Algebraic shortcuts for leave-one-out cross-validation in supervised network inference

The causal relationship between gut microbiota and nine infectious diseases: a two-sample Mendelian randomization analysis

Background

Evidence from observational studies and clinical trials has associated gut microbiota with infectious diseases. However, the causal relationship between gut microbiota and infectious diseases remains unclear.

Methods

We identified gut microbiota based on phylum, class, order, family, and genus classifications, and obtained infectious disease datasets from the IEU OpenGWAS database. The two-sample Mendelian Randomization (MR) analysis was then performed to determine whether the gut microbiota were causally associated with different infectious diseases. In addition, we performed reverse MR analysis to test for causality.

Results

Herein, we characterized causal relationships between genetic predispositions in the gut microbiota and nine infectious diseases. Eight strong associations were found between genetic predisposition in the gut microbiota and infectious diseases. Specifically, the abundance of class Coriobacteriia, order Coriobacteriales, and family Coriobacteriaceae was found to be positively associated with the risk of lower respiratory tract infections (LRTIs). On the other hand, family Acidaminococcaceae, genus Clostridiumsensustricto1, and class Bacilli were positively associated with the risk of endocarditis, cellulitis, and osteomyelitis, respectively. We also discovered that the abundance of class Lentisphaeria and order Victivallales lowered the risk of sepsis.

Conclusion

Through MR analysis, we found that gut microbiota were causally associated with infectious diseases. This finding offers new insights into the microbe-mediated infection mechanisms for further clinical research.

Causal associations between platelet count, alcohol consumption, and the risk of liver hepatocellular carcinoma based on a Mendelian randomization study

Background and aims

Liver hepatocellular carcinoma (LIHC) exhibits a multifactorial etiology, insidious onset, and a significantly low 5-year survival rate. We aimed to evaluate the causal impact of exposure factors (Alzheimer's disease, platelet count, ambidextrousness, cigarettes smoked per day, alcohol consumption, and endocarditis) on the risk of LIHC using a two-sample Mendelian randomization (MR) study.

Methods

Independent single nucleotide polymorphisms (SNPs) strongly associated with Alzheimer's disease, platelet count, ambidextrousness, daily cigarette consumption, alcohol intake, and endocarditis were selected as instrumental variables (IVs) from the corresponding genome-wide association studies (GWAS). Genetic summary statistics for LIHC came from a GWAS that included 168 cases and 372,016 controls of European individuals. Multivariable MR analyses were performed to find the causal association between 6 exposure factors and LIHC risk. The inverse-variance weighted (IVW)-MR was employed as the primary analysis, and the MR-Egger regression, LASSO regression, and weighted Median approaches were performed as complementary analyses.

Results

Multivariable MR analysis showed causal association between Alzheimer's disease [Odds ratio (OR) = 0.9999, 95% confidence intervals (CI) = 0.9998-0.9999, p = 0.0010], platelet count (OR = 0.9997, 95% CI = 0.9995-0.9999, p = 0.0066), alcohol consumption (OR = 0.9994, 95% CI = 0.9990-0.9999, p = 0.0098) and the LIHC outcome. After IVW-MR, MR-Egger and LASSO tests, the results are still significant. Next, we used different MR Methods to analyze platelet count, alcohol consumption, and Alzheimer's disease separately. Moreover, both funnel plots and MR-Egger intercepts provided compelling evidence to refute the presence of directional pleiotropy in the association between platelet count, alcohol consumption, Alzheimer's disease and the risk of LIHC. The IVW-MR analysis revealed a significant causal association between an elevated platelet count and a reduced risk of LIHC (OR = 0.9996, 95% CI= 0.9995-0.9998, p = 0.0005). Similarly, the analysis of weighted median revealed a negative correlation between platelet count and the risk of LIHC (OR = 0.9995, 95% CI = 0.9993-0.9999; p = 0.0160). Conversely, we observed a positive causal effect of alcohol consumption on the incidence of LIHC (OR = 1.0004, 95% CI = 0.9999-1.0009). However, no significant causal relationship was found between alcohol assumption, Alzheimer's disease, and LIHC susceptibility.

Conclusions

A significant causal relationship exists between platelet count, alcohol consumption, Alzheimer's disease, and an increased risk of LIHC. The study presents compelling evidence for a genetically predicted decreased susceptibility to LIHC based on platelet count. The research implies that elevated platelet count may serve as a protective mechanism against LIHC. These findings may inform clinical strategies for LIHC prevention.

Associations between type 2 diabetes mellitus and chronic liver diseases: evidence from a Mendelian randomization study in Europeans and East Asians

Objective

Multiple observational studies have demonstrated an association between type 2 diabetes mellitus (T2DM) and chronic liver diseases (CLDs). However, the causality of T2DM on CLDs remained unknown in various ethnic groups.

