genes (IQGAP1, SLC7A11, CD2AP, MYH10, PDLIM1) in predicting EMs. Moreover, we screened 12 specific compounds for EMs based on ectopic signature genes and the pharmacological impact of tretinoin on signature genes was further verified in the ectopic lesion in the EMs murine model.ConclusionThis study verified a close association between disulfidptosis and EMs based on bioinformatics analysis, machine learning, and animal experiments. Further investigation on the biological mechanism of disulfidptosis in EMs is anticipated to yield novel advancements for searching for potential diagnostic biomarkers and revolutionary therapeutic approaches in EMs.

Use of Landsat Imagery Time-Series and Random Forests Classifier to Reconstruct Eelgrass Bed Distribution Maps in Eeyou Istchee

The eastern coastline of James Bay is known to have been home to sizeable eelgrass beds (Zostera marina L.) which thrived in the bay’s shallow, subarctic waters. The region was subjected to substantial hydroelectric dams, large fires, and other human activities in the past half-century. To assess the impact of these factors on eelgrass beds, a historical reconstruction of eelgrass bed distribution was performed from images acquired by Landsat-5 Thematic Mapper (TM) in 1988, 1991, and 1996 and images of the Landsat-8 Operational Land Imager (OLI) in 2019. All the images were classified using the Random Forests classifier (RF) and assessed for accuracy each year on a bay-wide scale using an independent field validation dataset. The validation data were extracted from an eelgrass bed map established using aerial photos and field surveys in 1986, 1991, and 1995 and from a field survey in 2019. The overall validation accuracy of the classified images (between 72% and 85%) showed good agreement with the other datasets for most locations, providing reassurance about the reliability of the research. This makes it possible to use satellite imagery to detect past changes to eelgrass distribution within a bay. The classified images of 1988 and 1996 were also compared to aerial photos taken in years close to each other at ten sites to determine their ability to assess small eelgrass beds’ shape and presence. Such a comparison revealed that the classified images accurately portrayed eelgrass distribution even at finer scales.

Integration of machine learning and remote sensing for above ground biomass estimation through Landsat-9 and field data in temperate forests of the Himalayan region

Accurately estimating aboveground biomass (AGB) in forest ecosystems facilitates efficient resource management, carbon accounting, and conservation efforts. This study examines the relationship between predictors from Landsat-9 remote sensing data and several topographical features. While Landsat-9 provides reliable data crucial for long-term monitoring, it is part of a broader suite of available remote sensing technologies. We employ machine learning algorithms such as Extreme Gradient Boosting (XGBoost), Support Vector Regression (SVR), and Random Forest (RF), alongside linear regression techniques like Multiple Linear Regression (MLR). The primary objectives of this study encompass two key aspects. Firstly, the research methodically selects optimal predictor combinations from four distinct variable groups: Landsat-9 (L1) data, a fusion of Landsat-9 data and Vegetation-based indices (L2), and the integration of Landsat-9 data with the Shuttle Radar Topography Mission Digital Elevation Model (SRTM DEM) (L3) and the combination of best predictors (L4) derived from L1, L2, and L3. Secondly, the research systematically assesses the effectiveness of different algorithms to identify the most precise method for establishing any potential relationship between field-measured AGB and predictor variables. Our study revealed that the Random Forest (RF) model was the most efficient method utilizing Landsat-9 OLI and SRTM DEM (L3) predictors, achieving remarkable accuracy. This conclusion was reached by assessing its outstanding performance when compared to an independent validation dataset. The RF model exhibited remarkable accuracy, presenting relative mean absolute error (RMAE), relative root mean square error (RRMSE), and R2 values of 14.33%, 22.23%, and 0.81, respectively. The XGBoost model is the subsequent choice with RMAE, RRMSE, and R2 values of 15.54%, 23.85%, and 0.77, respectively. The study further highlights the significance of specific spectral bands, notably B4 and B5 from Landsat 9 OLI data, in capturing spatial AGB distribution patterns. Integration of vegetation-based indices, including TNDVI, NDVI, RVI, and GNDVI, further refines AGB mapping precision. Elevation, slope, and the Topographic Wetness Index (TWI) are crucial proxies for representing biophysical and biological mechanisms impacting AGB. Through the utilization of openly accessible fine-resolution data and employing the RF algorithm, the research demonstrated promising outcomes in the identification of optimal predictor-algorithm combinations for forest AGB mapping. This comprehensive approach offers a valuable avenue for informed decision-making in forest management, carbon assessment, and ecological monitoring initiatives.

