Build Machine Learning Models Research Articles

Machine learning-based prediction of lithium-ion battery life cycle for capacity degradation modelling

With a particular focus on capacity degradation modeling, this paper offers a ground-breaking examination of the use of machine learning approaches for the precise prediction of lithium-ion battery life cycles. Because lithium-ion batteries are essential to many technological applications, it's critical to comprehend and anticipate their life cycles in order to maximize performance and guarantee sustainable energy solutions. The study starts with an extensive examination of the literature, assessing current approaches critically and setting the stage for the introduction of models based on machine learning. The process entails the methodical collecting of data across a range of operational settings, environmental variables, and charging-discharging cycles. Thorough preprocessing guarantees the dataset's consistency and quality for further machine learning model training. Predictive models are created using a variety of machine learning algorithms, including regression models, support vector machines, and deep neural networks. In order to improve prediction accuracy, the paper focuses on the reasoning behind model selection, parameter tuning, and the incorporation of ensemble approaches. In order to uncover important elements influencing the life cycles of lithium-ion batteries and provide important insights into degradation mechanisms, feature selection approaches are used. Using cross-validation techniques and real-world lithium-ion battery datasets, the built machine learning models go through rigorous evaluation and validation processes to determine their robustness, capacity for generalization, and performance metrics. Comparing machine learning-based predictions with conventional models, the results are presented and discussed, offering insights into the interpretability of the models and the identification of important affecting elements. In order to promote proactive maintenance and optimize battery usage, predictive models are integrated into real-time monitoring systems. The consequences for battery management systems are examined. The paper continues by discussing the challenges that come with using machine learning to estimate the life cycle of batteries and outlining possible directions for further research and development, such as scalability, interpretability, and the incorporation of emerging technologies. This research contributes to the ongoing efforts to increase the reliability and sustainability of lithium-ion battery technologies by highlighting the potential impact of machine learning on energy storage system optimization.

Identification of New Modulators and Inhibitors of Palmitoyl-Protein Thioesterase 1 for CLN1 Batten Disease and Cancer.

Palmitoyl-protein thioesterase 1 (PPT1) is an understudied enzyme that is gaining attention due to its role in the depalmitoylation of several proteins involved in neurodegenerative diseases and cancer. PPT1 is overexpressed in several cancers, specifically cholangiocarcinoma and esophageal cancers. Inhibitors of PPT1 lead to cell death and have been shown to enhance the killing of tumor cells alongside known chemotherapeutics. PPT1 is hence a viable target for anticancer drug development. Furthermore, mutations in PPT1 cause a lysosomal storage disorder called infantile neuronal ceroid lipofuscinosis (CLN1 disease). Molecules that can inhibit, stabilize, or modulate the activity of this target are needed to address these diseases. We used PPT1 enzymatic assays to identify molecules that were subsequently tested by using differential scanning fluorimetry and microscale thermophoresis. Selected compounds were also tested in neuroblastoma cell lines. The resulting PPT1 screening data was used for building machine learning models to help select additional compounds for testing. We discovered two of the most potent PPT1 inhibitors reported to date, orlistat (IC50 178.8 nM) and palmostatin B (IC50 11.8 nM). When tested in HepG2 cells, it was found that these molecules had decreased activity, indicating that they were likely not penetrating the cells. The combination of in vitro enzymatic and biophysical assays enabled the identification of several molecules that can bind or inhibit PPT1 and may aid in the discovery of modulators or chaperones. The molecules identified could be used as a starting point for further optimization as treatments for other potential therapeutic applications outside CLN1 disease, such as cancer and neurological diseases.

Reduced representative methylome profiling of cell-free DNA for breast cancer detection.

