Training Set Research Articles

A clinical-radiomics nomogram based on automated segmentation of chest CT to discriminate PRISm and COPD patients

PurposeIt is vital to develop noninvasive approaches with high accuracy to discriminate the preserved ratio impaired spirometry (PRISm) group from the chronic obstructive pulmonary disease (COPD) groups. Radiomics has emerged as an image analysis technique. This study aims to develop and confirm the new radiomics-based noninvasive approach to discriminate these two groups. MethodsTotally 1066 subjects from 4 centers were included in this retrospective research, and classified into training, internal validation or external validation sets. The chest computed tomography (CT) images were segmented by the fully automated deep learning segmentation algorithm (Unet231) for radiomics feature extraction. We established the radiomics signature (Rad-score) using the least absolute shrinkage and selection operator algorithm, then conducted ten-fold cross-validation using the training set. Last, we constructed a radiomics signature by incorporating independent risk factors using the multivariate logistic regression model. Model performance was evaluated by receiver operating characteristic (ROC) curve, calibration curve, and decision curve analyses (DCA). ResultsThe Rad-score, including 15 radiomic features in whole-lung region, which was suitable for diffuse lung diseases, was demonstrated to be effective for discriminating between PRISm and COPD. Its diagnostic accuracy was improved through integrating Rad-score with a clinical model, and the area under the ROC (AUC) were 0.82(95 %CI 0.79–0.86), 0.77(95 %CI 0.72–0.83) and 0.841(95 %CI 0.78–0.91) for training, internal validation and external validation sets, respectively. As revealed by analysis, radiomics nomogram showed good fit and superior clinical utility. ConclusionsThe present work constructed the new radiomics-based nomogram and verified its reliability for discriminating between PRISm and COPD.

Prediction of YY1 loop anchor based on multi-omics features

The three-dimensional structure of chromatin is crucial for the regulation of gene expression. YY1 promotes enhancer-promoter interactions in a manner analogous to CTCF-mediated chromatin interactions. However, little is known about which YY1 binding sites can form loop anchors. In this study, the LightGBM model was used to predict YY1-loop anchors by integrating multi-omics data. Due to the large imbalance in the number of positive and negative samples, we use AUPRC to reflect the quality of the classifier. The results show that the LightGBM model exhibits strong predictive performance (AUPRC≥0.93). To verify the robustness of the model, the dataset was divided into training and testing sets at a 4:1 ratio. The results show that the model performs well for YY1-loop anchor prediction on both the training and independent testing sets. Additionally, we ranked the importance of the features and found that the formation of YY1-loop anchors is primarily influenced by the co-binding of transcription factors CTCF, SMC3, and RAD21, as well as histone modifications and sequence context.

Automated diagnosis of brain tumor classification and segmentation of magnetic resonance imaging images

<span lang="EN-US">Brain tumors are one of the most prevalent disorders of the central nervous system and are dangerous. For patients to receive the best treatment, early diagnosis is crucial. For radiologists to correctly detect brain tumor images, an automated approach is required. The identification procedure can be time-consuming and prone to mistakes. In this work, the issue of fully automated brain tumor classification and segmentation of magnetic resonance imaging (MRI) including meningioma, glioma, pituitary, and no tumor is taken into consideration. In this study, convolutional neural network (CNN) and mask region-based convolutional neural network (R-CNN) are proposed for classification and segmentation problems respectively. This study employed 3,200 images as a training set and the system achieved an accuracy of 96% for classifying the tumors and 94% accuracy in segmentation of tumors.</span>

Explainable machine learning models for predicting the acute toxicity of pesticides to sheepshead minnow (Cyprinodon variegatus).

