Iterative random forests to discover predictive and stable high-order interactions

Persistence and heterogeneity of the effects of educating mothers to improve child immunisation uptake: Experimental evidence from Uttar Pradesh in India

Childhood vaccinations are among the most cost-effective health interventions. Yet, in India, where immunisation services are widely available free of charge, a substantial proportion of children remain unvaccinated. We revisit households 30 months after a randomised experiment of a health information intervention designed to educate mothers on the benefits of child vaccination in Uttar Pradesh, India. We find that the large short-term effects on the uptake of diphtheria-pertussis-tetanus and measles vaccination were sustained at 30 months, suggesting the intervention did not simply bring forward vaccinations. We apply causal forests and find that the intervention increased vaccination uptake, but that there was substantial variation in the magnitude of the estimated effects. We conclude that characterising those who benefited most and conversely those who benefited least provides policy-makers with insights on how the intervention worked, and how the targeting of households could be improved.

Epistasis regulates genetic control of cardiac hypertrophy

The combinatorial effect of genetic variants is often assumed to be additive. Although genetic variation can clearly interact non-additively, methods to uncover epistatic relationships remain in their infancy. We develop low-signal signed iterative random forests to elucidate the complex genetic architecture of cardiac hypertrophy. We derive deep learning-based estimates of left ventricular mass from the cardiac MRI scans of 29,661 individuals enrolled in the UK Biobank. We report epistatic genetic variation including variants close to CCDC141 , IGF1R , TTN , and TNKS. Several loci where variants were deemed insignificant in univariate genome-wide association analyses are identified. Functional genomic and integrative enrichment analyses reveal a complex gene regulatory network in which genes mapped from these loci share biological processes and myogenic regulatory factors. Through a network analysis of transcriptomic data from 313 explanted human hearts, we found strong gene co-expression correlations between these statistical epistasis contributors in healthy hearts and a significant connectivity decrease in failing hearts. We assess causality of epistatic effects via RNA silencing of gene-gene interactions in human induced pluripotent stem cell-derived cardiomyocytes. Finally, single-cell morphology analysis using a novel high-throughput microfluidic system shows that cardiomyocyte hypertrophy is non-additively modifiable by specific pairwise interactions between CCDC141 and both TTN and IGF1R . Our results expand the scope of genetic regulation of cardiac structure to epistasis.

Higher-order neurodynamical equation for simplex prediction

It is demonstrated that higher-order patterns beyond pairwise relations can significantly enhance the learning capability of existing graph-based models, and simplex is one of the primary form for graphically representing higher-order patterns. Predicting unknown (disappeared) simplices in real-world complex networks can provide us with deeper insights, thereby assisting us in making better decisions. Nevertheless, previous efforts to predict simplices suffer from two issues: (i) they mainly focus on 2- or 3-simplices, and there are few models available for predicting simplices of arbitrary orders, and (ii) they lack the ability to analyze and learn the features of simplices from the perspective of dynamics. In this paper, we present a Higher-order Neurodynamical Equation for Simplex Prediction of arbitrary order (HNESP), which is a framework that combines neural networks and neurodynamics. Specifically, HNESP simulates the dynamical coupling process of nodes in simplicial complexes through different relations (i.e., strong pairwise relation, weak pairwise relation, and simplex) to learn node-level representations, while explaining the learning mechanism of neural networks from neurodynamics. To enrich the higher-order information contained in simplices, we exploit the entropy and normalized multivariate mutual information of different sub-structures of simplices to acquire simplex-level representations. Furthermore, simplex-level representations and multi-layer perceptron are used to quantify the existence probability of simplices. The effectiveness of HNESP is demonstrated by extensive simulations on seven higher-order benchmarks. Experimental results show that HNESP improves the AUC values of the state-of-the-art baselines by an average of 8.32%. Our implementations will be publicly available at: https://github.com/jianruichen/HNESP.

An experimental manipulation of the value of effort

People who take on challenges and persevere longer are more likely to succeed in life. But individuals often avoid exerting effort, and there is limited experimental research investigating whether we can learn to value effort. We developed a paradigm to test the hypothesis that people can learn to value effort and will seek effortful challenges if directly incentivized to do so. We also dissociate the effects of rewarding people for choosing effortful challenges and performing well. The results provide limited evidence that rewarding effort increased people's willingness to choose harder tasks when rewards were no longer offered (near transfer). There was also mixed evidence that rewarding effort increased willingness to choose harder tasks in another unrelated and unrewarded task (far transfer). These heterogeneous results highlight the need for further research to understand when this paradigm may be the most effective for increasing and generalizing the value of effort.

Reviewing the essential roles of remote phenotyping, GWAS and explainable AI in practical marker-assisted selection for drought-tolerant winter wheat breeding

Marker-assisted selection (MAS) plays a crucial role in crop breeding improving the speed and precision of conventional breeding programmes by quickly and reliably identifying and selecting plants with desired traits. However, the efficacy of MAS depends on several prerequisites, with precise phenotyping being a key aspect of any plant breeding programme. Recent advancements in high-throughput remote phenotyping, facilitated by unmanned aerial vehicles coupled to machine learning, offer a non-destructive and efficient alternative to traditional, time-consuming, and labour-intensive methods. Furthermore, MAS relies on knowledge of marker-trait associations, commonly obtained through genome-wide association studies (GWAS), to understand complex traits such as drought tolerance, including yield components and phenology. However, GWAS has limitations that artificial intelligence (AI) has been shown to partially overcome. Additionally, AI and its explainable variants, which ensure transparency and interpretability, are increasingly being used as recognised problem-solving tools throughout the breeding process. Given these rapid technological advancements, this review provides an overview of state-of-the-art methods and processes underlying each MAS, from phenotyping, genotyping and association analyses to the integration of explainable AI along the entire workflow. In this context, we specifically address the challenges and importance of breeding winter wheat for greater drought tolerance with stable yields, as regional droughts during critical developmental stages pose a threat to winter wheat production. Finally, we explore the transition from scientific progress to practical implementation and discuss ways to bridge the gap between cutting-edge developments and breeders, expediting MAS-based winter wheat breeding for drought tolerance.

Learning epistatic polygenic phenotypes with Boolean interactions

Detecting epistatic drivers of human phenotypes is a considerable challenge. Traditional approaches use regression to sequentially test multiplicative interaction terms involving pairs of genetic variants. For higher-order interactions and genome-wide large-scale data, this strategy is computationally intractable. Moreover, multiplicative terms used in regression modeling may not capture the form of biological interactions. Building on the Predictability, Computability, Stability (PCS) framework, we introduce the epiTree pipeline to extract higher-order interactions from genomic data using tree-based models. The epiTree pipeline first selects a set of variants derived from tissue-specific estimates of gene expression. Next, it uses iterative random forests (iRF) to search training data for candidate Boolean interactions (pairwise and higher-order). We derive significance tests for interactions, based on a stabilized likelihood ratio test, by simulating Boolean tree-structured null (no epistasis) and alternative (epistasis) distributions on hold-out test data. Finally, our pipeline computes PCS epistasis p-values that probabilisticly quantify improvement in prediction accuracy via bootstrap sampling on the test set. We validate the epiTree pipeline in two case studies using data from the UK Biobank: predicting red hair and multiple sclerosis (MS). In the case of predicting red hair, epiTree recovers known epistatic interactions surrounding MC1R and novel interactions, representing non-linearities not captured by logistic regression models. In the case of predicting MS, a more complex phenotype than red hair, epiTree rankings prioritize novel interactions surrounding HLA-DRB1, a variant previously associated with MS in several populations. Taken together, these results highlight the potential for epiTree rankings to help reduce the design space for follow up experiments.

Prenatal metal exposures and childhood gut microbial signatures are associated with depression score in late childhood

BACKGROUND

Childhood depression is a major public health issue worldwide. Previous studies have linked both prenatal metal exposures and the gut microbiome to depression in children. However, few, if any, have studied their interacting effect in specific subgroups of children.