CACM-Net: Daytime Cloud Mask for AGRI Onboard the FY-4A Satellite

Accurate cloud detection is a crucial initial stage in optical satellite remote sensing. In this study, a daytime cloud mask model is proposed for the Advanced Geostationary Radiation Imager (AGRI) onboard the Fengyun 4A (FY-4A) satellite based on a deep learning approach. The model, named “Convolutional and Attention-based Cloud Mask Net (CACM-Net)”, was trained using the 2021 dataset with CALIPSO data as the truth value. Two CACM-Net models were trained based on a satellite zenith angle (SZA) < 70° and >70°, respectively. The study evaluated the National Satellite Meteorological Center (NSMC) cloud mask product and compared it with the method established in this paper. The results indicate that CACM-Net outperforms the NSMC cloud mask product overall. Specifically, in the SZA < 70° subset, CACM-Net enhances accuracy, precision, and F1 score by 4.8%, 7.3%, and 3.6%, respectively, while reducing the false alarm rate (FAR) by approximately 7.3%. In the SZA > 70° section, improvements of 12.2%, 19.5%, and 8% in accuracy, precision, and F1 score, respectively, were observed, with a 19.5% reduction in FAR compared to NSMC. An independent validation dataset for January–June 2023 further validates the performance of CACM-Net. The results show improvements of 3.5%, 2.2%, and 2.8% in accuracy, precision, and F1 scores for SZA < 70° and 7.8%, 11.3%, and 4.8% for SZA > 70°, respectively, along with reductions in FAR. Cross-comparison with other satellite cloud mask products reveals high levels of agreement, with 88.6% and 86.3% matching results with the MODIS and Himawari-9 products, respectively. These results confirm the reliability of the CACM-Net cloud mask model, which can produce stable and high-quality FY-4A AGRI cloud mask results.

Prompt engineering on leveraging large language models in generating response to InBasket messages.

Large Language Models (LLMs) have been proposed as a solution to address high volumes of Patient Medical Advice Requests (PMARs). This study addresses whether LLMs can generate high quality draft responses to PMARs that satisfies both patients and clinicians with prompt engineering. We designed a novel human-involved iterative processes to train and validate prompts to LLM in creating appropriate responses to PMARs. GPT-4 was used to generate response to the messages. We updated the prompts, and evaluated both clinician and patient acceptance of LLM-generated draft responses at each iteration, and tested the optimized prompt on independent validation data sets. The optimized prompt was implemented in the electronic health record production environment and tested by 69 primary care clinicians. After 3 iterations of prompt engineering, physician acceptance of draft suitability increased from 62% to 84% (P <.001) in the validation dataset (N = 200), and 74% of drafts in the test dataset were rated as "helpful." Patients also noted significantly increased favorability of message tone (78%) and overall quality (80%) for the optimized prompt compared to the original prompt in the training dataset, patients were unable to differentiate human and LLM-generated draft PMAR responses for 76% of the messages, in contrast to the earlier preference for human-generated responses. Majority (72%) of clinicians believed it can reduce cognitive load in dealing with InBasket messages. Informed by clinician and patient feedback synergistically, tuning in LLM prompt alone can be effective in creating clinically relevant and useful draft responses to PMARs.

Intestinal mRNA expression profiles associated with mucosal healing in ustekinumab-treated Crohn's disease patients: bioinformatics analysis and prospective cohort validation