A quantitative structure-activity relationship (QSAR) study was conducted on 313 pesticides to predict their acute toxicity to Sheepshead minnow (Cyprinodon variegatus) by using DRAGON descriptors. Essentials accounting for a reliable model were all considered carefully, giving full consideration to the OECD (Organization for Economic Co-operation and Development) principles for QSAR acceptability in regulation during the model construction and assessment process. Nine variables were selected through the forward stepwise regression method and used as inputs to construct both linear and nonlinear models. The obtained models were validated internally and externally. Generally, machine learning-based methods, namely support vector machine (SVM), random forest (RF), and projection pursuit regression (PPR), perform better than the multiple linear regression (MLR) model. The statistical results (R2 = 0.682-0.933, Q2LOO = 0.604-0.659, Q2F1 = 0.740-0.796, CCC = 0.861-0.882) of the developed models show that they are robust, reliable, reproducible, accurate and predictive. Comparatively, the RF model performs best, giving predictive correlation coefficient Q2 of 0.814, root mean squared error (RMSE) of 0.658 and mean absolute error (MAE) of 0.534 for the test set, respectively. The RF model (as well as SVM and PPR models) was visualized and explained by using the SHapley Additive explanation (SHAP) analysis to enhance its transparency and credibility. In addition, the applicability domain (AD) range of the RF model was characterized by the Williams plot and the tree manifold approximation and projection (TMAP) technology was utilized to illustrate similarity and diversity of the entire data space, to assist in the analysis of the outliers. Activity cliff detection was investigated by using Arithmetic Residuals in K-groups Analysis (ARKA) descriptors. It was found that none of the pesticides was identified as an activity cliff in the training set or a potential prediction cliff in the test set. Therefore, the RF model fulfills each OECD principle in regulation for QSAR models. The research in this work will aid in the in silico QSAR prediction of the acute toxicity to Sheepshead minnow (Cyprinodon variegatus) for untested and new toxic pesticides and can also be extended to other studies.

Fishy Forensics: FT-NIR and Machine Learning based authentication of Mediterranean Anchovies (Engraulis encrasicolus)

Seafood authentication and traceability is a challenging issue owing to its complex supply chain and fishing in international waters. NIR spectroscopy has been successfully used to authenticate food of animal and plant origin. In this study, FT-NIR was used to discriminate between Mediterranean anchovies (Engraulis encrasicolus) fished from Adriatic, Balearic, and Tyrrhenian Sea. The spectra were prepared using the standard normal variate (SNV) and the Savitzky-Golay 1st derivative, 2nd order (SG), as well as both together. The model was built, after outlier removal, with linear-support vector machine (L-SVM), polynomial-SVM (P-SVM), k-nearest neighbor (k-NN) and Random Forest (RF). Spectral preprocessing improved model classification accuracy for all algorithms. The data could not be put into groups by linear algorithms like L-SVM and k-NN because the NIR spectra were not linear and had many columns. Non-linear algorithms, P-SVM and RF when coupled with SG+SNV, successfully produced models with maximum robustness. P-SVM and RF models had 100% accuracy in training set and 95.7% and 95.5% accuracy in testing set, respectively and 95.2 and 95.1 accuracy in cross-validation set.

A nomogram for predicting nutritional risk before gastric cancer surgery.

Gastric cancer (GC) is the fourth leading cause of cancer death worldwide. Patients with GC have higher nutritional risk. This study aimed to construct a nomogram model for predicting preoperative nutritional risk in patients with GC in order to assess preoperative nutritional risk in patients more precisely. Patients diagnosed with GC and undergoing surgical treatment were included in this study. Data was collected through clinical information, laboratory testing, and radiomics-derived characteristics. Least absolute shrinkage selection operator (LASSO) regression analysis and multi-variable logistic regression were employed to construct a clinical prediction model, which takes the form of a logistic nomogram. The effectiveness of the nomogram model was evaluated using receiver operat-ing characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA). A total of three predictors, namely body mass index (BMI), hemoglobin (Hb) and radiomics characteristic score (Radscore) were identified by LASSO regression analysis from a total of 21 variables studied. The model constructed using these three predictors displayed medium prediction ability. The area under the ROC curve was 0.895 (95% CI 0.844-0.945) in the training set, with a cutoff value of 0.651, precision of 0.957, and sensitivity of 0.718. In the validation set, it was 0.880 (95% CI 0.806-0.954), with a cutoff value of 0.655, precision of 0.930, and sensitivity of 0.698. DCA also confirmed the clinical benefit of the combined model. This simple and dependable nomogram model for clinical prediction can assist physicians in assessing preoperative nutritional risk in GC patients in a time-efficient and accurate manner to facilitate early identification and diagnosis.