OBJECTIVES

Using an interpretable machine-learning method, this study investigates whether children with specific combinations of prenatal metals and childhood microbial signatures (cliques or groups of metals and microbes) were more likely to have higher depression scores at 9-11 years of age.

METHODS

We leveraged data from a well-characterized pediatric longitudinal birth cohort in Mexico City and its microbiome substudy (n = 112). Eleven metal exposures were measured in maternal whole blood samples in the second and third trimesters of pregnancy. The gut microbial abundances were measured at 9-11-year-olds using shotgun metagenomic sequencing. Depression symptoms were assessed using the Child Depression Index (CDI) t-scores at 9-11 years of age. We used Microbial and Chemical Exposure Analysis (MiCxA), which combines interpretable machine-learning into a regression framework to identify and estimate joint associations of metal-microbial cliques in specific subgroups. Analyses were adjusted for relevant covariates.

RESULTS

We identified a subgroup of children (11.6 % of the sample) characterized by a four-component metal-microbial clique that had a significantly high depression score (15.4 % higher than the rest) in late childhood. This metal-microbial clique consisted of high Zinc in the second trimester, low Cobalt in the third trimester, a high abundance of Bacteroides fragilis, a high abundance of Faecalibacterium prausnitzii. All combinations of cliques (two-, three-, and four-components) were significantly associated with increased log-transformed t-scored CDI (β = 0.14, 95%CI = [0.05,0.23], P < 0.01 for the four-component clique).

SIGNIFICANCE

This study offers a new approach to chemical-microbial analysis and a novel demonstration that children with specific gut microbiome cliques and metal exposures during pregnancy may have a higher likelihood of elevated depression scores.

Akkermansia muciniphila modifies the association between metal exposure during pregnancy and depressive symptoms in late childhood

Emerging research suggests that exposures to metals during pregnancy and gut microbiome (GM) disruptions are associated with depressive disorders in childhood. Akkermansia muciniphila, a GM bacteria, has been studied for its potential antidepressant effects. However, its role in the influence of prenatal metal exposures on depressive symptoms during childhood is unknown. Leveraging a well-characterized pediatric longitudinal birth cohort and its microbiome substudy (n=112) and using a state-of-the-art machine-learning model, we investigated whether the presence of A.muciniphila in GM of 9-11-year-olds modifies the associations between exposure to a specific group of metals (or metal-clique) during pregnancy and concurrent childhood depressive symptoms. Among children with no A.muciniphila, a metal-clique of Zinc-Chromium-Cobalt was strongly associated with increased depression score (P<0.0001), whereas, for children with A.muciniphila, this same metal-clique was weakly associated with decreased depression score(P<0.4). Our analysis provides the first exploratory evidence hypothesizing A. muciniphila as a probiotic intervention attenuating the effect of prenatal metal-exposures-associated depressive disorders in late childhood.

Predicting the spatial variation in cost-efficiency for agricultural greenhouse gas mitigation programs in the U.S

BACKGROUND

Two major factors that determine the efficiency of programs designed to mitigate greenhouse gases by encouraging voluntary changes in U.S. agricultural land management are the effect of land use changes on producers' profitability and the net sequestration those changes create. In this work, we investigate how the interaction of these factors produces spatial heterogeneity in the cost-efficiency of voluntary programs incentivizing tillage reduction and cover-cropping practices. We map county-level predicted rates of adoption for each practice with the greenhouse gas mitigation or carbon sequestration benefits expected from their use. Then, we use these bivariate maps to describe how the cost efficiency of agricultural mitigation efforts is likely to vary spatially in the United States.

RESULTS

Our results suggest the combination of high adoption rates and large reductions in net emissions make reduced tillage programs most cost efficient in the Chesapeake Bay watershed or the Upper Mississippi and Lower Missouri sub-basins of the Mississippi River. For programs aiming to reduce net emissions by incentivizing cover-cropping, we expect cost-efficiency to be greatest in the areas near the main stem of the Mississippi River within its Middle and Lower sections.

CONCLUSIONS

Many voluntary agricultural conservation programs offer the same incentives across the United States. Yet spatial variation in profitability and efficacy of conservation practices suggest that these uniform approaches are not cost-effective. Spatial targeting of voluntary agricultural conservation programs has the potential to increase the cost-efficiency of these programs due to regional heterogeneity in the profitability and greenhouse gas mitigation benefits of agricultural land management practices across the continental United States. We illustrate how predicted rates of adoption and greenhouse gas sequestration might be used to target regions where efforts to incentivize cover-cropping and reductions in tillage are most likely to be cost -effective.

d -Alanine Affects the Circadian Clock to Regulate Glucose Metabolism in the Kidney

Key Points

d -Alanine affects the circadian clock to regulate gluconeogenesis in the kidney. d -Alanine itself has a clear intrinsic circadian rhythm, which is regulated by urinary excretion, and acts on the circadian rhythm. d -Alanine is a signal activator for circadian rhythm and gluconeogenesis through circadian transcriptional network.

Background

The aberrant glucose circadian rhythm is associated with the pathogenesis of diabetes. Similar to glucose metabolism in the kidney and liver, d -alanine, a rare enantiomer of alanine, shows circadian alteration, although the effect of d- alanine on glucose metabolism has not been explored. Here, we show that d- alanine acts on the circadian clock and affects glucose metabolism in the kidney.

Methods

The blood and urinary levels of d -alanine in mice were measured using two-dimensional high-performance liquid chromatography system. Metabolic effects of d -alanine were analyzed in mice and in primary culture of kidney proximal tubular cells from mice. Behavioral and gene expression analyses of circadian rhythm were performed using mice bred under constant darkness.

Results

d- Alanine levels in blood exhibited a clear intrinsic circadian rhythm. Since this rhythm was regulated by the kidney through urinary excretion, we examined the effect of d -alanine on the kidney. In the kidney, d -alanine induced the expressions of genes involved in gluconeogenesis and circadian rhythm. Treatment of d- alanine mediated glucose production in mice. Ex vivo glucose production assay demonstrated that the treatment of d -alanine induced glucose production in primary culture of kidney proximal tubular cells, where d -amino acids are known to be reabsorbed, but not in that of liver cells. Gluconeogenetic effect of d -alanine has an intraday variation, and this effect was in part mediated through circadian transcriptional network. Under constant darkness, treatment of d- alanine normalized the circadian cycle of behavior and kidney gene expressions.

Conclusions

d- Alanine induces gluconeogenesis in the kidney and adjusts the period of the circadian clock. Normalization of circadian cycle by d -alanine may provide the therapeutic options for life style–related diseases and shift workers.

Transcriptome Analysis of Choroidal Endothelium Links Androgen Receptor Role to Central Serous Chorioretinopathy

BACKGROUND

Central Serous Chorioretinopathy (CSCR) manifests as fluid accumulation between the neurosensory retina and the retinal pigment epithelium (RPE). Elevated levels of steroid hormones have been implicated in CSCR pathogenesis. This investigation aims to delineate the gene expression patterns of CSCR-associated risk and steroid receptors across human choroidal cell types and RPE cells to discern potential underlying mechanisms.

METHODS

This study utilized a comprehensive query of transcriptomic data derived from non-pathological human choroid and RPE cells.

FINDINGS

CSCR-associated genes such as PTPRB, CFH, and others are predominantly expressed in the choroidal endothelium as opposed to the RPE. The androgen receptor, encoded by the AR gene, demonstrates heightened expression in the macular endothelium compared to peripheral regions, unlike other steroid receptor genes. AR-expressing endothelial cells display an augmented responsiveness to Transforming growth factor beta (TGF-β), indicating a propensity towards endothelial to mesenchymal transition (endMT) transcriptional profiling.