BackgroundVariations exist in the response of patients with Crohn’s disease (CD) to ustekinumab (UST) treatment, but the underlying cause remains unknown. Our objective was to investigate the involvement of immune cells and identify potential biomarkers that could predict the response to interleukin (IL) 12/23 inhibitors in patients with CD.MethodsThe GSE207022 dataset, which consisted of 54 non-responders and 9 responders to UST in a CD cohort, was analyzed. Differentially expressed genes (DEGs) were identified and subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. Least absolute shrinkage and selection operator (LASSO) regression was used to screen the most powerful hub genes. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the predictive performances of these genes. Single-sample Gene Set Enrichment Analysis (ssGSEA) was used to estimate the proportions of immune cell types. These significantly altered genes were subjected to cluster analysis into immune cell-related infiltration. To validate the reliability of the candidates, patients prescribed UST as a first-line biologic in a prospective cohort were included as an independent validation dataset.ResultsA total of 99 DEGs were identified in the integrated dataset. GO and KEGG analyses revealed significant enrichment of immune response pathways in patients with CD. Thirteen genes (SOCS3, CD55, KDM5D, IGFBP5, LCN2, SLC15A1, XPNPEP2, HLA-DQA2, HMGCS2, DDX3Y, ITGB2, CDKN2B and HLA-DQA1), which were primarily associated with the response versus nonresponse patients, were identified and included in the LASSO analysis. These genes accurately predicted treatment response, with an area under the curve (AUC) of 0.938. T helper cell type 1 (Th1) cell polarization was comparatively strong in nonresponse individuals. Positive connections were observed between Th1 cells and the LCN2 and KDM5D genes. Furthermore, we employed an independent validation dataset and early experimental verification to validate the LCN2 and KDM5D genes as effective predictive markers.ConclusionsTh1 cell polarization is an important cause of nonresponse to UST therapy in patients with CD. LCN2 and KDM5D can be used as predictive markers to effectively identify nonresponse patients.Trial registration: Trial registration number: NCT05542459; Date of registration: 2022-09-14; URL: https://www.clinicaltrials.gov.

Use of Machine Learning to Predict Onset of NAFLD in an All-Comers Cohort—Development and Validation in 2 Large Asian Cohorts

Background and AimsNon-alcoholic fatty liver disease (NAFLD) is one of the most common liver diseases. There are no universally accepted models that accurately predict time to onset of NAFLD. Machine learning (ML) models may allow prediction of such time-to-event (i.e. survival) outcomes. This study aims to develop and independently validate ML derived models to allow personalised prediction of time to onset of NAFLD in individuals who have no NAFLD at baseline. MethodsThe development dataset comprised 25599 individuals from a South Korean NAFLD registry. A random 70:30 split divided it into training and internal validation sets. ML survival models (Random Survival Forest (RSF), Extra Survival Trees (XST)) were fitted, with time to NAFLD diagnosis in months as the target variable and routine anthropometric and laboratory parameters as predictors. The independent validation dataset comprised 16173 individuals from a Chinese open dataset. Models were evaluated using the concordance index (c-index) and Brier score on both the internal and independent validation sets. ResultsThe datasets (development vs independent validation) had 1331107 vs 543874 person months of follow up, NAFLD incidence of 25.7% (6584 individuals) vs 14.4% (2322 individuals), and median time to NAFLD onset of 60 (IQR 38-75) vs 24 (IQR 13-37) months, respectively. The ML models achieved a good c-index of >0.7 in the validation cohort - RSF 0.751 (95% CI 0.742-0.759), XST 0.752 (95% CI 0.744-0.762). ConclusionML models can predict time-to-onset of NAFLD based on routine patient data. They can be used by clinicians to deliver personalised predictions to patients, which may facilitate patient counselling and clinical decision making on interval imaging timing.

Dynamic Evolution of Infarct Volumes at MRI in Ischemic Stroke Due to Large Vessel Occlusion.