Multi-headed Ensemble Residual CNN: A Powerful Tool for Fibroblast Growth Factor Prediction

Fibroblast Growth Factor (FGF) performs a significant role in the repair, nervous system, development, and maintenance, making it a promising target for treating neurological diseases such as Parkinson's, stroke, Alzheimer's, and Huntington's disorders. However, the detection of FGF remains challenging due to the high cost and time required by traditional methods. To address this issue, we propose the first sequence-based computational method, FGF-MERCNN, for identifying FGFs using deep learning. Novel datasets were constructed by converting primary sequences into numerical representations through Enhanced Group Amino Acid Composition (EGAAC), Dipeptide Deviation from Expected Mean (DDE), and Group Tripeptide Composition (GTPC). These features were combined and analyzed using several deep learning models, including Convolutional Neural Networks (CNN), Multilayer Perceptrons (MLP), Gated Recurrent Units (GRU), and Multiheaded Ensemble Residual CNN (MERCNN). Among these, the MERCNN model achieved superior performance with accuracies of 88.60% on the training set and 83.68% on the test set, as validated by 5-fold cross-validation. The results of this study represent a significant advancement in FGF detection, offering a faster, cost-effective solution with the potential to improve diagnostic and therapeutic strategies for neurological diseases.

Enhancing the content of phycoerythrin through the application of microplastics from Porphyridium cruentum produced in wastewater using machine learning methods

Microalgae can produce secondary metabolites like phycoerythrin (Phy). The effects of some microplastics (MPs), wastewater (WW), and light intensity (LI) parameters, including complex data sets, on Phy concentration from Porphyridium cruentum were investigated using machine learning methods in this study. Also, the deep learning (DL) model was developed to get the maximum phy concentration from the dataset. The dataset (232 data groups), including a feature set, polyethylene (PE), polypropylene (PP), polystyrene (PS), polyvinyl chloride (PVC), WW, LI, and an output variable, Phy, were randomly divided into training and test sets to create and evaluate the models. The highest experimental and predicted Phy concentrations were 52.3 mg/g and 58.32 mg/g in a scenario with 15% WW, 80 mg/L PE, PP, PS, and PVC, and a LI of 175 μmolm−2 s-1, respectively. The Pearson correlation coefficient (r) indicates a positive correlation between Phy and the variables PE (r = 0.35), PVC (r = 0.69), PP (r = 0.27), PS (r = 0.29), and LI (r = 0.22). However, variables such as WW (r = -0.05) have a weak correlation, and while PVC and PE showed the most significant effect on Phy concentration, WW had the lowest effect. Furthermore, LIME (local interpretable model-agnostic explanations) and SHAP (shapley additive explanations) provided us with important results for interpreting the random forest regression (RF) and DL models' predictions, respectively. The LIME and SHAP analyses suggest that the system with more PVC has a higher predicted Phy value. For WW, the reverse is true; higher WW values result in lower Phy predictions. Researchers were given the model explainability decision tree (DT) structure to study reactants' effects on output (Phy). In conclusion, the dye industry can use microalgae to treat WW contaminated with MPs while also producing high amounts of Phy using a DL model.

Predicting Antitumor Activity of Anthrapyrazole Derivatives using Advanced Machine Learning Techniques.

Anthrapyrazoles are a new class of antitumor agents and successors to anthracyclines possessing a broad range of antitumor activity in various model tumors. The present study introduces novel QSAR models for the prediction of antitumor activity of anthrapyrazole analogues. The predictive performance of four machine learning algorithms, namely artificial neural networks, boosted trees, multivariate adaptive regression splines, and random forest, was studied in terms of variation of the observed and predicted data, internal validation, predictability, precision, and accuracy. ANN and boosted trees algorithms met the validation criteria. It means that these procedures may be able to forecast the anticancer effects of the anthrapyrazoles studied. Evaluation of validation metrics, calculated for each approach, indicated the artificial neural network (ANN) procedure as the algorithm of choice, especially with regard to the obtained predictability as well as the lowest value of mean absolute error. The designed multilayer perceptron (MLP)-15-7-1 network displayed a high correlation between the predicted and the experimental pIC50 value for the training, test, and validation set. A conducted sensitivity analysis enabled an indication of the most important structural features of the studied activity. The ANN strategy combines topographical and topological information and can be used for the design and development of novel anthrapyrazole analogues as anticancer molecules.

