OBJECTIVE: To identify the key risk factors for venous thromboembolism (VTE) in urological inpatients based on the Caprini scale using an interpretable machine learning method.
METHODS: VTE risk data of urological inpatients were obtained based on the Caprini scale in the case hospital. Based on the data, the Boruta method was used to further select the key variables from the 37 variables in the Caprini scale. Furthermore, decision rules corresponding to each risk level were generated using the rough set (RS) method. Finally, random forest (RF), support vector machine (SVM), and backpropagation artificial neural network (BPANN) were used to verify the data accuracy and were compared with the RS method.
RESULTS: Following the screening, the key risk factors for VTE in urology were \"(C1) Age,\" \"(C2) Minor Surgery planned,\" \"(C3) Obesity (BMI > 25),\" \"(C8) Varicose veins,\" \"(C9) Sepsis (< 1 month),\" (C10) \"Serious lung disease incl. pneumonia (< 1month) \" (C11) COPD,\" \"(C16) Other risk,\" \"(C18) Major surgery (> 45 min),\" \"(C19) Laparoscopic surgery (> 45 min),\" \"(C20) Patient confined to bed (> 72 h),\" \"(C18) Malignancy (present or previous),\" \"(C23) Central venous access,\" \"(C31) History of DVT/PE,\" \"(C32) Other congenital or acquired thrombophilia,\" and \"(C34) Stroke (< 1 month.\" According to the decision rules of different risk levels obtained using the RS method, \"(C1) Age,\" \"(C18) Major surgery (> 45 minutes),\" and \"(C21) Malignancy (present or previous)\" were the main factors influencing mid- and high-risk levels, and some suggestions on VTE prevention were indicated based on these three factors. The average accuracies of the RS, RF, SVM, and BPANN models were 79.5%, 87.9%, 92.6%, and 97.2%, respectively. In addition, BPANN had the highest accuracy, recall, F1-score, and precision.
CONCLUSIONS: The RS model achieved poorer accuracy than the other three common machine learning models. However, the RS model provides strong interpretability and allows for the identification of high-risk factors and decision rules influencing high-risk assessments of VTE in urology. This transparency is very important for clinicians in the risk assessment process.

Acute kidney injury (AKI) is one of the most important lethal factors for patients admitted to intensive care units (ICUs), and timely high-risk prognostic assessment and intervention are essential to improving patient prognosis. In this study, a stacking model using the MIMIC-III dataset with a two-tier feature selection approach was developed to predict the risk of in-hospital mortality in ICU patients admitted for AKI. External validation was performed using separate MIMIC-IV and eICU-CRD. The area under the curve (AUC) was calculated using the stacking model, and features were selected using the Boruta and XGBoost feature selection methods. This study compares the performance of a stacking model using two-tier feature selection with a model using single-tier feature selection (XGBoost: 85; Boruta: 83; two-tier: 0.91). The predictive effectiveness of the stacking model was further validated by using different datasets (Validation 1: 0.83; Validation 2: 0.85) and comparing it with a simpler model and traditional clinical scores (SOFA: 0.65; APACH IV: 0.61). In addition, this study combined interpretable techniques and causal inference to analyze the causal relationship between features and predicted outcomes.

Asparagine peptide lyase (APL) is among the seven groups of proteases, also known as proteolytic enzymes, which are classified according to their catalytic residue. APLs are synthesized as precursors or propeptides that undergo self-cleavage through autoproteolytic reaction. At present, APLs are grouped into 10 families belonging to six different clans of proteases. Recognizing their critical roles in many biological processes including virus maturation, and virulence, accurate identification and characterization of APLs is indispensable. Experimental identification and characterization of APLs is laborious and time-consuming. Here, we developed APLpred, a novel support vector machine (SVM) based predictor that can predict APLs from the primary sequences. APLpred was developed using Boruta-based optimal features derived from seven encodings and subsequently trained using five machine learning algorithms. After evaluating each model on an independent dataset, we selected APLpred (an SVM-based model) due to its consistent performance during cross-validation and independent evaluation. We anticipate APLpred will be an effective tool for identifying APLs. This could aid in designing inhibitors against these enzymes and exploring their functions. The APLpred web server is freely available at https://procarb.org/APLpred/.