INTERPRETATION

These results highlight the proclivity of CSCR to manifest primarily within the choroidal vasculature rather than the RPE, suggesting its categorization as a vascular eye disorder. This study accentuates the pivotal role of androgenic steroids, in addition to glucocorticoids. The observed linkage to TGF-β-mediated endMT provides a potential mechanistic insight into the disease's etiology.

Understanding the role of neutral species by means of high-order interaction in the rock-paper-scissors dynamics

The existence of neutral species carries profound ecological implications that warrant further investigation. In this paper, we study the impact of neutral species on biodiversity in a spatial tritrophic system of cyclic competition, in which the neutral species are identified as the fourth species that may affect the competition process of the other three species under the rock-paper-scissors (RPS) rule. Extensive simulations showed that neutral species can promote coexistence in a high mobility regime within the system. When coexistence occurs, we found that the state can be maintained by two mechanisms: Species can either (i) adhere to traditional RPS rule or (ii) form patches to resist invasion. Our findings might aid in understanding the impact of neutral species on biodiversity in ecosystems.

Neural correlates of negative life events and their relationships with alcohol and cannabis use initiation

OBJECTIVE

Negative life events (NLEs), e.g., poor academic performance (controllable) or being the victim of a crime (uncontrollable), can profoundly affect the trajectory of one's life. Yet, their impact on how the brain develops is still not well understood. This investigation examined the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA) dataset for the impact of NLEs on the initiation of alcohol and cannabis use, as well as underlying neural mechanisms.

METHODS

This study evaluated the impact of controllable and uncontrollable NLEs on substance use initiation in 207 youth who initiated alcohol use, 168 who initiated cannabis use, and compared it to 128 youth who remained substance-naïve, using generalised linear regression models. Mediation analyses were conducted to determine neural pathways of NLE impacting substance use trajectories.

RESULTS

Dose-response relationships between controllable NLEs and substance use initiation were observed. Having one controllable NLE increased the odds of alcohol initiation by 50% (95%CI [1.18, 1.93]) and cannabis initiation by 73% (95%CI [1.36, 2.24]), respectively. Greater cortical thickness in left banks of the superior temporal sulcus mediated effects of controllable NLEs on alcohol and cannabis initiations. Greater left caudate gray-matter volumes mediated effects of controllable NLEs on cannabis initiation.

CONCLUSIONS

Controllable but not uncontrollable NLEs increased the odds of alcohol and cannabis initiation. Moreover, those individuals with less mature brain structures at the time of the NLEs experienced a greater impact of NLEs on subsequent initiation of alcohol or cannabis use. Targeting youth experiencing controllable NLEs may help mitigate alcohol and cannabis initiation.

Seminar: Scalable Preprocessing Tools for Exposomic Data Analysis

BACKGROUND

The exposome serves as a popular framework in which to study exposures from chemical and nonchemical stressors across the life course and the differing roles that these exposures can play in human health. As a result, data relevant to the exposome have been used as a resource in the quest to untangle complicated health trajectories and help connect the dots from exposures to adverse outcome pathways.

OBJECTIVES

The primary aim of this methods seminar is to clarify and review preprocessing techniques critical for accurate and effective external exposomic data analysis. Scalability is emphasized through an application of highly innovative combinatorial techniques coupled with more traditional statistical strategies. The Public Health Exposome is used as an archetypical model. The novelty and innovation of this seminar's focus stem from its methodical, comprehensive treatment of preprocessing and its demonstration of the positive effects preprocessing can have on downstream analytics.

DISCUSSION

State-of-the-art technologies are described for data harmonization and to mitigate noise, which can stymie downstream interpretation, and to select key exposomic features, without which analytics may lose focus. A main task is the reduction of multicollinearity, a particularly formidable problem that frequently arises from repeated measurements of similar events taken at various times and from multiple sources. Empirical results highlight the effectiveness of a carefully planned preprocessing workflow as demonstrated in the context of more highly concentrated variable lists, improved correlational distributions, and enhanced downstream analytics for latent relationship discovery. The nascent field of exposome science can be characterized by the need to analyze and interpret a complex confluence of highly inhomogeneous spatial and temporal data, which may present formidable challenges to even the most powerful analytical tools. A systematic approach to preprocessing can therefore provide an essential first step in the application of modern computer and data science methods. https://doi.org/10.1289/EHP12901.

Complex 33-beam simulated galactic cosmic radiation exposure impacts cognitive function and prefrontal cortex neurotransmitter networks in male mice

Astronauts will encounter extended exposure to galactic cosmic radiation (GCR) during deep space exploration, which could impair brain function. Here, we report that in male mice, acute or chronic GCR exposure did not modify reward sensitivity but did adversely affect attentional processes and increased reaction times. Potassium (K+)-stimulation in the prefrontal cortex (PFC) elevated dopamine (DA) but abolished temporal DA responsiveness after acute and chronic GCR exposure. Unlike acute GCR, chronic GCR increased levels of all other neurotransmitters, with differences evident between groups after higher K+-stimulation. Correlational and machine learning analysis showed that acute and chronic GCR exposure differentially reorganized the connection strength and causation of DA and other PFC neurotransmitter networks compared to controls which may explain space radiation-induced neurocognitive deficits.

Machine Learning Assisted Discovery of Interactions between Pesticides, Phthalates, Phenols, and Trace Elements in Child Neurodevelopment

A growing body of literature suggests that developmental exposure to individual or mixtures of environmental chemicals (ECs) is associated with autism spectrum disorder (ASD). However, investigating the effect of interactions among these ECs can be challenging. We introduced a combination of the classical exposure-mixture Weighted Quantile Sum (WQS) regression and a machine-learning method termed Signed iterative Random Forest (SiRF) to discover synergistic interactions between ECs that are (1) associated with higher odds of ASD diagnosis, (2) mimic toxicological interactions, and (3) are present only in a subset of the sample whose chemical concentrations are higher than certain thresholds. In a case-control Childhood Autism Risks from Genetics and Environment (CHARGE) study, we evaluated multiordered synergistic interactions among 62 ECs measured in the urine samples of 479 children in association with increased odds for ASD diagnosis (yes vs no). WQS-SiRF identified two synergistic two-ordered interactions between (1) trace-element cadmium (Cd) and the organophosphate pesticide metabolite diethyl-phosphate (DEP); and (2) 2,4,6-trichlorophenol (TCP-246) and DEP. Both interactions were suggestively associated with increased odds of ASD diagnosis in the subset of children with urinary concentrations of Cd, DEP, and TCP-246 above the 75th percentile. This study demonstrates a novel method that combines the inferential power of WQS and the predictive accuracy of machine-learning algorithms to discover potentially biologically relevant chemical-chemical interactions associated with ASD.

Molecular signature to predict quality of life and survival with glioblastoma using Multiview omics model

Glioblastoma multiforme (GBM) patients show a variety of signs and symptoms that affect their quality of life (QOL) and self-dependence. Since most existing studies have examined prognostic factors based only on clinical factors, there is a need to consider the value of integrating multi-omics data including gene expression and proteomics with clinical data in identifying significant biomarkers for GBM prognosis. Our research aimed to isolate significant features that differentiate between short-term (≤ 6 months) and long-term (≥ 2 years) GBM survival, and between high Karnofsky performance scores (KPS ≥ 80) and low (KPS ≤ 60), using the iterative random forest (iRF) algorithm. Using the Cancer Genomic Atlas (TCGA) database, we identified 35 molecular features composed of 19 genes and 16 proteins. Our findings propose molecular signatures for predicting GBM prognosis and will improve clinical decisions, GBM management, and drug development.