Radiomic features of primary retroperitoneal sarcomas: a prognostic study

BackgroundRisk-stratification of patients with retroperitoneal sarcomas (RPS) relies on validated nomograms, such as Sarculator. This retrospective study investigated whether radiomic features extracted from computed tomography (CT) imaging could i) enhance the performance of Sarculator and ii) identify G3 dedifferentiated liposarcoma (DDLPS) or leiomyosarcoma (LMS), which are currently consider in a randomized clinical trial testing neoadjuvant chemotherapy. MethodsPatients with primary localized RPS treated with curative-intent surgery (2011-2015) and available pre-operative CT imaging were included. Regions of interest (ROIs) were manually annotated on both unenhanced and portal venous phase acquisitions. Top performing radiomic features were selected with outcome-specific random forest models, through generation of replicative experiments (contexts) where patients were split into training and testing sets. Endpoints were overall and disease-free survival (OS, DFS).Prognostic models for DFS and OS included the top five selected radiomic features and the Sarculator nomogram score.Models accuracy was assessed with Harrell’s Concordance (C-)index. ResultsThe study included 112 patients, with a median follow-up of 77 months (IQR 65-92 months).Sarculator alone achieved a C-index of 0.622 and 0.686 for DFS and OS, respectively. Radiomic features only marginally enhanced the prediction accuracy of Sarculator for OS (C-index=0.726, C-index gain: 0.04) or DFS (C-index=0.639, C-index gain: 0.017). Finally, radiomic features identified patients with G3 DDLPS or LMS with an accuracy of 0.806. ConclusionRadiomic features marginally improved the performance of Sarculator in RPS.However, they accurately identified G3 DDLPS or LMS at diagnosis, potentially improving patients selection for neoadjuvant treatments.

Machine Learning Model for Early Prediction of Survival in Gallbladder Adenocarcinoma: A comparison Study

The prognosis for gallbladder adenocarcinoma (GBAC), a highly malignant cancer, is not good. In order to facilitate individualized risk stratification and improve clinical decision-making, this study set out to create and validate a machine learning model that could accurately predict early survival outcomes in GBAC patients. Five models—RSF, Cox regression, GBM, XGBoost, and Deepsurv—were compared using data from the SEER database (2010–2020). The dataset was divided into training (70%) and validation (30%) sets, and the C-index, ROC curves, calibration curves, and decision curve analysis (DCA) were used to assess the model's performance. At 1, 2, and 3-year survival intervals, the RSF model performed better than the others in terms of calibration, discrimination, and clinical net benefit. The most important predictor of survival, according to SHAP analysis, is AJCC stage. Patients were divided into high, medium, and low-risk groups according to RSF-derived risk scores, which revealed notable variations in survival results. These results demonstrate the RSF model's potential as an early survival prediction tool for GBAC patients, which could enhance individualized treatment and decision-making.

Structure and Intensity Unbiased Translation for 2D Medical Image Segmentation.

Data distribution gaps often pose significant challenges to the use of deep segmentation models. However, retraining models for each distribution is expensive and time-consuming. In clinical contexts, device-embedded algorithms and networks, typically unretrainable and unaccessable post-manufacture, exacerbate this issue. Generative translation methods offer a solution to mitigate the gap by transferring data across domains. However, existing methods mainly focus on intensity distributions while ignoring the gaps due to structure disparities. In this paper, we formulate a new image-to-image translation task to reduce structural gaps. We propose a simple, yet powerful Structure-Unbiased Adversarial (SUA) network which accounts for both intensity and structural differences between the training and test sets for segmentation. It consists of a spatial transformation block followed by an intensity distribution rendering module. The spatial transformation block is proposed to reduce the structural gaps between the two images. The intensity distribution rendering module then renders the deformed structure to an image with the target intensity distribution. Experimental results show that the proposed SUA method has the capability to transfer both intensity distribution and structural content between multiple pairs of datasets and is superior to prior arts in closing the gaps for improving segmentation.

Eye movement disorders: A new approach to preliminary screening of Parkinson's disease.