Whole-genome methylation sequencing of cfDNA is not cost-effective for tumor detection. Here, we introduce reduced representative methylome profiling (RRMP), which employs restriction enzyme for depletion of AT-rich sequence to achieve enrichment and deep sequencing of CG-rich sequences. We first verified the ability of RRMP to enrich CG-rich sequences using tumor cell genomic DNA and analyzed differential methylation regions between tumor cells and normal whole blood cells. We then analyzed cfDNA from 29 breast cancer patients and 27 non-breast cancer individuals to detect breast cancer by building machine learning models. RRMP captured 81.9% CpG islands and 75.2% gene promoters when sequenced to 10 billion base pairs, with an enrichment efficiency being comparable to RRBS. RRMP allowed us to assess DNA methylation changes between tumor cells and whole blood cells. Applying our approach to cfDNA from 29 breast cancer patients and 27 non-breast cancer individuals, we developed machine learning models that could discriminate between breast cancer and non-breast cancer controls (AUC = 0.85), suggesting possibilities for truly non-invasive cancer detection. We developed a new method to achieve reduced representative methylome profiling of cell-free DNA for tumor detection.

A Generalised Method for Friction Optimisation of Surface Textured Seals by Machine Learning

Friction behaviour is an important characteristic of dynamic seals. Surface texturing is an effective method to control the friction level without the need to change materials or lubricants. However, it is difficult to put the manual prediction of optimal friction reducing textures as a function of operating conditions into practice. Therefore, in this paper, we use machine learning techniques for the prediction of optimal texture parameters for friction optimisation. The application of pneumatic piston seals serves as an illustrative example to demonstrate the machine learning method and results. The analyses of this work are based on experimentally determined data of surface texture parameters, defined by the dimple diameter, distance, and depth. Furthermore friction data between the seal and the pneumatic cylinder are measured in different friction regimes from boundary over mixed up to hydrodynamic lubrication. A particular innovation of this work is the definition of a generalised method that guides the entire machine learning process from raw data acquisition to model prediction, without committing to only a few learning algorithms. A large number of 26 regression learning algorithms are used to build machine learning models through supervised learning to evaluate the suitability of different models in the specific application context. In order to select the best model, mathematical metrics and tribological relationships, like Stribeck curves, are applied and compared with each other. The resulting model is utilised in the subsequent friction optimisation step, in which optimal surface texture parameter combinations with the lowest friction coefficients are predicted over a defined interval of relative velocities. Finally, the friction behaviour is evaluated in the context of the model and optimal value combinations of the surface texture parameters are identified for different lubrication conditions.

Global, regional, and national mortality of tuberculosis attributable to alcohol and tobacco from 1990 to 2019: A modelling study based on the Global Burden of Disease study 2019.

Tuberculosis (TB) is expected to become the second leading single cause of death with several risk factors, but the related disease burden is currently unknown. We aimed to analyse the pre-coronavirus disease 2019 (COVID-19) changes in mortality of TB attributable to alcohol and tobacco worldwide from 1990 to 2019. We obtained data of TB deaths and age-standardised death rates attributed to alcohol and cigarette in 204 countries and territories from the Global Burden of Disease 2019 public database. We performed a spatial-temporal analysis of age-standardised death rate and the average annual per cent change (AAPC), after which we analysed the effects of gender and socio-demographic index on age-standardised death rate using an age-period-cohort model. Finally, we built machine learning models to predict the TB age-standardised death rate in 2035. We found that the global age-standardised death rate of TB attributable to alcohol consumption declined from 5.35 (95% uncertainty interval (UI) = 3.51, 7.00) in 1990 to 2.54 (95% UI = 1.65, 3.33) in 2019, with significant declines in Andean Latin America (AAPC = -7.59; 95% confidence interval (CI) = -8, -7.16, P < 0.05), East Asia (AAPC = -7.32; 95% CI = -8.00, -6.62, P < 0.05), and Central Latin America (AAPC = -7.31; 95% CI = -7.63, -6.99, P < 0.05). However, there was an increase in a few regions, especially in parts of Central Asia. The age-standardised death rate for TB attributable to cigarette smoking declined more rapidly than that for TB attributable to alcohol, from 7.45 (95% UI = 6.17, 8.72) to 2.21 (95% UI = 1.78, 2.64), especially in East Asia (AAPC = -6.64; 95% CI = -7.07, -6.2, P < 0.05), North Africa and Middle East (AAPC = -6.47; 95% CI = -6.67, -6.28, P < 0.05), and Andean Latin America (AAPC = -6.31; 95% CI = -6.87, -5.75, P < 0.05). However, TB attributable to cigarette smoking increased in parts of Eastern Europe. In both TB attributable to alcohol consumption and to cigarette smoking, the age-standardised death rate was much higher in men than in women. The age-period-cohort model results showed that TB attributable to alcohol consumption was the highest in older adults aged 60-80 years, while TB attributable to cigarette smoking was the highest in adults aged 40-60 years. Machine learning models projected that by 2035, the age-standardised death rate for TB attributable to alcohol consumption would be 1.29 (per 100 000 population), while the age-standardised death rate for TB attributable to cigarette consumption would be 0.37 (per 100 000 population), which might not achieve the 2035 global target for eliminating TB. Globally, the age-standardised death rate for TB attributable to alcohol consumption declined slower than that attributable to cigarette smoking. Controlling these two factors would help achieve the global goal of TB elimination, especially for the elderly who are at high risk.