Prenatal Lead Exposure Is Associated with Reduced Abundance of Beneficial Gut Microbial Cliques in Late Childhood: An Investigation Using Microbial Co-Occurrence Analysis (MiCA)

Many analytical methods used in gut microbiome research focus on either single bacterial taxa or the whole microbiome, ignoring multibacteria relationships (microbial cliques). We present a novel analytical approach to identify microbial cliques within the gut microbiome of children at 9-11 years associated with prenatal lead (Pb) exposure. Data came from a subset of participants (n = 123) in the Programming Research in Obesity, Growth, Environment and Social Stressors cohort. Pb concentrations were measured in maternal whole blood from the second and third trimesters of pregnancy. Stool samples collected at 9-11 years old underwent metagenomic sequencing to assess the gut microbiome. Using a novel analytical approach, Microbial Co-occurrence Analysis (MiCA), we paired a machine learning algorithm with randomization-based inference to first identify microbial cliques that were predictive of prenatal Pb exposure and then estimate the association between prenatal Pb exposure and microbial clique abundance. With second-trimester Pb exposure, we identified a two-taxa microbial clique that included Bifidobacterium adolescentis and Ruminococcus callidus and a three-taxa clique that also included Prevotella clara. Increasing second-trimester Pb exposure was associated with significantly increased odds of having the two-taxa microbial clique below the median relative abundance (odds ratio (OR) = 1.03, 95% confidence interval (CI) [1.01-1.05]). Using a novel combination of machine learning and causal inference, MiCA identified a significant association between second-trimester Pb exposure and the reduced abundance of a probiotic microbial clique within the gut microbiome in late childhood.

Quantum biological insights into CRISPR-Cas9 sgRNA efficiency from explainable-AI driven feature engineering

CRISPR-Cas9 tools have transformed genetic manipulation capabilities in the laboratory. Empirical rules-of-thumb have been developed for only a narrow range of model organisms, and mechanistic underpinnings for sgRNA efficiency remain poorly understood. This work establishes a novel feature set and new public resource, produced with quantum chemical tensors, for interpreting and predicting sgRNA efficiency. Feature engineering for sgRNA efficiency is performed using an explainable-artificial intelligence model: iterative Random Forest (iRF). By encoding quantitative attributes of position-specific sequences for Escherichia coli sgRNAs, we identify important traits for sgRNA design in bacterial species. Additionally, we show that expanding positional encoding to quantum descriptors of base-pair, dimer, trimer, and tetramer sequences captures intricate interactions in local and neighboring nucleotides of the target DNA. These features highlight variation in CRISPR-Cas9 sgRNA dynamics between E. coli and H. sapiens genomes. These novel encodings of sgRNAs enhance our understanding of the elaborate quantum biological processes involved in CRISPR-Cas9 machinery.

Four Markers Useful for the Distinction of Intrauterine Growth Restriction in Sheep

Intrauterine growth restriction (IUGR) is a common perinatal complication in animal reproduction, with long-lasting negative effects on neonates and postnatal animals, which seriously negatively affects livestock production. In this study, we aimed to identify potential genes associated with the diagnosis of IUGR through bioinformatics analysis. Based on the 73 differentially expressed related genes obtained by differential analysis and weighted gene co-expression network analysis, we used three machine learning algorithms to identify 4 IUGR-related hub genes (IUGR-HGs), namely, ADAM9, CRYL1, NDP52, and SERPINA7, whose ROC curves showed that they are a good diagnostic target for IUGR. Next, we identified two molecular subtypes of IUGR through consensus clustering analysis and constructed a gene scoring system based on the IUGR-HGs. The results showed that the IUGR score was positively correlated with the risk of IUGR. The AUC value of IUGR scoring accuracy was 0.970. Finally, we constructed a new artificial neural network model based on the four IUGR-HGs to diagnose sheep IUGR, and its accuracy reached 0.956. In conclusion, the IUGR-HGs we identified provide new potential molecular markers and models for the diagnosis of IUGR in sheep; they can better diagnose whether sheep have IUGR. The present findings provide new perspectives on the diagnosis of IUGR.

An interpretable machine learning model of cross-sectional U.S. county-level obesity prevalence using explainable artificial intelligence

BACKGROUND

There is considerable geographic heterogeneity in obesity prevalence across counties in the United States. Machine learning algorithms accurately predict geographic variation in obesity prevalence, but the models are often uninterpretable and viewed as a black-box.

OBJECTIVE

The goal of this study is to extract knowledge from machine learning models for county-level variation in obesity prevalence.

METHODS

This study shows the application of explainable artificial intelligence methods to machine learning models of cross-sectional obesity prevalence data collected from 3,142 counties in the United States. County-level features from 7 broad categories: health outcomes, health behaviors, clinical care, social and economic factors, physical environment, demographics, and severe housing conditions. Explainable methods applied to random forest prediction models include feature importance, accumulated local effects, global surrogate decision tree, and local interpretable model-agnostic explanations.

RESULTS

The results show that machine learning models explained 79% of the variance in obesity prevalence, with physical inactivity, diabetes, and smoking prevalence being the most important factors in predicting obesity prevalence.

CONCLUSIONS

Interpretable machine learning models of health behaviors and outcomes provide substantial insight into obesity prevalence variation across counties in the United States.

Identifying proteomic risk factors for overall, aggressive and early onset prostate cancer using Mendelian randomization and tumor spatial transcriptomics

Background

Understanding the role of circulating proteins in prostate cancer risk can reveal key biological pathways and identify novel targets for cancer prevention.

Methods

We investigated the association of 2,002 genetically predicted circulating protein levels with risk of prostate cancer overall, and of aggressive and early onset disease, using cis-pQTL Mendelian randomization (MR) and colocalization. Findings for proteins with support from both MR, after correction for multiple-testing, and colocalization were replicated using two independent cancer GWAS, one of European and one of African ancestry. Proteins with evidence of prostate-specific tissue expression were additionally investigated using spatial transcriptomic data in prostate tumor tissue to assess their role in tumor aggressiveness. Finally, we mapped risk proteins to drug and ongoing clinical trials targets.

Results

We identified 20 proteins genetically linked to prostate cancer risk (14 for overall [8 specific], 7 for aggressive [3 specific], and 8 for early onset disease [2 specific]), of which a majority were novel and replicated. Among these were proteins associated with aggressive disease, such as PPA2 [Odds Ratio (OR) per 1 SD increment = 2.13, 95% CI: 1.54-2.93], PYY [OR = 1.87, 95% CI: 1.43-2.44] and PRSS3 [OR = 0.80, 95% CI: 0.73-0.89], and those associated with early onset disease, including EHPB1 [OR = 2.89, 95% CI: 1.99-4.21], POGLUT3 [OR = 0.76, 95% CI: 0.67-0.86] and TPM3 [OR = 0.47, 95% CI: 0.34-0.64]. We confirm an inverse association of MSMB with prostate cancer overall [OR = 0.81, 95% CI: 0.80-0.82], and also find an inverse association with both aggressive [OR = 0.84, 95% CI: 0.82-0.86] and early onset disease [OR = 0.71, 95% CI: 0.68-0.74]. Using spatial transcriptomics data, we identified MSMB as the genome-wide top-most predictive gene to distinguish benign regions from high grade cancer regions that had five-fold lower MSMB expression. Additionally, ten proteins that were associated with prostate cancer risk mapped to existing therapeutic interventions.

Conclusion

Our findings emphasize the importance of proteomics for improving our understanding of prostate cancer etiology and of opportunities for novel therapeutic interventions. Additionally, we demonstrate the added benefit of in-depth functional analyses to triangulate the role of risk proteins in the clinical aggressiveness of prostate tumors. Using these integrated methods, we identify a subset of risk proteins associated with aggressive and early onset disease as priorities for investigation for the future prevention and treatment of prostate cancer.