To investigate the characteristics and diagnostic values of the eye movement disorders in patients with Parkinson's disease (PD-EMDs), this cross-sectional study enrolled 127 Chinese patients with PD and 80 healthy controls, and divided them into training and validation sets based on enrollment time. Performance in the five oculomotor paradigms was assessed using an infrared pupil and a corneal reflection tracking device. The primary characteristics of PD-EMDs were elucidated as inaccurate fixation with high deviation (frequency and total quantity); inaccurate saccades with delayed reaction and low velocity; saccadic pursuit with high deviation, delayed reaction, and velocity; and decreased visual search ability. Subgroup comparison shows that PD-EMDs can be related with PD stages, motor subtypes, frozen gait, and drowsiness. Finally, we developed and externally validated a model for PD preliminary screening using multivariate stepwise logistic regression analysis, comprising four oculomotor parameters (fixation accuracy, pro-saccade velocity, anti-saccade accuracy, and visual search duration), cognition score and educational years. The model has good feasibility with satisfactory performance on the receiver operating characteristic, calibration, and decision curves, and broad clinical applicability with better discrimination for more advanced PD patients and non-tremor-dominant PD patients. A nomogram was created to make the model more user-friendly in the clinical setting. Overall, we have demonstrated the presence of PD-EMDs and their prospective value for PD preliminary screening.

Multimodal ultrasound deep learning to detect fibrosis in early chronic kidney disease.

We developed a multimodal ultrasound (US) deep learning (DL) fusion model to automatically classify early fibrosis in patients with chronic kidney disease (CKD). This prospective study included patients with CKD who underwent continuous gray-scale US, superb microvascular imaging, and strain elastography from May to November 2022. According to the pathological tubular atrophy and interstitial fibrosis score, patients were divided into minimal and mild groups (affected area ≤10% and 11 - 25% of the total cortical volume, respectively). The dataset was divided into training (70%) and test (30%) sets. A DL model combining the features of the three US modes was developed to predict early fibrosis in patients with CKD. We compared these findings with the area under the receiver operating characteristic curve (AUC) of the clinical model by analyzing the receiver operating characteristic curve in the test set. The AUC of single-mode DL based on gray-scale US, superb microvascular imaging, and strain elastography was 0.682, 0.745, and 0.648, respectively, while that of the multimodal US DL model was 0.86. The accuracy, specificity, and sensitivity of the multimodal US DL model were 0.779, 0.767, and 0.796, respectively, and the negative and positive predictive values were 0.842 and 0.706, respectively. The AUC of the multimodal US DL model was significantly better than that of the single-mode DL and clinical models. The DL algorithm developed using multimodal US images can effectively predict early fibrosis in patients with CKD with significantly greater accuracy than single-mode DL or clinical models.

Generating implicit object fragment datasets for machine learning

One of the primary challenges inherent in utilizing deep learning models is the scarcity and accessibility hurdles associated with acquiring datasets of sufficient size to facilitate effective training of these networks. This is particularly significant in object detection, shape completion, and fracture assembly. Instead of scanning a large number of real-world fragments, it is possible to generate massive datasets with synthetic pieces. However, realistic fragmentation is computationally intensive in the preparation (e.g., pre-factured models) and generation. Otherwise, simpler algorithms such as Voronoi diagrams provide faster processing speeds at the expense of compromising realism. In this context, it is required to balance computational efficiency and realism. This paper introduces a GPU-based framework for the massive generation of voxelized fragments derived from high-resolution 3D models, specifically prepared for their utilization as training sets for machine learning models. This rapid pipeline enables controlling how many pieces are produced, their dispersion and the appearance of subtle effects such as erosion. We have tested our pipeline with an archaeological dataset, producing more than 1M fragmented pieces from 1,052 Iberian vessels (). Although this work primarily intends to provide pieces as implicit data represented by voxels, triangle meshes and point clouds can also be inferred from the initial implicit representation. To underscore the unparalleled benefits of CPU and GPU acceleration in generating vast datasets, we compared against a realistic fragment generator that highlights the potential of our approach, both in terms of applicability and processing time. We also demonstrate the synergies between our pipeline and realistic simulators, which frequently cannot select the number and size of resulting pieces. To this end, a deep learning model was trained over realistic fragments and our dataset, showcasing similar results.