Text-mining-based feature selection for anticancer drug response prediction.

Predicting anticancer treatment response from baseline genomic data is a critical obstacle in personalized medicine. Machine learning methods are commonly used for predicting drug response from gene expression data. In the process of constructing these machine learning models, one of the most significant challenges is identifying appropriate features among a massive number of genes. In this study, we utilize features (genes) extracted using the text-mining of scientific literatures. Using two independent cancer pharmacogenomic datasets, we demonstrate that text-mining-based features outperform traditional feature selection techniques in machine learning tasks. In addition, our analysis reveals that text-mining feature-based machine learning models trained on in vitro data also perform well when predicting the response of in vivo cancer models. Our results demonstrate that text-mining-based feature selection is an easy to implement approach that is suitable for building machine learning models for anticancer drug response prediction. https://github.com/merlab/text_features.

Reputation-Aware Federated Learning Client Selection based on Stochastic Integer Programming

Federated Learning(FL) has attracted wide research interest due to its potential in building machine learning models while preserving users&#x0027; data privacy. However, due to the distributive nature of FL, it is vulnerable to misbehavior from participating worker nodes. Thus, it is important to select clients to participate in FL. Recent studies on FL client selection focus on the perspective of improving model training efficiency and performance, without holistically considering potential misbehavior and the cost of hiring. To bridge this gap, we propose a first-of-its-kind reputation-aware <u>S</u>tochastic integer programming-based FL <u>C</u>lient <u>S</u>election method (SCS). It can optimally select and compensate clients with different reputation profiles. Extensive experiments show that SCS achieves the most advantageous performance-cost trade-off compared to other existing state-of-the-art approaches.

MACHINE LEARNING AND END-TO-END DEEP LEARNING FOR THE DETECTION OF CHRONIC HEART FAILURE FROM HEART SOUNDS

Chronic heart failure (CHF) affects over 26 million of people worldwide, and its incidence is increasing by 2% annually. Despite the significant burden that CHF poses and despite the ubiquity of sensors in our lives, methods for automatically detecting CHF are surprisingly scarce, even in the research community. We present a method for CHF detection based on heart sounds. The method combines classic MachineLearning (ML) and end-to-end Deep Learning (DL). The classic ML learns from expert features, and the DL learns from a spectro-temporal representation of the signal. The method was evaluated on recordings from 947 subjects from six publicly available datasets and one CHF dataset that was collected for this study. Using the same evaluation method as a recent PhysoNet challenge, the proposed method achieved a score of 89.3, which is 9.1 higher than the challenge’s baseline method. The method’s aggregated accuracy is 92.9% (error of 7.1%); while the experimental results are not directly comparable, this error rate is relatively close to the percentage of recordings labeled as “unknown” by experts (9.7%). Finally, we identified 15 expert features that are useful for building ML models to differentiate between CHF phases (i.e., in the decompensated phase during hospitalization and in the recompensated phase) with an accuracy of 93.2%. The proposed method shows promising results both for the distinction of recordings between healthy subjects and patients and for the detection of different CHF phases. This may lead to the easier identification of new CHF patients and the development of home-based CHF monitors for avoiding hospitalizations.

Discovery of novel ULK1 inhibitors through machine learning-guided virtual screening and biological evaluation.