A blood-based metabolomic signature predictive of risk for pancreatic cancer

Emerging evidence implicates microbiome involvement in the development of pancreatic cancer (PaCa). Here, we investigate whether increases in circulating microbial-related metabolites associate with PaCa risk by applying metabolomics profiling to 172 sera collected within 5 years prior to PaCa diagnosis and 863 matched non-subject sera from participants in the Prostate, Lung, Colorectal, and Ovarian (PLCO) cohort. We develop a three-marker microbial-related metabolite panel to assess 5-year risk of PaCa. The addition of five non-microbial metabolites further improves 5-year risk prediction of PaCa. The combined metabolite panel complements CA19-9, and individuals with a combined metabolite panel + CA19-9 score in the top 2.5th percentile have absolute 5-year risk estimates of >13%. The risk prediction model based on circulating microbial and non-microbial metabolites provides a potential tool to identify individuals at high risk of PaCa that would benefit from surveillance and/or from potential cancer interception strategies.

An Oral Microbial Biomarker for Early Detection of Recurrence of Oral Squamous Cell Carcinoma

Changes in the oral microbiome are associated with oral squamous cell carcinoma (OSCC). Oral microbe-derived signatures have been utilized as markers of OSCC. However, the structure of the oral microbiome during OSCC recurrence and biomarkers for the prediction of OSCC recurrence remains unknown. To identify OSCC recurrence-associated microbial biomarkers for the prediction of OSCC recurrence, we performed 16S rRNA amplicon sequencing on 54 oral swab samples from OSCC patients. Differences in bacterial compositions were observed in patients with vs without recurrence. We found that Granulicatella, Peptostreptococcus, Campylobacter, Porphyromonas, Oribacterium, Actinomyces, Corynebacterium, Capnocytophaga, and Dialister were enriched in OSCC recurrence. Functional analysis of the oral microbiome showed altered functions associated with OSCC recurrence compared with nonrecurrence. A random forest prediction model was constructed with five microbial signatures including Leptotrichia trevisanii, Capnocytophaga sputigena, Capnocytophaga, Cardiobacterium, and Olsenella to discriminate OSCC recurrence from original OSCC (accuracy = 0.963). Moreover, we validated the prediction model in another independent cohort (46 OSCC patients), achieving an accuracy of 0.761. We compared the accuracy of the prediction of OSCC recurrence between the five microbial signatures and two clinicopathological parameters, including resection margin and lymph node counts. The results predicted by the model with five microbial signatures showed a higher accuracy than those based on the clinical outcomes from the two clinicopathological parameters. This study demonstrated the validity of using recurrence-related microbial biomarkers, a noninvasive and effective method for the prediction of OSCC recurrence. Our findings may contribute to the prognosis and treatment of OSCC recurrence.

Application of machine learning to discover interactions predictive of dietary lapses

The purpose of this study it to build a machine learning model to predict dietary lapses with comparable accuracy, sensitivity, and specificity to previous literature while recovering predictor interactions. The sample for the current study consisted of merged data from two separate studies of individuals with obesity/overweight (total N = 87). Participants completed six ecological momentary assessment surveys per day where they were asked about 16 risk factors of lapse and if they had lapsed from their dietary prescriptions since the previous survey. Alcohol consumption and self-efficacy were the most prevalent in the top 10 stable interactions. Alcohol consumption decreased the protective effect of self-efficacy, motivation, and planning. Higher planning predicted higher risk for lapse only when consuming alcohol. Low motivation, hunger, cravings, and lack of healthy food availability increased the protective effect of self-efficacy. Higher self-efficacy increased risk effect of positive mood and having recently eaten a meal on lapse. For individuals with lower levels of self-efficacy, planning increased the risk of lapse. Alcohol intake and self-efficacy interact with several variables to predict dietary lapses, and these interactions should be targeted in just-in-time adaptive interventions that deliver interventions for lapses.

Using iterative random forest to find geospatial environmental and Sociodemographic predictors of suicide attempts

Introduction

Despite a recent global decrease in suicide rates, death by suicide has increased in the United States. It is therefore imperative to identify the risk factors associated with suicide attempts to combat this growing epidemic. In this study, we aim to identify potential risk factors of suicide attempt using geospatial features in an Artificial intelligence framework.

Methods

We use iterative Random Forest, an explainable artificial intelligence method, to predict suicide attempts using data from the Million Veteran Program. This cohort incorporated 405,540 patients with 391,409 controls and 14,131 attempts. Our predictive model incorporates multiple climatic features at ZIP-code-level geospatial resolution. We additionally consider demographic features from the American Community Survey as well as the number of firearms and alcohol vendors per 10,000 people to assess the contributions of proximal environment, access to means, and restraint decrease to suicide attempts. In total 1,784 features were included in the predictive model.

Results

Our results show that geographic areas with higher concentrations of married males living with spouses are predictive of lower rates of suicide attempts, whereas geographic areas where males are more likely to live alone and to rent housing are predictive of higher rates of suicide attempts. We also identified climatic features that were associated with suicide attempt risk by age group. Additionally, we observed that firearms and alcohol vendors were associated with increased risk for suicide attempts irrespective of the age group examined, but that their effects were small in comparison to the top features.

Discussion

Taken together, our findings highlight the importance of social determinants and environmental factors in understanding suicide risk among veterans.

Association between Sexual Activity during Pregnancy, Pre- and Early-Term Birth, and Vaginal Cytokine Inflammation: A Prospective Study of Black Women

This study aimed to investigate the association between sexual activity during pregnancy and adverse birth outcomes among Black women, and to explore whether vaginal cytokine inflammation mediates this association. Data from 397 Black pregnant women through questionnaires on sexual activity and vaginal biosamples during early (8-14 weeks) and late (24-30 weeks) pregnancy, and birth outcomes were analyzed. Using a data-driven approach, the study found that vaginal sex during late pregnancy was associated with spontaneous early-term birth (sETB, 38-39 completed weeks' gestation) (OR = 0.39, 95% CI: [0.21, 0.72], p-value = 0.003) but not with spontaneous preterm birth (sPTB) (OR = 1.08, p-value = 0.86) compared to full-term birth. Overall, despite vaginal sex in late pregnancy showing an overall positive effect on sETB (total effect = -0.1580, p-value = 0.015), we observed a negative effect of vaginal sex on sETB (indirect effect = 0.0313, p-value = 0.026) due to the fact that having vaginal sex could lead to elevated IL6 levels, which in turn increased the odds of sETB. In conclusion, the study found an overall positive association between sexual activity on ETB and a negative partial mediation effect via increased vaginal cytokine inflammation induced by vaginal sexual activity. This inconsistent mediation model suggested that vaginal sexual activity is a complex behavior that could have both positive and negative effects on the birth outcome.

Machine learning in bioprocess development: from promise to practice

Fostered by novel analytical techniques, digitalization, and automation, modern bioprocess development provides large amounts of heterogeneous experimental data, containing valuable process information. In this context, data-driven methods like machine learning (ML) approaches have great potential to rationally explore large design spaces while exploiting experimental facilities most efficiently. Herein we demonstrate how ML methods have been applied so far in bioprocess development, especially in strain engineering and selection, bioprocess optimization, scale-up, monitoring, and control of bioprocesses. For each topic, we will highlight successful application cases, current challenges, and point out domains that can potentially benefit from technology transfer and further progress in the field of ML.

Prenatal Pb exposure is associated with reduced abundance of beneficial gut microbial cliques in late childhood: an investigation using Microbial Co-occurrence Analysis (MiCA)

Background

Many analytical methods used in gut microbiome research focus on either single bacterial taxa or the whole microbiome, ignoring multi-bacteria relationships (microbial cliques). We present a novel analytical approach to identify multiple bacterial taxa within the gut microbiome of children at 9-11 years associated with prenatal Pb exposure.

Methods

Data came from a subset of participants (n=123) in the Programming Research in Obesity, Growth, Environment and Social Stressors (PROGRESS) cohort. Pb concentrations were measured in maternal whole blood from the second and third trimesters of pregnancy. Stool samples collected at 9-11 years old underwent metagenomic sequencing to assess the gut microbiome. Using a novel analytical approach, Microbial Co-occurrence Analysis (MiCA), we paired a machine-learning algorithm with randomization-based inference to first identify microbial cliques that were predictive of prenatal Pb exposure and then estimate the association between prenatal Pb exposure and microbial clique abundance.