Development of A Deep Learning Algorithm for Paneth Cell Density Quantification for Inflammatory Bowel Disease

Alterations in ileal Paneth cell (PC) density have been described in gut inflammatory diseases such as Crohn's disease (CD) and could be used as a biomarker for disease prognosis. However, quantifying PCs is time-intensive, a barrier for clinical workflow. Deep learning (DL) has transformed the development of robust and accurate tools for complex image evaluation. Our aim was to use DL to quantify PCs for use as a quantitative biomarker. A retrospective cohort of whole slide images (WSI) of ileal tissue samples from patients with/without inflammatory bowel disease (IBD) was used for the study. A pathologist-annotated training set of WSI were used to train a U-net two-stage DL model to quantify PC number, crypt number, and PC density. For validation, a cohort of 48 WSIs were manually quantified by study pathologists and compared to the DL algorithm, using root mean square error (RMSE) and the coefficient of determination (r2) as metrics. To test the value of PC quantification as a biomarker, resection specimens from patients with CD (n=142) and without IBD (n=48) patients were analysed with the DL model. Finally, we compared time to disease recurrence in patients with CD with low versus high DL-quantified PC density using Log-rank test. Initial one-stage DL model showed moderate accuracy in predicting PC density in cross-validation tests (RMSE=1.880, r2=0.641), but adding a second stage significantly improved accuracy (RMSE=0.802, r2=0.748). In the validation of the two-stage model compared to expert pathologists, the algorithm showed good performance up to RMSE=1.148, r2=0.708. The retrospective cross-sectional cohort had mean ages of 62.1 years in the patients without IBD and 38.6 years for the patients with CD. In the non-IBD cohort, 43.75% of the patients were male, compared to 49.3% of the patients with CD. Analysis by the DL model showed significantly higher PC density in non-IBD controls compared to the patients with CD (4.04 versus 2.99 PC/crypt). Finally, the algorithm quantification of PCs density in patients with CD showed patients with the lowest 25% PC density (Quartile 1) have significantly shorter recurrence-free interval (p=0.0399). The current model performance demonstrates the feasibility of developing a DL-based tool to measure PC density as a predictive biomarker for future clinical practice. This study was funded by the National Institutes of Health (NIH).

Playing for 3D Human Recovery.

Image- and video-based 3D human recovery (i.e., pose and shape estimation) have achieved substantial progress. However, due to the prohibitive cost of motion capture, existing datasets are often limited in scale and diversity. In this work, we obtain massive human sequences by playing the video game with automatically annotated 3D ground truths. Specifically, we contribute GTA-Human, a large-scale 3D human dataset generated with the GTA-V game engine, featuring a highly diverse set of subjects, actions, and scenarios. More importantly, we study the use of game-playing data and obtain five major insights. First, game-playing data is surprisingly effective. A simple frame-based baseline trained on GTA-Human outperforms more sophisticated methods by a large margin. For video-based methods, GTA-Human is even on par with the in-domain training set. Second, we discover that synthetic data provides critical complements to the real data that is typically collected indoor. We highlight that our investigation into domain gap provides explanations for our data mixture strategies that are simple yet useful, which offers new insights to the research community. Third, the scale of the dataset matters. The performance boost is closely related to the additional data available. A systematic study on multiple key factors (such as camera angle and body pose) reveals that the model performance is sensitive to data density. Fourth, the effectiveness of GTA-Human is also attributed to the rich collection of strong supervision labels (SMPL parameters), which are otherwise expensive to acquire in real datasets. Fifth, the benefits of synthetic data extend to larger models such as deeper convolutional neural networks (CNNs) and Transformers, for which a significant impact is also observed. We hope our work could pave the way for scaling up 3D human recovery to the real world.

In silico analysis of Design of Experiment methods for metabolic pathway optimization

Microbial cell factories allow the production of chemicals presenting an alternative to traditional fossil fuel-dependent production. However, finding the optimal expression of production pathway genes is crucial for the development of efficient production strains. Unlike sequential experimentation, combinatorial optimization captures the relationships between pathway genes and production, albeit at the cost of conducting multiple experiments. Fractional factorial designs followed by linear modeling and statistical analysis reduce the experimental workload while maximizing the information gained during experimentation. Although tools to perform and analyze these designs are available, guidelines for selecting appropriate factorial designs for pathway optimization are missing. In this study, we leverage a kinetic model of a seven-genes pathway to simulate the performance of a full factorial strain library. We compare this approach to resolution V, IV, III, and Plackett Burman (PB) designs. Additionally, we evaluate the performance of these designs as training sets for a random forest algorithm aimed at identifying best-producing strains. Evaluating the robustness of these designs to noise and missing data, traits inherent to biological datasets, we find that while resolution V designs capture most information present in full factorial data, they necessitate the construction of a large number of strains. On the other hand, resolution III and PB designs fall short in identifying optimal strains and miss relevant information. Besides, given the small number of experiments required for the optimization of a pathway with seven genes, linear models outperform random forest. Consequently, we propose the use of resolution IV designs followed by linear modeling in Design-Build-Test-Learn (DBTL) cycles targeting the screening of multiple factors. These designs enable the identification of optimal strains and provide valuable guidance for subsequent optimization cycles.