The typical infarct volume trajectories in stroke patients, categorized as slow or fast progressors, remain largely unknown. This study aimed to reveal the characteristic spatiotemporal evolutions of infarct volumes caused by large vessel occlusion (LVO) and show that such growth charts help anticipate clinical outcomes. We conducted a secondary analysis from prospectively collected databases (FRAME, 2017-2019; ETIS, 2015-2022). We selected acute MRI data from anterior LVO stroke patients with witnessed onset, which were divided into training and independent validation datasets. In the training dataset, using Gaussian mixture analysis, we classified the patients into 3 growth groups based on their rate of infarct growth (diffusion volume/time-to-imaging). Subsequently, we extrapolated pseudo-longitudinal models of infarct growth for each group and generated sequential frequency maps to highlight the spatial distribution of infarct growth. We used these charts to attribute a growth group to the independent patients from the validation dataset. We compared their 3-month modified Rankin scale (mRS) with the predicted values based on a multivariable regression model from the training dataset that used growth group as an independent variable. We included 804 patients (median age 73.0 years [interquartile range 61.2-82.0 years]; 409 men). The training dataset revealed nonsupervised clustering into 11% (74/703) slow, 62% (437/703) intermediate, and 27% (192/703) fast progressors. Infarct volume evolutions were best fitted with a linear (r = 0.809; p < 0.001), cubic (r = 0.471; p < 0.001), and power (r = 0.63; p < 0.001) function for the slow, intermediate, and fast progressors, respectively. Notably, the deep nuclei and insular cortex were rapidly affected in the intermediate and fast groups with further cortical involvement in the fast group. The variable growth group significantly predicted the 3-month mRS (multivariate odds ratio 0.51; 95% CI 0.37-0.72, p < 0.0001) in the training dataset, yielding a mean area under the receiver operating characteristic curve of 0.78 (95% CI 0.66-0.88) in the independent validation dataset. We revealed spatiotemporal archetype dynamic evolutions following LVO stroke according to 3 growth phenotypes called slow, intermediate, and fast progressors, providing insight into anticipating clinical outcome. We expect this could help in designing neuroprotective trials aiming at modulating infarct growth before EVT.

Spectroscopy-based chemometrics combined machine learning modeling predicts cashew foliar macro- and micronutrients

Precision nutrient management in orchard crops needs precise, accurate, and real-time information on the plant’s nutritional status. This is limited by the fact that it requires extensive leaf sampling and chemical analysis when it is to be done over more extensive areas like field- or landscape scale. Thus, rapid, reliable, and repeatable means of nutrient estimations are needed. In this context, lab-based remote sensing or spectroscopy has been explored in the current study to predict the foliar nutritional status of the cashew crop. Novel spectral indices (normalized difference and simple ratio), chemometric modeling, and partial least square regression (PLSR) combined machine learning modeling of the visible near-infrared hyperspectral data were employed to predict macro- and micronutrients content of the cashew leaves. The full dataset was divided into calibration (70 % of the full dataset) and validation (30 % of the full dataset) datasets. An independent validation dataset was used for the validation of the algorithms tested. The approach of spectral indices yielded very poor and unreliable predictions for all eleven nutrients. Among the chemometric models tested, the performance of the PLSR was the best, but still, the predictions were not acceptable. The PLSR combined machine learning modeling approach yielded acceptable to excellent predictions for all the nutrients except sulphur and copper. The best predictions were observed when PLSR was combined with Cubist for nitrogen, phosphorus, potassium, manganese, and zinc; support vector machine regression for calcium, magnesium, iron, copper, and boron; elastic net for sulphur. The current study showed hyperspectral remote sensing-based models could be employed for non-destructive and rapid estimation of cashew leaf macro- and micro-nutrients. The developed approach is suggested to employ within the operational workflows for site-specific and precision nutrient management of the cashew orchards.

Discriminative analysis of schizophrenia and major depressive disorder using fNIRS

BackgroundResearch into the shared and distinct brain dysfunctions in patients with schizophrenia (SCZ) and major depressive disorder (MDD) has been increasing. However, few studies have explored the application of functional near-infrared spectroscopy (fNIRS) in investigating brain dysfunction and enhancing diagnostic methodologies in these two conditions. MethodsA general linear model was used for analysis of brain activation following task-state fNIRS from 131 patients with SCZ, 132 patients with MDD and 130 healthy controls (HCs). Subsequently, seventy-seven time-frequency analysis methods were used to construct new features of fNIRS, followed by the implementation of five machine learning algorithms to develop a differential diagnosis model for the three groups. This model was evaluated by comparing it to both a diagnostic model relying on traditional fNIRS features and assessments made by two psychiatrists. ResultsBrain activation analysis revealed significantly lower activation in Broca's area, the dorsolateral prefrontal cortex, and the middle temporal gyrus for both the SCZ and MDD groups compared to HCs. Additionally, the SCZ group exhibited notably lower activation in the superior temporal gyrus and the subcentral gyrus compared to the MDD group. When distinguishing among the three groups using independent validation datasets, the models utilizing new fNIRS features achieved an accuracy of 85.90 % (AUC = 0.95). In contrast, models based on traditional fNIRS features reached an accuracy of 52.56 % (AUC = 0.66). The accuracies of the two psychiatrists were 42.00 % (AUC = 0.60) and 38.00 % (AUC = 0.50), respectively. ConclusionThis investigation brings to light the shared and distinct neurobiological abnormalities present in SCZ and MDD, offering potential enhancements for extant diagnostic systems.