Results

With second-trimester Pb exposure, we identified a 2-taxa microbial clique that included Bifidobacterium adolescentis and Ruminococcus callidus, and a 3-taxa clique that added Prevotella clara. Increasing second-trimester Pb exposure was associated with significantly increased odds of having the 2-taxa microbial clique below the 50th percentile relative abundance (OR=1.03,95%CI[1.01-1.05]). In an analysis of Pb concentration at or above vs. below the United States and Mexico guidelines for child Pb exposure, odds of the 2-taxa clique in low abundance were 3.36(95%CI[1.32-8.51]) and 6.11(95%CI[1.87-19.93]), respectively. Trends were similar with the 3-taxa clique but not statistically significant.

Discussion

Using a novel combination of machine-learning and causal-inference, MiCA identified a significant association between second-trimester Pb exposure and reduced abundance of a probiotic microbial clique within the gut microbiome in late childhood. Pb exposure levels at the guidelines for child Pb poisoning in the United States, and Mexico are not sufficient to protect against the potential loss of probiotic benefits.

The Impact of Stability Considerations on Genetic Fine-Mapping

Fine-mapping methods, which aim to identify genetic variants responsible for complex traits following genetic association studies, typically assume that sufficient adjustments for confounding within the association study cohort have been made, e.g., through regressing out the top principal components (i.e., residualization). Despite its widespread use, however, residualization may not completely remove all sources of confounding. Here, we propose a complementary stability-guided approach that does not rely on residualization, which identifies consistently fine-mapped variants across different genetic backgrounds or environments. We demonstrate the utility of this approach by applying it to fine-map eQTLs in the GEUVADIS data. Using 378 different functional annotations of the human genome, including recent deep learning-based annotations (e.g., Enformer), we compare enrichments of these annotations among variants for which the stability and traditional residualization-based fine-mapping approaches agree against those for which they disagree, and find that the stability approach enhances the power of traditional fine-mapping methods in identifying variants with functional impact. Finally, in cases where the two approaches report distinct variants, our approach identifies variants comparably enriched for functional annotations. Our findings suggest that the stability principle, as a conceptually simple device, complements existing approaches to fine-mapping, reinforcing recent advocacy of evaluating cross-population and cross-environment portability of biological findings. To support visualization and interpretation of our results, we provide a Shiny app, available at: https://alan-aw.shinyapps.io/stability_v0/.

Selection of Optimal Cell Lines for High-Content Phenotypic Screening

High-content microscopy offers a scalable approach to screen against multiple targets in a single pass. Prior work has focused on methods to select "optimal" cellular readouts in microscopy screens. However, methods to select optimal cell line models have garnered much less attention. Here, we provide a roadmap for how to select the cell line or lines that are best suited to identify bioactive compounds and their mechanism of action (MOA). We test our approach on compounds targeting cancer-relevant pathways, ranking cell lines in two tasks: detecting compound activity ("phenoactivity") and grouping compounds with similar MOA by similar phenotype ("phenosimilarity"). Evaluating six cell lines across 3214 well-annotated compounds, we show that optimal cell line selection depends on both the task of interest (e.g., detecting phenoactivity vs inferring phenosimilarity) and distribution of MOAs within the compound library. Given a task of interest and a set of compounds, we provide a systematic framework for choosing optimal cell line(s). Our framework can be used to reduce the number of cell lines required to identify hits within a compound library and help accelerate the pace of early drug discovery.

Learning high-order interactions for polygenic risk prediction

Within the framework of precision medicine, the stratification of individual genetic susceptibility based on inherited DNA variation has paramount relevance. However, one of the most relevant pitfalls of traditional Polygenic Risk Scores (PRS) approaches is their inability to model complex high-order non-linear SNP-SNP interactions and their effect on the phenotype (e.g. epistasis). Indeed, they incur in a computational challenge as the number of possible interactions grows exponentially with the number of SNPs considered, affecting the statistical reliability of the model parameters as well. In this work, we address this issue by proposing a novel PRS approach, called High-order Interactions-aware Polygenic Risk Score (hiPRS), that incorporates high-order interactions in modeling polygenic risk. The latter combines an interaction search routine based on frequent itemsets mining and a novel interaction selection algorithm based on Mutual Information, to construct a simple and interpretable weighted model of user-specified dimensionality that can predict a given binary phenotype. Compared to traditional PRSs methods, hiPRS does not rely on GWAS summary statistics nor any external information. Moreover, hiPRS differs from Machine Learning-based approaches that can include complex interactions in that it provides a readable and interpretable model and it is able to control overfitting, even on small samples. In the present work we demonstrate through a comprehensive simulation study the superior performance of hiPRS w.r.t. state of the art methods, both in terms of scoring performance and interpretability of the resulting model. We also test hiPRS against small sample size, class imbalance and the presence of noise, showcasing its robustness to extreme experimental settings. Finally, we apply hiPRS to a case study on real data from DACHS cohort, defining an interaction-aware scoring model to predict mortality of stage II-III Colon-Rectal Cancer patients treated with oxaliplatin.

NSPA: characterizing the disease association of multiple genetic interactions at single-subject resolution

Motivation

The interaction between genetic variables is one of the major barriers to characterizing the genetic architecture of complex traits. To consider epistasis, network science approaches are increasingly being used in research to elucidate the genetic architecture of complex diseases. Network science approaches associate genetic variables' disease susceptibility to their topological importance in the network. However, this network only represents genetic interactions and does not describe how these interactions attribute to disease association at the subject-scale. We propose the Network-based Subject Portrait Approach (NSPA) and an accompanying feature transformation method to determine the collective risk impact of multiple genetic interactions for each subject.

Results

The feature transformation method converts genetic variants of subjects into new values that capture how genetic variables interact with others to attribute to a subject's disease association. We apply this approach to synthetic and genetic datasets and learn that (1) the disease association can be captured using multiple disjoint sets of genetic interactions and (2) the feature transformation method based on NSPA improves predictive performance comparing with using the original genetic variables. Our findings confirm the role of genetic interaction in complex disease and provide a novel approach for gene-disease association studies to identify genetic architecture in the context of epistasis.

Availability and implementation

The codes of NSPA are now available in: https://github.com/MIB-Lab/Network-based-Subject-Portrait-Approach.

Contact

ting.hu@queensu.ca.

Supplementary information

Supplementary data are available at Bioinformatics Advances online.

A primer on artificial intelligence in plant digital phenomics: embarking on the data to insights journey

Artificial intelligence (AI) has emerged as a fundamental component of global agricultural research that is poised to impact on many aspects of plant science. In digital phenomics, AI is capable of learning intricate structure and patterns in large datasets. We provide a perspective and primer on AI applications to phenome research. We propose a novel human-centric explainable AI (X-AI) system architecture consisting of data architecture, technology infrastructure, and AI architecture design. We clarify the difference between post hoc models and 'interpretable by design' models. We include guidance for effectively using an interpretable by design model in phenomic analysis. We also provide directions to sources of tools and resources for making data analytics increasingly accessible. This primer is accompanied by an interactive online tutorial.

Selection of optimal cell lines for high-content phenotypic screening

High-content microscopy offers a scalable approach to screen against multiple targets in a single pass. Prior work has focused on methods to select "optimal" cellular readouts in microscopy screens. However, methods to select optimal cell line models have garnered much less attention. Here, we provide a roadmap for how to select the cell line or lines that are best suited to identify bioactive compounds and their mechanism of action (MOA). We test our approach on compounds targeting cancer-relevant pathways, ranking cell lines in two tasks: detecting compound activity ("phenoactivity") and grouping compounds with similar MOA by similar phenotype ("phenosimilarity"). Evaluating six cell lines across 3214 well-annotated compounds, we show that optimal cell line selection depends on both the task of interest (e.g. detecting phenoactivity vs. inferring phenosimilarity) and distribution of MOAs within the compound library. Given a task of interest and set of compounds, we provide a systematic framework for choosing optimal cell line(s). Our framework can be used to reduce the number of cell lines required to identify hits within a compound library and help accelerate the pace of early drug discovery.