OBJECTIVE: Headache is one of the most prevalent and disabling health conditions globally. We prospectively explored the urban exposome in relation to weekly occurrence of headache episodes using data from the Dutch population-based Occupational and Environmental Health Cohort Study (AMIGO).
METHODS: Participants (N = 7,339) completed baseline and follow-up questionnaires in 2011 and 2015, reporting headache frequency. Information on the urban exposome covered 80 exposures across 10 domains, such as air pollution, electromagnetic fields, and lifestyle and socio-demographic characteristics. We first identified all relevant exposures using the Boruta algorithm and then, for each exposure separately, we estimated the average treatment effect (ATE) and related standard error (SE) by training causal forests adjusted for age, depression diagnosis, painkiller use, general health indicator, sleep disturbance index and weekly occurrence of headache episodes at baseline.
RESULTS: Occurrence of weekly headache was 12.5 % at baseline and 11.1 % at follow-up. Boruta selected five air pollutants (NO2, NOX, PM10, silicon in PM10, iron in PM2.5) and one urban temperature measure (heat island effect) as factors contributing to the occurrence of weekly headache episodes at follow-up. The estimated causal effect of each exposure on weekly headache indicated positive associations. NO2 showed the largest effect (ATE = 0.007 per interquartile range (IQR) increase; SE = 0.004), followed by PM10 (ATE = 0.006 per IQR increase; SE = 0.004), heat island effect (ATE = 0.006 per one-degree Celsius increase; SE = 0.007), NOx (ATE = 0.004 per IQR increase; SE = 0.004), iron in PM2.5 (ATE = 0.003 per IQR increase; SE = 0.004), and silicon in PM10 (ATE = 0.003 per IQR increase; SE = 0.004).
CONCLUSIONS: Our results suggested that exposure to air pollution and heat island effects contributed to the reporting of weekly headache episodes in the study population.

Regression models are constructed to predict glucose and lactate concentrations from near-infrared spectra in culture media. The partial least-squares (PLS) regression technique is employed, and we investigate the improvement in the predictive ability of PLS models that can be achieved using wavelength selection and transfer learning. We combine Boruta, a nonlinear variable selection method based on random forests, with variable importance in projection (VIP) in PLS to produce the proposed variable selection method, VIP-Boruta. Furthermore, focusing on the situation where both culture medium samples and pseudo-culture medium samples can be used, we transfer pseudo media to culture media. Data analysis with an actual dataset of culture media and pseudo media confirms that VIP-Boruta can effectively select appropriate wavelengths and improves the prediction ability of PLS models, and that transfer learning with pseudo media enhances the predictive ability. The proposed method could reduce the prediction errors by about 61% for glucose and about 16% for lactate, compared to the traditional PLS model.

Machine learning methods have been used in identifying omics markers for a variety of phenotypes. We aimed to examine whether a supervised machine learning algorithm can improve identification of alcohol-associated transcriptomic markers. In this study, we analysed array-based, whole-blood derived expression data for 17 873 gene transcripts in 5508 Framingham Heart Study participants. By using the Boruta algorithm, a supervised random forest (RF)-based feature selection method, we selected twenty-five alcohol-associated transcripts. In a testing set (30 % of entire study participants), AUC (area under the receiver operating characteristics curve) of these twenty-five transcripts were 0·73, 0·69 and 0·66 for non-drinkers v. moderate drinkers, non-drinkers v. heavy drinkers and moderate drinkers v. heavy drinkers, respectively. The AUC of the selected transcripts by the Boruta method were comparable to those identified using conventional linear regression models, for example, AUC of 1958 transcripts identified by conventional linear regression models (false discovery rate < 0·2) were 0·74, 0·66 and 0·65, respectively. With Bonferroni correction for the twenty-five Boruta method-selected transcripts and three CVD risk factors (i.e. at P < 6·7e-4), we observed thirteen transcripts were associated with obesity, three transcripts with type 2 diabetes and one transcript with hypertension. For example, we observed that alcohol consumption was inversely associated with the expression of DOCK4, IL4R, and SORT1, and DOCK4 and SORT1 were positively associated with obesity, and IL4R was inversely associated with hypertension. In conclusion, using a supervised machine learning method, the RF-based Boruta algorithm, we identified novel alcohol-associated gene transcripts.

The rising demand for cocoa powder has resulted in an upsurge in market prices, leading to the emergence of adulteration practices aimed at achieving economic benefits. This study aimed to detect and quantify cocoa powder adulteration using visible and near-infrared spectroscopy (Vis-NIRS). The adulterants used in this study were powdered carob, cocoa shell, foxtail millet, soybean, and whole wheat. The NIRS data could not be resolved using Savitzky-Golay smoothing. Nevertheless, the application of a random forest and support vector machine successfully classified the samples with 100% accuracy. Quantification of adulteration using partial least squares (PLS), Lasso, Ridge, elastic Net, and RF regressions provided R2 higher than 0.96 and root mean square error <2.6. Coupling PLS with the Boruta algorithm produced the most reliable regression model (R2 = 1, RMSE = 0.0000). Finally, an online application was prepared to facilitate the determination of adulterants in the cocoa powder.