Integrative analyses of multi-omics data constructing tumor microenvironment and immune-related molecular prognosis model in human colorectal cancer

The increasing prevalence and incidence of colorectal cancer (CRC), particularly in young adults, underscore the imperative to comprehend its fundamental mechanisms, discover novel diagnostic and prognostic markers, and enhance therapeutic strategies. Here, we integrated multi-omics data, including gene expression, somatic mutation data and DNA methylation data, to unravel the intricacies of tumor microenvironment (TME) in CRC and search for novel prognostic markers. By calculating the immune score for each patient from the expression profile, we delineated the differential immune cell fraction, constructed an immune-related multi-omics atlas, and identified molecular characteristics. The entire colorectal dataset (n = 343) was randomly divided into training (n = 249) and testing datasets (n = 94). We screened 144 immune-related genes, 6 mutant genes, and 38 methylation probes associated with overall survival (OS). These makers were then incorporated into a 10-gene prognostic model using Lasso and Cox regression in the training dataset, and the model's performance was evaluated in an independent validation dataset. The model exhibited satisfactory results (average concordance index [C-index] = 0.77), with the average 1-year, 3-year, and 5-year AUCs being 0.79, 0.76, and 0.76 in the training dataset and 0.74, 0.80, and 0.90 in the testing dataset. Furthermore, the prognostic model demonstrated applicability in guiding chemotherapy for CRC patients and exhibited a degree of pan-cancer utility in risk stratification. In conclusion, our integrated analysis of multi-omics data revealed immune-related genetic and epigenetic characteristics of the TME. We propose an integrative prognostic model that can stratify risk and guide chemotherapy for CRC patients. The generalizability of the model in risk stratification across different cancer types was validated in Pan-Cancer cohort.

Use of an AI-powered visualization platform to support precision clinical decision-making in patients with breast cancer undergoing neoadjuvant chemotherapy.

e12618 Background: Chemotherapy to reduce cancer burden and tumor size ("downstaging") is widely used in neoadjuvant treatment of early-stage breast cancer (ESBC) and can have a dramatic impact on selection of further treatment courses, including surgery. The resulting decision-making process between physicians and cancer patients requires detailed and effective communication of the patient's disease status, and clear presentation of this information in Tumor Board settings. Artificial intelligence (AI) -powered visualization platforms offer the potential to aid in evaluating breast tumor location and morphology, and accordingly optimize treatment. Additionally, these tools offer enhanced communication with patients regarding future treatment options, particularly serving as clinical decision support (CDS) tools. Methods: We assessed the performance of an AI platform that employs dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) to render three-dimensional (3D) visualizations of breast tumors and 5 additional chest tissues on an independent validation dataset comprise of 100 ESBC cases with ground-truth tumor labels vetted by 2 breast-specialized radiologists. The platform provides 3D depictions of the tumor and its multi-tissue context, as well as functionalities for quantifying tumor morphology, tumor-to-landmark structure distances, excision volumes, and approximate surgical margins. We subsequently evaluated the tool's utility in a real-world setting to understand how the tool impacts clinical decision-making and patient communication, specifically around surgical planning in the context of neoadjuvant chemotherapy (NACT). Results: The platform generated accurate 3D tumor maps, as evidenced by a median Dice score of 75.6% and a median Hausdorff distances of 15.1 mm relative to the radiologist validation set. The automatic longest tumor dimension calculations were linearly correlated with radiologist direct measurements (Pearson correlation coefficient = 0.78) and had a median absolute difference of 6.3 mm. We found strong overlap between the platform’s ground-truth convex hulls, as shown by a median Dice score of 81.5%. When used by a skilled breast-specialized surgeon, the tool enabled rapid and intuitive patient-physician communication of the location and extent of disease, and potential degrees of downstaging to breast-conserving surgery. Conclusions: Detailed 3D tumor and surrounding multi-tissue depictions offer both qualitative and quantitative comprehension of cancer topology and may aid in formulating an optimal treatment approach for breast cancer treatment. We show that detailed depictions enhance communication between physicians and patients and may aid in streamlining decision-making.