Interpreting tree ensemble machine learning models with endoR

Tree ensemble machine learning models are increasingly used in microbiome science as they are compatible with the compositional, high-dimensional, and sparse structure of sequence-based microbiome data. While such models are often good at predicting phenotypes based on microbiome data, they only yield limited insights into how microbial taxa may be associated. We developed endoR, a method to interpret tree ensemble models. First, endoR simplifies the fitted model into a decision ensemble. Then, it extracts information on the importance of individual features and their pairwise interactions, displaying them as an interpretable network. Both the endoR network and importance scores provide insights into how features, and interactions between them, contribute to the predictive performance of the fitted model. Adjustable regularization and bootstrapping help reduce the complexity and ensure that only essential parts of the model are retained. We assessed endoR on both simulated and real metagenomic data. We found endoR to have comparable accuracy to other common approaches while easing and enhancing model interpretation. Using endoR, we also confirmed published results on gut microbiome differences between cirrhotic and healthy individuals. Finally, we utilized endoR to explore associations between human gut methanogens and microbiome components. Indeed, these hydrogen consumers are expected to interact with fermenting bacteria in a complex syntrophic network. Specifically, we analyzed a global metagenome dataset of 2203 individuals and confirmed the previously reported association between Methanobacteriaceae and Christensenellales. Additionally, we observed that Methanobacteriaceae are associated with a network of hydrogen-producing bacteria. Our method accurately captures how tree ensembles use features and interactions between them to predict a response. As demonstrated by our applications, the resultant visualizations and summary outputs facilitate model interpretation and enable the generation of novel hypotheses about complex systems.

Comparison of machine learning methods with logistic regression analysis in creating predictive models for risk of critical in-hospital events in COVID-19 patients on hospital admission

BACKGROUND

Machine learning (ML) algorithms have been trained to early predict critical in-hospital events from COVID-19 using patient data at admission, but little is known on how their performance compares with each other and/or with statistical logistic regression (LR). This prospective multicentre cohort study compares the performance of a LR and five ML models on the contribution of influencing predictors and predictor-to-event relationships on prediction model´s performance.

METHODS

We used 25 baseline variables of 490 COVID-19 patients admitted to 8 hospitals in Germany (March-November 2020) to develop and validate (75/25 random-split) 3 linear (L1 and L2 penalty, elastic net [EN]) and 2 non-linear (support vector machine [SVM] with radial kernel, random forest [RF]) ML approaches for predicting critical events defined by intensive care unit transfer, invasive ventilation and/or death (composite end-point: 181 patients). Models were compared for performance (area-under-the-receiver-operating characteristic-curve [AUC], Brier score) and predictor importance (performance-loss metrics, partial-dependence profiles).

RESULTS

Models performed close with a small benefit for LR (utilizing restricted cubic splines for non-linearity) and RF (AUC means: 0.763-0.731 [RF-L1]); Brier scores: 0.184-0.197 [LR-L1]). Top ranked predictor variables (consistently highest importance: C-reactive protein) were largely identical across models, except creatinine, which exhibited marginal (L1, L2, EN, SVM) or high/non-linear effects (LR, RF) on events.

CONCLUSIONS

Although the LR and ML models analysed showed no strong differences in performance and the most influencing predictors for COVID-19-related event prediction, our results indicate a predictive benefit from taking account for non-linear predictor-to-event relationships and effects. Future efforts should focus on leveraging data-driven ML technologies from static towards dynamic modelling solutions that continuously learn and adapt to changes in data environments during the evolving pandemic.

TRIAL REGISTRATION NUMBER

NCT04659187.

Structural variants identified using non-Mendelian inheritance patterns advance the mechanistic understanding of autism spectrum disorder

The heritability of autism spectrum disorder (ASD), based on 680,000 families and five countries, is estimated to be nearly 80%, yet heritability reported from SNP-based studies are consistently lower, and few significant loci have been identified with genome-wide association studies. This gap in genomic information may reside in rare variants, interaction among variants (epistasis), or cryptic structural variation (SV) and may provide mechanisms that underlie ASD. Here we use a method to identify potential SVs based on non-Mendelian inheritance patterns in pedigrees using parent-child genotypes from ASD families and demonstrate that they are enriched in ASD-risk genes. Most are in non-coding genic space and are over-represented in expression quantitative trait loci, suggesting that they affect gene regulation, which we confirm with their overlap of differentially expressed genes in postmortem brain tissue of ASD individuals. We then identify an SV in the GRIK2 gene that alters RNA splicing and a regulatory region of the ACMSD gene in the kynurenine pathway as significantly associated with a non-verbal ASD phenotype, supporting our hypothesis that these currently excluded loci can provide a clearer mechanistic understanding of ASD. Finally, we use an explainable artificial intelligence approach to define subgroups demonstrating their use in the context of precision medicine.

Soil Metabolomics Predict Microbial Taxa as Biomarkers of Moisture Status in Soils from a Tidal Wetland

We present observations from a laboratory-controlled study on the impacts of extreme wetting and drying on a wetland soil microbiome. Our approach was to experimentally challenge the soil microbiome to understand impacts on anaerobic carbon cycling processes as the system transitions from dryness to saturation and vice-versa. Specifically, we tested for impacts on stress responses related to shifts from wet to drought conditions. We used a combination of high-resolution data for small organic chemical compounds (metabolites) and biological (community structure based on 16S rRNA gene sequencing) features. Using a robust correlation-independent data approach, we further tested the predictive power of soil metabolites for the presence or absence of taxa. Here, we demonstrate that taking an untargeted, multidimensional data approach to the interpretation of metabolomics has the potential to indicate the causative pathways selecting for the observed bacterial community structure in soils.

Characteristics of Microbial Distribution in Different Oral Niches of Oral Squamous Cell Carcinoma

Oral squamous cell carcinoma (OSCC), one of the most common malignant tumors of the head and neck, is closely associated with the presence of oral microbes. However, the microbiomes of different oral niches in OSCC patients and their association with OSCC have not been adequately characterized. In this study, 305 samples were collected from 65 OSCC patients, including tumor tissue, adjacent normal tissue (paracancerous tissue), cancer surface tissue, anatomically matched contralateral normal mucosa, saliva, and tongue coat. 16S ribosomal DNA (16S rDNA) sequencing was used to compare the microbial composition, distribution, and co-occurrence network of different oral niches. The association between the microbiome and the clinical features of OSCC was also characterized. The oral microbiome of OSCC patients showed a regular ecological distribution. Tumor and paracancerous tissues were more microbially diverse than other oral niches. Cancer surface, contralateral normal mucosa, saliva, and tongue coat showed similar microbial compositions, especially the contralateral normal mucosa and saliva. Periodontitis-associated bacteria of the genera Fusobacterium, Prevotella, Porphyromonas, Campylobacter, and Aggregatibacter, and anaerobic bacteria were enriched in tumor samples. The microbiome was highly correlated with tumor clinicopathological features, with several genera (Lautropia, Asteroleplasma, Parvimonas, Peptostreptococcus, Pyramidobacter, Roseburia, and Propionibacterium) demonstrating a relatively high diagnostic power for OSCC metastasis, potentially providing an indicator for the development of OSCC.

Predictability and stability testing to assess clinical decision instrument performance for children after blunt torso trauma

OBJECTIVE

The Pediatric Emergency Care Applied Research Network (PECARN) has developed a clinical-decision instrument (CDI) to identify children at very low risk of intra-abdominal injury. However, the CDI has not been externally validated. We sought to vet the PECARN CDI with the Predictability Computability Stability (PCS) data science framework, potentially increasing its chance of a successful external validation.