Aim: Build a virtual screening model for ULK1 inhibitors based on artificial intelligence.Materials & methods: Build machine learning and deep learning classification models and combine molecular docking and biological evaluation to screen ULK1 inhibitors from 13million compounds. And molecular dynamics was used to explore the binding mechanism of active compounds.Results & conclusion: Possibly due to less available training data, machine learning models significantly outperform deep learning models. Among them, the Naive Bayes model has the best performance. Through virtual screening, we obtained three inhibitors with IC50 of μM level and they all bind well to ULK1. This study provides an efficient virtual screening model and three promising compounds for the study of ULK1 inhibitors.

Diurnal Carbon Monoxide Retrieval from FY-4B/GIIRS Using a Novel Machine Learning Method

Carbon monoxide (CO) is one of the primary reactive trace gases in the Earth’s atmosphere and plays an important role in atmospheric chemistry. The Geostationary Interferometric Infrared Sounder (GIIRS) onboard the FY-4 series satellites is currently the only geostationary hyperspectral thermal infrared sensor capable of monitoring the unprecedented hourly CO concentrations in East Asia during both daytime and nighttime. In this study, we presented a radiative transfer model-driven machine learning approach to quickly convert CO spectral features extracted from FY-4B/GIIRS into CO total columns. We built machine learning models for land and ocean regions separately from July 2022 to June 2023, and these models reproduced more than 97.77% (land) and 98.49% (ocean) of the CO column variance in the training set. We estimated the absolute uncertainty of the retrieved CO column based on error propagation theory and found that it is dominated by GIIRS measurement noise. We compared the machine learning retrieval results with optimal estimation and ground-based Fourier transform infrared measurements, and the results reveal the consistent spatial distribution and temporal variation across these different datasets. Our results confirm that the machine learning method has the potential to provide reliable CO products without the computationally intensive iterative process required by traditional retrieval methods. The diel cycle and monthly variation of CO over land and ocean demonstrate the value of GIIRS in monitoring the long-range transport of anthropogenic pollutants and biomass burning emissions.

A Machine Learning-Based Approach to Identifying Key Innovation Determinants for Industrial Firms: The Case of the Pharmaceutical Industry in China

With abundant resources consumed by innovation practices, it is of strategic importance for industrial firms to identify key determinants of innovation performance from both internal and external environments for better and effective innovation management. To this end, this article develops a new machine learning-based approach to examine a variety of factors influencing innovation performance. In particular, a two-process modeling approach is developed to build machine learning models efficiently for identifying key internal and external innovation factors. With extensive experiments using data collected from China's pharmaceutical industry, the approach identifies a set of key internal and external innovation factors and their best configurations in terms of two commonly used innovation performance measures. The findings enable pharmaceutical firms and local governments in China to manage innovation efforts and formulate management strategies for promoting industrial innovation. The article contributes to innovation management research by providing an objective and data-oriented approach for comprehensively exploring internal and external innovation factors and establishing their relationships with expected innovation performance. Although the approach is exemplified by China's pharmaceutical industry and the results cannot be generalized, it has general applicability to other industries or countries.

Application of machine learning methods for predicting the risk of stroke occurrence

In the paper, research was carried out in the medical field, which is very important for people and is gaining more and more importance every year. The study was aimed at predicting the occurrence of a stroke, this disease is a serious threat to people's health and lives. To build machine learning models that could solve the problem of predicting the occurrence of a stroke, a very unbalanced dataset was used, which made the work difficult. The best results were shown by the Random Forest model, which reached precision, recall, and f1-score equal to 90%. The obtained results can be useful for doctors and medical workers engaged in the diagnosis and treatment of stroke.

An integrative analysis of single-cell and bulk transcriptome and bidirectional mendelian randomization analysis identified C1Q as a novel stimulated risk gene for Atherosclerosis.