Detection of forest disturbance across California using deep-learning on PlanetScope imagery

California forests have recently experienced record breaking wildfires and tree mortality from droughts, However, there is inadequate monitoring, and limited data to inform policies and management strategies across the state. Although forest surveys and satellite observations of forest cover changes exist at medium to coarse resolutions (30–500 m) annually, they remain less effective in mapping small disturbances of forest patches (&amp;lt;5 m) occurring multiple times a year. We introduce a novel method of tracking California forest cover using a supervised U-Net deep learning architecture and PlanetScope’s Visual dataset which provides 3-band RGB (Red, Green, and Blue) mosaicked imagery. We created labels of forest and non-forest to train the U-Net model to map tree cover based on a semi-unsupervised classification method. We then detected changes of tree cover and disturbance with the U-Net model, achieving an overall accuracy of 98.97% over training data set, and 95.5% over an independent validation dataset, obtaining a precision of 82%, and a recall of 74%. With the predicted tree cover mask, we created wall to wall monthly tree cover maps over California at 4.77 m resolution for 2020, 2021, and 2022. These maps were then aggregated in a post-processing step to develop annual maps of disturbance, while accounting for the time of disturbance and other confounding factors such as topography, phenological and snow cover variability. We compared our high-resolution disturbance maps with wildfire GIS survey data from CALFIRE, and satellite-based forest cover changes and achieved an F-1 score of 54% and 88% respectively. The results suggest that high-resolution maps capture variability of forest disturbance and fire that wildfire surveys and medium resolution satellite products cannot. From 2020 to 2021, California maintained 30,923.5 sq km of forest while 5,994.9 sq km were disturbed. The highest observed forest loss rate was located at the Sierra Nevada mountains at 21.4% of the forested area being disturbed between 2020 and 2021. Our findings highlight the strong potential of deep learning and high-resolution RGB optical imagery for mapping complex forest ecosystems and their changes across California, as well as the application of these techniques on a national to global scale.

WilsonGenAI a deep learning approach to classify pathogenic variants in Wilson Disease.

Advances in Next Generation Sequencing have made rapid variant discovery and detection widely accessible. To facilitate a better understanding of the nature of these variants, American College of Medical Genetics and Genomics and the Association of Molecular Pathologists (ACMG-AMP) have issued a set of guidelines for variant classification. However, given the vast number of variants associated with any disorder, it is impossible to manually apply these guidelines to all known variants. Machine learning methodologies offer a rapid way to classify large numbers of variants, as well as variants of uncertain significance as either pathogenic or benign. Here we classify ATP7B genetic variants by employing ML and AI algorithms trained on our well-annotated WilsonGen dataset. We have trained and validated two algorithms: TabNet and XGBoost on a high-confidence dataset of manually annotated, ACMG & AMP classified variants of the ATP7B gene associated with Wilson's Disease. Using an independent validation dataset of ACMG & AMP classified variants, as well as a patient set of functionally validated variants, we showed how both algorithms perform and can be used to classify large numbers of variants in clinical as well as research settings. We have created a ready to deploy tool, that can classify variants linked with Wilson's disease as pathogenic or benign, which can be utilized by both clinicians and researchers to better understand the disease through the nature of genetic variants associated with it.

Towards good modelling practice for parallel hybrid models for wastewater treatment processes.

This study explores various approaches to formulating a parallel hybrid model (HM) for Water and Resource Recovery Facilities (WRRFs) merging a mechanistic and a data-driven model. In the study, the HM is constructed by training a neural network (NN) on the residual of the mechanistic model for effluent nitrate. In an initial experiment using the Benchmark Simulation Model no. 1, a parallel HM effectively addressed limitations in the mechanistic model's representation of autotrophic bacteria growth and the data-driven model's incapability to extrapolate. Next, different versions of a parallel HM of a large pilot-scale WRRF are constructed, using different calibration/training datasets and different versions of the mechanistic model to investigate the balance between the calibration effort for the mechanistic model and the compensation by the NN component. The HM can improve predictions compared to the mechanistic model. Training the NN on an independent validation dataset produced better results than on the calibration dataset. Interestingly, the best performance is achieved for the HM based on a mechanistic model using default (uncalibrated) parameters. Both long short-term memory (LSTM) and convolutional neural network (CNN) are tested as data-driven components, with a CNN HM (root-mean-squared error (RMSE) = 1.58 mg NO3-N/L) outperforming an LSTM HM (RMSE = 4.17 mg NO3-N/L).