MATERIALS & METHODS

We performed a secondary analysis of two prospectively collected datasets: PECARN (12,044 children from 20 emergency departments) and an independent external validation dataset from the Pediatric Surgical Research Collaborative (PedSRC; 2,188 children from 14 emergency departments). We used PCS to reanalyze the original PECARN CDI along with new interpretable PCS CDIs developed using the PECARN dataset. External validation was then measured on the PedSRC dataset.

RESULTS

Three predictor variables (abdominal wall trauma, Glasgow Coma Scale Score <14, and abdominal tenderness) were found to be stable. A CDI using only these three variables would achieve lower sensitivity than the original PECARN CDI with seven variables on internal PECARN validation but achieve the same performance on external PedSRC validation (sensitivity 96.8% and specificity 44%). Using only these variables, we developed a PCS CDI which had a lower sensitivity than the original PECARN CDI on internal PECARN validation but performed the same on external PedSRC validation (sensitivity 96.8% and specificity 44%).

CONCLUSION

The PCS data science framework vetted the PECARN CDI and its constituent predictor variables prior to external validation. We found that the 3 stable predictor variables represented all of the PECARN CDI's predictive performance on independent external validation. The PCS framework offers a less resource-intensive method than prospective validation to vet CDIs before external validation. We also found that the PECARN CDI will generalize well to new populations and should be prospectively externally validated. The PCS framework offers a potential strategy to increase the chance of a successful (costly) prospective validation.

Using Explainable Artificial Intelligence to Discover Interactions in an Ecological Model for Obesity

Ecological theories suggest that environmental, social, and individual factors interact to cause obesity. Yet, many analytic techniques, such as multilevel modeling, require manual specification of interacting factors, making them inept in their ability to search for interactions. This paper shows evidence that an explainable artificial intelligence approach, commonly employed in genomics research, can address this problem. The method entails using random intersection trees to decode interactions learned by random forest models. Here, this approach is used to extract interactions between features of a multi-level environment from random forest models of waist-to-height ratios using 11,112 participants from the Adolescent Brain Cognitive Development study. This study shows that methods used to discover interactions between genes can also discover interacting features of the environment that impact obesity. This new approach to modeling ecosystems may help shine a spotlight on combinations of environmental features that are important to obesity, as well as other health outcomes.

Correlates of past year suicidal thoughts among sexual and gender minority young adults: A machine learning analysis

Sexual and gender minority populations are at elevated risk of experiencing suicidal thoughts and attempting suicide. The COVID-19 pandemic exacerbated mental health and substance use challenges among this population. We aimed to examine the relative importance and effects of intersectional factors and strong interactions associated with the risk of suicidal thoughts among Canadian lesbian, gay, bisexual, transgender, queer, questioning, intersex and Two Spirit (LGBTQI2S+) young adults. A cross-sectional online survey was conducted among LGBTQI2S + participants aged 16-29 years living in two Canadian provinces (Ontario, Quebec). Among 1414 participants (mean age 21.90 years), 61% (n = 857) participants reported suicidal thoughts in last 12 months. We built a random forest model to predict the risk of having past year suicidal thoughts, which achieved high performance with an area under the receiver operating characteristic curve (AUC) of 0.84. The top 10 correlates identified were: seeking help from health professionals for mental health or substance use issues since the start of the pandemic, current self-rated mental health status, insulted by parents or adults in childhood, ever heard that being identifying as LGBTQI2S+ is not normal, age in years, past week feeling depressed, lifetime diagnosis of mental illness, lifetime diagnosis of depressive disorder, past week feeling sad, ever pretended to be straight or cisgender to be accepted. The increase in the risk of suicidal thoughts for those having mental health challenges or facing minority stressors is more pronounced in those living in urban areas or being unemployed than those living in rural areas or being employed.

Incorporation of Data-mined Knowledge into Black-box SVM for Interpretability

The lack of interpretability often makes black-box models challenging to be applied in many practical domains. For this reason, the current work, from the black-box model input port, proposes to incorporate data-mined knowledge into the black-box soft-margin SVM model to enhance accuracy and interpretability. The concept and incorporation mechanism of data-mined knowledge are successively developed, based on which a partially interpretable soft-margin SVM ( pTsm -SVM) optimization model is designed and then solved through reformulating the optimization problem as standard quadratic programming. An algorithm for mining linear positive (negative) class knowledge from general data sets is also proposed, which generates a linear two-dimensional discriminative rule with specificity (sensitivity) equal to 1 and the highest possible sensitivity (specificity) among all two-dimensional feature spaces. The knowledge-integrated pTsm -SVM works by achieving a good trade-off among the “large margin”, “high specificity”, and “high sensitivity”. Our experimental results on eight UCI datasets demonstrate the superiority of the proposed pTsm -SVM over the standard soft-margin SVM both in terms of accuracy and interpretability.

Nontargeted metabolomics-based multiple machine learning modeling boosts early accurate detection for citrus Huanglongbing

Early accurate detection of crop disease is extremely important for timely disease management. Huanglongbing (HLB), one of the most destructive citrus diseases, has brought about severe economic losses for the global citrus industry. The direct strategies for HLB identification, such as quantitative real-time polymerase chain reaction (qPCR) and chemical staining, are robust for the symptomatic plants but powerless for the asymptomatic ones at the early stage of affection. Thus, it is very necessary to develop a practical method used for the early detection of HLB. In this study, a novel method combining ultra-high performance liquid chromatography/mass spectrometry (UHPLC/MS)-based nontargeted metabolomics and machine learning (ML) was developed for conducting the early detection of HLB for the first time. Six ML algorithms were selected to build the classifiers. Regularized logistic regression (LR-L2) and gradient-boosted decision tree (GBDT) outperformed with the highest average accuracy of 95.83% to not only classify healthy and infected plants but identify significant features. The proposed method proved to be practical for early detection of HLB, which tackled the shortcomings of low sensitivity in the conventional methods and avoid the problems such as lighting condition interference in spectrum/image recognition-based ML methods. Additionally, the discovered biomarkers were verified by the metabolic pathway analysis and content change analysis, which was remarkably consistent with the previous reports.

Evaluating the Performance of Random Forest and Iterative Random Forest Based Methods when Applied to Gene Expression Data

Gene-to-gene networks, such as Gene Regulatory Networks (GRN) and Predictive Expression Networks (PEN) capture relationships between genes and are beneficial for use in downstream biological analyses. There exists multiple network inference tools to produce these gene-to-gene networks from matrices of gene expression data. Random Forest-Leave One Out Prediction (RF-LOOP) is a method that has been shown to be efficient at producing these gene-to-gene networks, frequently known as GEne Network Inference with Ensemble of trees (GENIE3). Random Forest can be replaced in this process by iterative Random Forest (iRF), which performs variable selection and boosting. Here we validate that iterative Random Forest-Leave One Out Prediction (iRF-LOOP) produces higher quality networks than GENIE3 (RF-LOOP). We use both synthetic and empirical networks from the Dialogue for Reverse Engineering Assessment and Methods (DREAM) Challenges by Sage Bionetworks, as well as two additional empirical networks created from Arabidopsis thaliana and Populus trichocarpa expression data.

Provable Boolean interaction recovery from tree ensemble obtained via random forests

SignificanceRandom Forests (RFs) are among the most successful machine-learning algorithms in terms of prediction accuracy. In many domain problems, however, the primary goal is not prediction, but to understand the data-generation process-in particular, finding important features and feature interactions. There exists strong empirical evidence that RF-based methods-in particular, iterative RF (iRF)-are very successful in terms of detecting feature interactions. In this work, we propose a biologically motivated, Boolean interaction model. Using this model, we complement the existing empirical evidence with theoretical evidence for the ability of iRF-type methods to select desirable interactions. Our theoretical analysis also yields deeper insights into the general interaction selection mechanism of decision-tree algorithms and the importance of feature subsampling.