The role of complement component 1q (C1Q) related genes on human atherosclerotic plaques (HAP) is less known. Our aim is to establish C1Q associated hub genes using single-cell RNA sequencing (scRNA-seq) and bulk RNA analysis to diagnose and predict HAP patients more effectively and investigate the association between C1Q and HAP (ischemic stroke) using bidirectional Mendelian randomization (MR) analysis. HAP scRNA-seq and bulk-RNA data were download from the Gene Expression Omnibus (GEO) database. The C1Q-related hub genes was screened using the GBM, LASSO and XGBoost algorithms. We built machine learning models to diagnose and distinguish between types of atherosclerosis using generalized linear models and receiver operating characteristics (ROC) analyses. Further, we scored the HALLMARK_COMPLEMENT signaling pathway using ssGSEA and confirmed hub gene expression through qRT-PCR in RAW264.7 macrophages and apoE-/- mice. Furthermore, the risk association between C1Q and HAP was assessed through bidirectional MR analysis, with C1Q as exposure and ischemic stroke (IS, large artery atherosclerosis) as outcomes. Inverse variance weighting (IVW) was used as the main method. We utilized scRNA-seq dataset (GSE159677) to identify 24 cell clusters and 12 cell types, and revealed seven C1Q associated DEGs in both the scRNA-seq and GEO datasets. We then used GBM, LASSO and XGBoost to select C1QA and C1QC from the seven DEGs. Our findings indicated that both training and validation cohorts had satisfactory diagnostic accuracy for identifying patients with HPAs. Additionally, we confirmed SPI1 as a potential TF responsible for regulating the two hub genes in HAP. Our analysis further revealed that the HALLMARK_COMPLEMENT signaling pathway was correlated and activated with C1QA and C1QC. We confirmed high expression levels of C1QA, C1QC and SPI1 in ox-LDL-treated RAW264.7 macrophages and apoE-/- mice using qPCR. The results of MR indicated that there was a positive association between the genetic risk of C1Q and IS, as evidenced by an odds ratio (OR) of 1.118 (95%CI: 1.013-1.234, P = 0.027). The authors have effectively developed and validated a novel diagnostic signature comprising two genes for HAP, while MR analysis has provided evidence supporting a favorable association of C1Q on IS.

Detection of Helicobacter pylori Infection in Human Gastric Fluid Through Surface-Enhanced Raman Spectroscopy Coupled With Machine Learning Algorithms

Diagnostic methods for Helicobacter pylori infection include, but are not limited to, urea breath test, serum antibody test, fecal antigen test, and rapid urease test. However, these methods suffer drawbacks such as low accuracy, high false-positive rate, complex operations, invasiveness, etc. Therefore, there is a need to develop simple, rapid, and noninvasive detection methods for H. pylori diagnosis. In this study, we propose a novel technique for accurately detecting H. pylori infection through machine learning analysis of surface-enhanced Raman scattering (SERS) spectra of gastric fluid samples that were noninvasively collected from human stomachs via the string test. One hundred participants were recruited to collect gastric fluid samples noninvasively. Therefore, 12,000 SERS spectra (n = 120 spectra/participant) were generated for building machine learning models evaluated by standard metrics in model performance assessment. According to the results, the Light Gradient Boosting Machine algorithm exhibited the best prediction capacity and time efficiency (accuracy = 99.54% and time = 2.61 seconds). Moreover, the Light Gradient Boosting Machine model was blindly tested on 2,000 SERS spectra collected from 100 participants with unknown H. pylori infection status, achieving a prediction accuracy of 82.15% compared with qPCR results. This novel technique is simple and rapid in diagnosing H. pylori infection, potentially complementing current H. pylori diagnostic methods.

Bridging the Coordination Chemistry of Small Compounds and Metalloproteins Using Machine Learning.

Metalloproteins require metal ions as cofactors to catalyze specific reactions with remarkable efficiency and specificity. In various electron transfer reactions, metals in the active sites change their oxidation states to facilitate the biochemical reactions. Cryogenic electron microscopy, X-ray, and X-ray free electron laser (XFEL) crystallography are used to image metalloproteins to understand the reaction mechanisms. However, radiation damage in cryoEM and X-ray crystallography, and the challenge of generating homogeneous crystals and keeping the appropriate experimental conditions for all the crystals in XFEL crystallography, may alter the oxidation states. Here, we build machine learning models trained on a large data set from the Cambridge Crystallographic Data Center to evaluate the metal oxidation states. The models yield high accuracy scores (from 82% to 94%) for all metals in the small molecules. Then, they were used to predict the oxidation states of more than 30 000 metal clusters in metalloproteins with Fe, Mn, Co, and Cu in their active sites. We found that most of the metals exist in the lower oxidation states (Fe2+ 77%, Mn2+ 85%, Co2+ 65%, and Cu+ 64%), and these populations correlate with the standard reduction potentials of the metal ions. Furthermore, we found no clear correlation between these populations and the resolution of the structures, which suggests no significant dependence of these predictions on the resolution. Our models represent a valuable tool for evaluating the oxidation states of the metals in metalloproteins imaged with different techniques. The data files and the machine learning code are available in a public GitHub repository: https://github.com/mamin03/OxitationStatesMetalloprotein.git.