Reliable detection of stochastic epigenetic mutations and associations with cardiovascular aging.

Stochastic epigenetic mutations (SEMs) have been proposed as novel aging biomarkers to capture heterogeneity in age-related DNA methylation changes. SEMs are defined as outlier methylation patterns at cytosine-guanine dinucleotide sites, categorized as hypermethylated (hyperSEM) or hypomethylated (hypoSEM) relative to a reference. Because SEMs are defined by their outlier status, it is critical to differentiate extreme values due to technical noise or data artifacts from those due to real biology. Using technical replicate data, we found SEM detection is not reliable: across 3 datasets, 24 to 39% of hypoSEM and 46 to 67% of hyperSEM are not shared between replicates. We identified factors influencing SEM reliability-including blood cell type composition, probe beta-value statistics, genomic location, and presence of SNPs. We used these factors in a training dataset to build a machine learning-based filter that removes unreliable SEMs, and found this filter enhances reliability in two independent validation datasets. We assessed associations between SEM loads and aging phenotypes in the Framingham Heart Study and discovered that associations with aging outcomes were in large part driven by hypoSEMs at baseline methylated probes and hyperSEMs at baseline unmethylated probes, which are the same subsets that demonstrate highest technical reliability. These aging associations were preserved after filtering out unreliable SEMs and were enhanced after adjusting for blood cell composition. Finally, we utilized these insights to formulate best practices for SEM detection and introduce a novel R package, SEMdetectR, which uses parallel programming for efficient SEM detection with comprehensive options for detection, filtering, and analysis.

Abstract PO5-15-10: Plasticity marker FOXC1 expression accurately predicts efficacy of Adjuvant Tamoxifen + Chemotherapy in reducing all-cause mortality in ER+LN- Breast Cancer: Validation in the SCAN-B Prospective Study (NCT02306096)

Abstract INTRODUCTION: Despite the development and validation of multimarker gene panels like OncotypeDx®, Mammaprint® and EndoPredict®, an affordable and globally accessible biomarker approach that can predict increased risk of all-cause mortality in patients diagnosed with ER+LN- breast cancer, as well as the efficacy of adjuvant endocrine + chemotherapy in mitigating such elevated risk, continues to be an unmet medical need. Lack of such a pragmatic solution continues to be the cause of preventable recurrent/metastatic disease in patients diagnosed with ER+LN- breast cancer in resource-challenged settings around the world. We hypothesized that a predictive biomarker strategy that measures expression of the plasticity marker FOXC1, in combination with clinical parameters like tumor size (TS) and tumor grade (TG), may offer equivalent results to that of the above multimarker gene panels at a significantly lower economic cost without compromising accuracy of prediction results. METHODS: Pre-treatment tumor RNA data obtained from a training cohort (compendium of gene expression microarray datasets, n=2857) and a single large validation cohort (SCAN-B Prospective Multicenter Observational Study, n=3520) for patients diagnosed with ER+LN- breast cancer were analyzed for FOXC1 expression, tumor size (TS) and tumor grade (TG) and correlated with Recurrence-Free Survival (RFS) and Overall Survival (OS). Optimized biomarker cutoff values based on model area-under-curve were leave-one-out cross validated and Risk-of-Recurrence (ROR) prediction algorithm derived utilizing the (compendium) training dataset. The unmodified strategy was then validated in the independent (SCAN-B) validation dataset. RESULTS: A predetermined High ROR score calculated using FOXC1 expression, TS and TG (trained using the compendium dataset) predicted efficacy of adjuvant endocrine + chemotherapy over that of adjuvant endocrine therapy alone in ER+LN- patients in terms of statistically significant reduction in all-cause mortality at 8-years post-diagnosis in the SCAN-B validation dataset (4.66% vs 24.06%, n=326, OR=6.48, 95%CI [2.97-14.11], p&amp;lt; 0.0001). OncotypeDx® (5.7% vs 14.4%, n=762, OR=2.79, 95%CI [1.37-5.67], p=0.002), Mammaprint® (5.1% vs 18.4%, n=665, OR= 95%CI [2.04-8.43], p&amp;lt; 0.0001) and Endopredict® (4.8% vs 15.2%, n=923, OR=3.58 95%CI [1.78-7.21], p=0.0002) were also statistically significant predictors of the same. CONCLUSION: Pre-treatment tumor FOXC1 mRNA or protein expression (assessed using qRT-PCR or routine immunohistochemistry (IHC), respectively, when combined with TS and TG presents a unique and economical alternative solution to multimarker gene panel tests like OncotypeDx®, Mammaprint® or Endopredict®, for guiding therapy of patients diagnosed with ER+LN- breast cancer in resource challenged settings. Such an approach to identify elevated risk of recurrence in patients diagnosed with ER+LN- breast cancer and prevent the same by guiding adjuvant endocrine + chemotherapy decisions, could help to extend recurrence-free and overall survival. Such an approach merits testing in real world ER+LN- patient cohorts in resource-challenged settings to help support implementation of this FOXC1-driven predictive biomarker strategy in the clinic. Citation Format: Partha Ray, Tania Ray, Clive Taylor, Robert Hussa. Plasticity marker FOXC1 expression accurately predicts efficacy of Adjuvant Tamoxifen + Chemotherapy in reducing all-cause mortality in ER+LN- Breast Cancer: Validation in the SCAN-B Prospective Study (NCT02306096) [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO5-15-10.