Methods

We obtained instrumental variables for T2DM and conducted a two-sample mendelian randomization (MR) study to examine the causal effect on nonalcoholic fatty liver disease (NAFLD), hepatocellular carcinoma (HCC), viral hepatitis, hepatitis B virus (HBV) infection, and hepatitis C virus (HCV) infection risk in Europeans and East Asians. The primary analysis utilized the inverse variance weighting (IVW) technique to evaluate the causal relationship between T2DM and CLDs. In addition, we conducted a series of rigorous analyses to bolster the reliability of our MR results.

Results

In Europeans, we found that genetic liability to T2DM has been linked with increased risk of NAFLD (IVW : OR =1.3654, 95% confidence interval [CI], 1.2250-1.5219, p=1.85e-8), viral hepatitis (IVW : OR =1.1173, 95%CI, 1.0271-1.2154, p=0.0098), and a suggestive positive association between T2DM and HCC (IVW : OR=1.2671, 95%CI, 1.0471-1.5333, p=0.0150), HBV (IVW : OR=1.1908, 95% CI, 1.0368-1.3677, p=0.0134). No causal association between T2DM and HCV was discovered. Among East Asians, however, there was a significant inverse association between T2DM and the proxies of NAFLD (ALT: IVW OR=0.9752, 95%CI 0.9597-0.9909, p=0.0021; AST: IVW OR=0.9673, 95%CI, 0.9528-0.9821, p=1.67e-5), and HCV (IVW: OR=0.9289, 95%CI, 0.8852-0.9747, p=0.0027). Notably, no causal association was found between T2DM and HCC, viral hepatitis, or HBV.

Conclusion

Our MR analysis revealed varying causal associations between T2DM and CLDs in East Asians and Europeans. Further research is required to investigate the potential mechanisms in various ethnic groups, which could yield new insights into early screening and prevention strategies for CLDs in T2DM patients.

Causal relationship between type 1 diabetes mellitus and six high-frequency infectious diseases: A two-sample mendelian randomization study

Purpose

Type 1 diabetes mellitus (T1DM) is associated with different types of infections; however, studies on the causal relationship between T1DM and infectious diseases are lacking. Therefore, our study aimed to explore the causalities between T1DM and six high-frequency infections using a Mendelian randomization (MR) approach.

Methods

Two-sample MR studies were conducted to explore the causalities between T1DM and six high-frequency infections: sepsis, acute lower respiratory infections (ALRIs), intestinal infections (IIs), infections of the genitourinary tract (GUTIs) in pregnancy, infections of the skin and subcutaneous tissues (SSTIs), and urinary tract infections (UTIs). Data on summary statistics for T1DM and infections were obtained from the European Bioinformatics Institute database, the United Kingdom Biobank, FinnGen biobank, and Medical Research Council Integrative Epidemiology Unit. All data obtained for summary statistics were from European countries. The inverse-variance weighted (IVW) method was employed as the main analysis. Considering the multiple comparisons, statistical significance was set at p< 0.008. If univariate MR analyses found a significant causal association, multivariable MR (MVMR) analyses were performed to adjust body mass index (BMI) and glycated hemoglobin (HbA1c). MVMR-IVW was performed as the primary analysis, and the least absolute shrinkage and selection operator (LASSO) regression and MVMR-Robust were performed as complementary analyses.

Results

MR analysis showed that susceptibility to IIs increased in patients with T1DM by 6.09% using the IVW-fixed method [odds ratio (OR)=1.0609; 95% confidence interval (CI): 1.0281-1.0947, p=0.0002]. Results were still significant after multiple testing. Sensitivity analyses did not show any significant horizontal pleiotropy or heterogeneity. After adjusting for BMI and HbA1c, MVMR-IVW (OR=1.0942; 95% CI: 1.0666-1.1224, p<0.0001) showed significant outcomes that were consistent with those of LASSO regression and MVMR-Robust. However, no significant causal relationship was found between T1DM and sepsis susceptibility, ALRI susceptibility, GUTI susceptibility in pregnancy, SSTI susceptibility, and UTI susceptibility.

Conclusions

Our MR analysis genetically predicted increased susceptibility to IIs in T1DM. However, no causality between T1DM and sepsis, ALRIs, GUTIs in pregnancy, SSTIs, or UTIs was found. Larger epidemiological and metagenomic studies are required to further investigate the observed associations between the susceptibility of certain infectious diseases with T1DM.

Human-Machine Interface: Multiclass Classification by Machine Learning on 1D EOG Signals for the Control of an Omnidirectional Robot

People with severe disabilities require assistance to perform their routine activities; a Human-Machine Interface (HMI) will allow them to activate devices that respond according to their needs. In this work, an HMI based on electrooculography (EOG) is presented, the instrumentation is placed on portable glasses that have the task of acquiring both horizontal and vertical EOG signals. The registration of each eye movement is identified by a class and categorized using the one hot encoding technique to test precision and sensitivity of different machine learning classification algorithms capable of identifying new data from the eye registration; the algorithm allows to discriminate blinks in order not to disturb the acquisition of the eyeball position commands. The implementation of the classifier consists of the control of a three-wheeled omnidirectional robot to validate the response of the interface. This work proposes the classification of signals in real time and the customization of the interface, minimizing the user's learning curve. Preliminary results showed that it is possible to generate trajectories to control an omnidirectional robot to implement in the future assistance system to control position through gaze orientation.