CircuitNet: An Open-Source Dataset for Machine Learning in VLSI CAD Applications With Improved Domain-Specific Evaluation Metric and Learning Strategies

The design automation community has been actively exploring machine learning for VLSI CAD. Many studies have explored learning-based techniques for cross-stage prediction tasks in the design flow. Although building machine learning models usually requires a large amount of data, most studies can only generate small internal datasets for validation due to the lack of large public datasets. Such a situation challenges the research in this field and raises potential issues like difficulty in benchmarking and reproducing results, limited research scope on small internal datasets, and high bar for new researchers. Therefore, in this paper, we present an open-source dataset called “CircuitNet” for machine learning tasks in VLSI CAD. The dataset consists of more than 10K samples extracted from versatile runs of commercial design tools based on 6 open-source RISC-V designs which support typical cross-stage prediction tasks, such as routability and IR drop prediction, with extensive benchmarking on recent models. With the dataset prepared, we identify two practical challenges, data imbalance and model transferability, for machine learning application in CAD. To overcome data imbalance, we propose a loss function, biased loss, to give more weight to the minority, leading to 2% congestion reduction in routability driven placement. We test the model transferability from RISC-V designs to ISPD 2015 contest designs in congestion prediction with several transfer learning methods, and further proposed a knowledge distillation based transfer learning framework with up to 20% accuracy improvement. We believe this dataset can open up new opportunities for machine learning in CAD research and beyond.

A Machine Learning Approach to Identify Individualized Baseline MRI, Plasma, Cardiovascular and Lifestyle Predictors of a 5‐Year Cognitive Decline

AbstractBackgroundThe mapping between risk factors and biomarkers with cognitive performance and decline in the older adults is complex and varies significantly across different age and sex groups. Our goals were to (i) develop a machine learning (ML) model to predict baseline cognition and cognitive decline over a 5‐year follow‐up using only baseline assessments, and (ii) identify important modifiable and non‐modifiable predictors of cognitive decline across different age strata.MethodWe utilized longitudinal data from 2219 participants (mean age (SD): 70 (12) years, 49% male) from Mayo Clinic Study of Aging who had 43 features belonging to cardiovascular health, plasma, MRI (diffusion and structural), and lifestyle measures. Cognition was measured using a global cognition z‐score. We built ML models to predict baseline cognition and rate of cognitive decline over 5‐years (linear regression slopes over all individual followup measurements for each participant) using baseline assessments as inputs. We then built a model for each category separately and as a pooled model with all measures included, incorporating education and sex as demographic features in all models (Fig. 1). We applied a multi‐layer stack ensemble ML technique that utilized repeated k‐fold bagging to maximize accuracy. Importance of the top predictors were assessed in three age strata (55‐65 years (31%), 65‐80 years (46%), 80‐95 years (23%)).ResultThe pooled model explained the largest variance for prediction of baseline cognition and cognitive decline (Fig. 2). In individuals ≤80 years, lower NfL was most predictive of higher baseline cognition. A combination of imaging and plasma features predicted the 5‐year cognitive decline. Additionally, in individuals ≤65 years, diabetes and hypertension were important predictors of faster cognitive decline and in those &gt;80, lower BMI and midlife physical/cognitive activity was an important predictor of faster cognitive decline (Fig. 3).ConclusionThe plasma model was best for baseline cognition prediction and imaging model was best for cognitive decline prediction. There was significant heterogeneity in important features across the age strata. In individuals &gt;65 years, plasma GFAP was the most important predictor of cognitive decline. Our model allowed identification of key risk factors that predicted cognitive decline.