National variability in soil organic carbon stock predictions: Impact of bulk density pedotransfer functions

Accurate soil organic carbon storage (SOCS) estimation is crucial for sustaining ecosystem health and mitigating climate change impacts. This study investigated the accuracy and variability of SOCS predictions, focusing on the role of pedotransfer functions (PTFs) in estimating soil bulk density (BD). Utilizing a comprehensive dataset from the Korean Rural Development Administration (RDA database), which includes 516 soil horizons, we evaluated 36 widely-used BD PTFs, well-established formulas that estimate BD by considering soil properties, including soil organic carbon (SOC), soil organic matter (OM), sand, gravel, silt, and clay. These PTFs demonstrated varying levels of precision, with root mean squared errors (RMSE) ranging from 0.177 to 0.377 Mg m−3 and coefficients of determination (R2) from 0.176 to 0.658; hence, the PTFs have been classified into excellent, moderate, and poor-performing groups for predicting BD. Further, a novel PTF based on an exponential function of SOC was developed, showing superior predictive power (R2 = 0.73) compared to existing PTFs, using an independent validation dataset. Our findings reveal significant differences in SOCS predictions and observations among the PTFs, with a p-value <0.05. The highest concentrations of SOCS were noted in forest soils, considerably above the national average, highlighting the importance of tailored soil management practices to enhance carbon sequestration. These findings are crucial for refining PTF precision to improve the accuracy of national SOCS estimates, supporting effective land management and climate change mitigation strategies.

Artificial intelligence for diagnosis and prognosis prediction of natural killer/T cell lymphoma using magnetic resonance imaging

Accurate diagnosis and prognosis prediction are conducive to early intervention and improvement of medical care for natural killer/T cell lymphoma (NKTCL). Artificial intelligence (AI)-based systems are developed based on nasopharynx magnetic resonance imaging. The diagnostic systems achieve areas under the curve of 0.905-0.960 in detecting malignant nasopharyngeal lesions and distinguishing NKTCL from nasopharyngeal carcinoma in independent validation datasets. In comparison to human radiologists, the diagnostic systems show higher accuracies than resident radiologists and comparable ones to senior radiologists. The prognostic system shows promising performance in predicting survival outcomes of NKTCL and outperforms several clinical models. For patients with early-stage NKTCL, only the high-risk group benefits from early radiotherapy (hazard ratio= 0.414 vs. late radiotherapy; 95% confidence interval, 0.190-0.900, p= 0.022), while progression-free survival does not differ in the low-risk group. In conclusion, AI-based systems show potential in assisting accurate diagnosis and prognosis prediction and may contribute to therapeutic optimization for NKTCL.