Integrative Bioinformatics and Experimental Analysis of Curcumin's Role in Regulating Ferroptosis to Combat Osteoporosis

This study utilized bioinformatics and data mining techniques to explore the molecular mechanism of curcumin in treating osteoporosis (OP) through the lens of ferroptosis, thereby identifying novel therapeutic targets. The datasets GSE35958, GSE35956, GSE7429, and GSE7158 were obtained from the Gene Expression Omnibus (GEO) database. GSE35958 and GSE35956 were employed as training sets for data integration, while GSE7429 and GSE7158 served as independent validation sets. Through retrieval from the FerrDb database, iron death-related genes (FRGs) were identified, and the differentially expressed genes (DEGs) were intersected to obtain differentially expressed FRGs (DEFRGs). Subsequently, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted, followed by the construction of a Protein-Protein Interaction (PPI) network to identify Hub genes. The DGIdb database was utilized to predict candidate drugs associated with the Hub genes, and molecular docking and in vitro experiments confirmed that curcumin targets the Hub genes EGFR and PTGS2.Through in vitro testing of curcumin on BMSC cells, researchers examined cell vitality, iron death, osteogenic differentiation, mineralization, and the impact on EGFR and PTGS2 levels. The analysis yielded 2212 DEGs, 484 FRGs, with 45 FDEGs at the intersection. GO analysis revealed involvement in regulating mitochondrial proteins, amino acid transport across plasma membranes, and protein ubiquitination. KEGG pathway analysis indicated associations with iron death, FoxO signaling, mTOR signaling, cell aging, osteoclast differentiation, and GSH metabolism. Utilizing the MCC algorithm, five Hub genes were identified: MAPK3, PTGS2, TGFB1, CYBB, and EGFR, showing diagnostic potential for iron death. Curcumin displayed affinity for EGFR and PTGS2, mitigating iron-induced effects on BMSCs such as increased reactive oxygen species, Fe3+ levels, and decreased mitochondrial membrane potential. Furthermore, curcumin reversed these effects, suggesting EGFR and PTGS2 as targets for curcumin to inhibit BMSC ferroptosis and potentially delay osteoporosis development. Maintaining iron homeostasis and targeting BMSC ferroptosis could offer therapeutic avenues for iron overload-induced osteoporosis.

Bridging real and simulated data for cross-spatial- resolution vegetation segmentation with application to rice crops

Accurate image segmentation is essential for image-based estimation of vegetation canopy traits, as it minimizes background interference. However, existing segmentation models often lack the generalization ability to effectively tackle both ground-based and aerial images across a wide range of spatial resolutions. To address this limitation, a cross-spatial-resolution image segmentation model for rice crop was trained using the integration of in-situ and in silico multi-resolution images. We collected more than 3,000 RGB images (real set) covering 17 different resolutions reflecting diverse canopy structures, illumination conditions and background in rice fields, with vegetation pixels annotated manually. Using the previously developed Digital Plant Phenotyping Platform, we created a simulated dataset (sim set) including 10,000 RGB images with resolutions ranging from 0.5 to 3.5 mm/pixel, accompanied by corresponding mask labels. By employing a domain adaptation technique, the simulated images were further transformed into visually realistic images while preserving the original labels, creating a simulated-to-realistic dataset (sim2real set). Building upon a SegFormer deep learning model, we demonstrated that training with multi-resolution samples led to more generalized segmentation results than single-resolution training on the real dataset. Our exploration of various integration strategies revealed that a training set of 9,600 sim2real images combined with only 60 real images achieved the same segmentation accuracy as 2,400 real images (IoU = 0.819, F1 = 0.901). Moreover, combining 2,400 real images and 1,200 sim2real images resulted in the best performing model, effective against six challenging situations, such as specular reflections and shadows. Compared with models trained with single-resolution samples and an established model (i.e., VegANN), our model effectively improved the estimation of both green fraction and green area index across spatial resoultions. The strategy of bridging real and simulated data for cross-resolution deep learning model is expected to be applicable to other crops. The best trained model is available at https://github.com/PheniX-Lab/crossGSD-seg.