Classification of Diseases Using Machine Learning Algorithms: A Comparative Study

Machine learning in the medical area has become a very important requirement. The healthcare professional needs useful tools to diagnose medical illnesses. Classifiers are important to provide tools that can be useful to the health professional for this purpose. However, questions arise: which classifier to use? What metrics are appropriate to measure the performance of the classifier? How to determine a good distribution of the data so that the classifier does not bias the medical patterns to be classified in a particular class? Then most important question: does a classifier perform well for a particular disease? This paper will present some answers to the questions mentioned above, making use of classification algorithms widely used in machine learning research with datasets relating to medical illnesses under the supervised learning scheme. In addition to state-of-the-art algorithms in pattern classification, we introduce a novelty: the use of meta-learning to determine, a priori, which classifier would be the ideal for a specific dataset. The results obtained show numerically and statistically that there are reliable classifiers to suggest medical diagnoses. In addition, we provide some insights about the expected performance of classifiers for such a task.

Cold-Start Problems in Data-Driven Prediction of Drug-Drug Interaction Effects

Combining drugs, a phenomenon often referred to as polypharmacy, can induce additional adverse effects. The identification of adverse combinations is a key task in pharmacovigilance. In this context, in silico approaches based on machine learning are promising as they can learn from a limited number of combinations to predict for all. In this work, we identify various subtasks in predicting effects caused by drug–drug interaction. Predicting drug–drug interaction effects for drugs that already exist is very different from predicting outcomes for newly developed drugs, commonly called a cold-start problem. We propose suitable validation schemes for the different subtasks that emerge. These validation schemes are critical to correctly assess the performance. We develop a new model that obtains AUC-ROC =0.843 for the hardest cold-start task up to AUC-ROC =0.957 for the easiest one on the benchmark dataset of Zitnik et al. Finally, we illustrate how our predictions can be used to improve post-market surveillance systems or detect drug–drug interaction effects earlier during drug development.

A Novel and Simple Mathematical Transform Improves the Perfomance of Lernmatrix in Pattern Classification

The Lernmatrix is a classic associative memory model. The Lernmatrix is capable of executing the pattern classification task, but its performance is not competitive when compared to state-of-the-art classifiers. The main contribution of this paper consists of the proposal of a simple mathematical transform, whose application eliminates the subtractive alterations between patterns. As a consequence, the Lernmatrix performance is significantly improved. To perform the experiments, we selected 20 datasets that are challenging for any classifier, as they exhibit class imbalance. The effectiveness of our proposal was compared against seven supervised classifiers of the most important approaches (Bayes, nearest neighbors, decision trees, logistic function, support vector machines, and neural networks). By choosing balanced accuracy as a performance measure, our proposal obtained the best results in 10 datasets. The elimination of subtractive alterations makes the new model competitive against the best classifiers, and sometimes beats them. After applying the Friedman test and the Holm post hoc test, we can conclude that within a 95% confidence, our proposal competes successfully with the most effective classifiers of the state of the art.

Guiding Mineralization Co-Culture Discovery Using Bayesian Optimization

Many disciplines rely on testing combinations of compounds, materials, proteins, or bacterial species to drive scientific discovery. It is time-consuming and expensive to determine experimentally, via trial-and-error or random selection approaches, which of the many possible combinations will lead to desirable outcomes. Hence, there is a pressing need for more rational and efficient experimental design approaches to reduce experimental effort. In this work, we demonstrate the potential of machine learning methods for the in silico selection of promising co-culture combinations in the application of bioaugmentation. We use the example of pollutant removal in drinking water treatment plants, which can be achieved using co-cultures of a specialized pollutant degrader with combinations of bacterial isolates. To reduce the experimental effort needed to discover high-performing combinations, we propose a data-driven experimental design. Based on a dataset of mineralization performance for all pairs of 13 bacterial species co-cultured with MSH1, we built a Gaussian process regression model to predict the Gompertz mineralization parameters of the co-cultures of two and three species, based on the single-strain parameters. We subsequently used this model in a Bayesian optimization scheme to suggest potentially high-performing combinations of bacteria. We achieved good performance with this approach, both for predicting mineralization parameters and for selecting effective co-cultures, despite the limited dataset. As a novel application of Bayesian optimization in bioremediation, this experimental design approach has promising applications for highlighting co-culture combinations for in vitro testing in various settings, to lessen the experimental burden and perform more targeted screenings.