Interpretable machine learning models are instrumental in disease diagnosis and clinical decision-making, shedding light on relevant features. Notably, Boruta, SHAP (SHapley Additive exPlanations), and BorutaShap were employed for feature selection, each contributing to the identification of crucial features. These selected features were then utilized to train six machine learning algorithms, including LR, SVM, ETC, AdaBoost, RF, and LR, using diverse medical datasets obtained from public sources after rigorous preprocessing. The performance of each feature selection technique was evaluated across multiple ML models, assessing accuracy, precision, recall, and F1-score metrics. Among these, SHAP showcased superior performance, achieving average accuracies of 80.17%, 85.13%, 90.00%, and 99.55% across diabetes, cardiovascular, statlog, and thyroid disease datasets, respectively. Notably, the LGBM emerged as the most effective algorithm, boasting an average accuracy of 91.00% for most disease states. Moreover, SHAP enhanced the interpretability of the models, providing valuable insights into the underlying mechanisms driving disease diagnosis. This comprehensive study contributes significant insights into feature selection techniques and machine learning algorithms for disease diagnosis, benefiting researchers and practitioners in the medical field. Further exploration of feature selection methods and algorithms holds promise for advancing disease diagnosis methodologies, paving the way for more accurate and interpretable diagnostic models.

BACKGROUND: In 2016, the UK had 4.5 million people with diabetes, predominantly Type-2 Diabetes Mellitus (T2DM). The NHS allocates £10 billion (9% of its budget) to manage diabetes. Metformin is the primary treatment for T2DM, but 35% of patients don\'t benefit from it, leading to complications. This study aims to delve into metformin\'s efficacy using clinical, genomic, and proteomic data to uncover new biomarkers and build a Machine Learning predictor for early metformin response detection.
METHODS: Here we report analysis from a T2DM dataset of individuals prescribed metformin monotherapy from the Diastrat cohort recruited at the Altnagelvin Area Hospital, Northern Ireland.
RESULTS: In the clinical data analysis, comparing responders (those achieving HbA1c ≤ 48 mmol/mol) to non-responders (with HbA1c > 48 mmol/mol), we identified that creatinine levels and bodyweight were more negatively correlated with response than non-response. In genomic analysis, we identified statistically significant (p-value <0.05) variants rs6551649 (LPHN3), rs6551654 (LPHN3), rs4495065 (LPHN3) and rs7940817 (TRPC6) which appear to differentiate the responders and non-responders. In proteomic analysis, we identified 15 statistically significant (p-value <0.05, q-value <0.05) proteomic markers that differentiate controls, responders, non-responders and treatment groups, out of which the most significant were HAOX1, CCL17 and PAI that had fold change ∼2. A machine learning model was build; the best model predicted non-responders with 83% classification accuracy.
CONCLUSIONS: Further testing in prospective validation cohorts is required to determine the clinical utility of the proposed model.

UNASSIGNED: This study used machine learning algorithms to identify important variables and predict postinduction hypotension (PIH) in patients undergoing colorectal tumor resection surgery.
UNASSIGNED: Data from 318 patients who underwent colorectal tumor resection under general anesthesia were analyzed. The training and test sets are divided based on the timeline. The Boruta algorithm was used to screen relevant basic characteristic variables and establish a model for the training set. Four models, regression tree, K-nearest neighbor, neural network, and random forest (RF), were built using repeated cross-validation and hyperparameter optimization. The best model was selected, and a sorting chart of the feature variables, a univariate partial dependency profile, and a breakdown profile were drawn. R2, mean absolute error (MAE), mean squared error (MSE), and root MSE (RMSE) were used to plot regression fitting curves for the training and test sets.
UNASSIGNED: The basic feature variables associated with the Boruta screening were age, sex, body mass index, L3 skeletal muscle index, and HUAC. In the optimal RF model, R2 was 0.7708 and 0.7591, MAE was 0.0483 and 0.0408, MSE was 0.0038 and 0.0028, and RMSE was 0.0623 and 0.0534 for the training and test sets, respectively.
UNASSIGNED: A high-performance algorithm was established and validated to demonstrate the degree of change in blood pressure after induction to control important characteristic variables and reduce PIH occurrence.