A Conceptual Framework of Data Readiness for the Health and Aging Brain Study‐Health Disparities

AbstractBackgroundThe Health and Aging Brain Study: Health Disparities (HABS‐HD) seeks to understand the biological, social and environmental factors that impact brain aging among diverse communities. HABS‐HD, like many other NIH funded data‐sharing projects, has important data assets for various uses, including social, environmental and behavioral data, and multiple data flow pathways. Machine learning (ML) develops algorithms and models to continuously improve itself over time, but the determination of data quality and its readiness are needed for these models to operate efficiently. Therefore, developing a data readiness reporting methodology has become a very urgent task for HABS‐HD.MethodIn this study, we developed a conceptual framework of data readiness. First, we analyzed the missing data percentage and used ML‐Based Multiple Imputation (MLMI) for missing data imputation. Then, we performed SVM based on Recursive Feature Elimination and Cross Validation (SVM‐RFE‐CV) for feature elimination and outlier removal. Lastly, we rated the data readiness based on the three metrics: missing data percentage, performance before feature engineering, and performance after feature engineering to rate data readiness. All the three scores were averaged to rate the overall readiness of data.ResultA framework for calculating overall average score for readiness of data was presented (1 stands for completely accessible, 0 for not accessible at all, and 0.5 for neutral). Our results show that the framework of data readiness was straightforward and useful in assessing how ready the HABS‐HD data is for ML.ConclusionThe systematic analysis of readiness of data before building ML models is of utmost importance. And it has a significant impact on biomarker discovery and disease prediction application for Alzheimer’s disease. The conceptual framework of data readiness works well for our Alzheimer’s disease models in HABS‐HD. It can also be applied to other disease data readiness reporting.

Building Machine Learning Models to Improve Hypertension Diagnosis

Hypertension (HT) is a major risk factor for heart disease, stroke, and kidney disease. A diagnosis of HT is based on two or more blood pressure (BP) readings taken on two or more visits, and such diagnosis may vary depending on the threshold BP values utilized by specific guidelines. Our hypertension identification approach has been recently proposed and validated for HT diagnosis and could be used as ground truth in modeling of latent HT diagnosis. Our approach to modeling of latent HT diagnosis with high precision will leverage analytics to “Big Data”, such as electronic health records (EHRs). In this work, we will review the time complexity underlying the classical ML methods XGBoost (XGB) and Artificial Neural Networks (ANN). In particular, we compare the XGB and ANN to leverage their strengths. The performance of all algorithms for diagnosing HT will be characterized using the area under the curve (AUC) approach on a big EHR longitudinal dataset. Predictor variable importance and model interpretability will be assessed using the Shapley values approach.

Predicting Binding Affinity Between MHC-I Receptor and Peptides Based on Molecular Docking and Protein-peptide Interaction Interface Characteristics

Background:Predicting protein-peptide binding affinity is one of the leading research subjects in peptide drug design and repositioning. In previous studies, models constructed by researchers just used features of peptide structures. These features had limited information and could not describe the proteinpeptide interaction mode. This made models and predicted results lack interpretability in pharmacy and biology, which led to the protein-peptide interaction mode not being reflected. Therefore, it was of little significance for the design of peptide drugs.Objective:Considering the protein-peptide interaction mode, we extracted protein-peptide interaction interface characteristics and built machine learning models to improve the performance and enhance the interpretability of models.Methods:Taking MHC-I protein and its binding peptides as the research object, protein-peptide complexes were obtained by molecular docking, and 94 protein-peptide interaction interface characteristics were calculated. Then ten important features were selected using recursive feature elimination to construct SVR, RF, and MLP models to predict protein-peptide binding affinity.Results:The MAE of the SVR, RF and MLP models constructed using protein-peptide interaction interface characteristics are 0.2279, 0.2939 and 0.2041, their MSE are 0.1289, 0.1308 and 0.0780, and their R2 reached 0.8711, 0.8692 and 0.9220, respectively.Conclusion:The model constructed using protein-peptide interaction interface characteristics showed better prediction results. The key features for predicting protein-peptide binding affinity are the bSASA of negatively charged species, hydrogen bond acceptor, hydrophobic group, planarity, and aromatic ring.