Applications of random forest feature selection for fine-scale genetic population assignment

Radiomics based on multiple machine learning methods for diagnosing early bone metastases not visible on CT images

OBJECTIVES

This study utilizes [99mTc]-methylene diphosphate (MDP) single photon emission computed tomography (SPECT) images as a reference standard to evaluate whether the integration of radiomics features from computed tomography (CT) and machine learning algorithms can identify microscopic early bone metastases. Additionally, we also determine the optimal machine learning approach.

MATERIALS AND METHODS

We retrospectively studied 63 patients with early bone metastasis from July 2020 to March 2023. The ITK-SNAP software was used to delineate early bone metastases and normal bone tissue in SPECT images of each patient, which were then registered onto CT images to outline the volume of interest (VOI). The VOI includes 63 early bone metastasis volumes and 63 normal bone tissue volumes. 126 VOIs were randomly distributed in a 7:3 ratio between the training and testing groups, and 944 radiomics features were extracted from every VOI. We established 20 machine learning models using 5 feature selection algorithms and 4 classification methods. Evaluate the performance of the model using the area under the receiver operating characteristic curve (AUC).

RESULTS

Most machine learning models demonstrated outstanding discriminative capacity, with AUCs higher than 0.70. Notably, the K-Nearest Neighbors (KNN) classifier exhibited significant performance improvement compared to the other four classifiers. Specifically, the model constructed utilizing eXtreme Gradient Boosting (XGBoost) feature selection method integrated with KNN classifier achieved the maximum AUC, which is 0.989 in the training set and 0.975 in the testing set.

CONCLUSIONS

Radiomics features integrated with machine learning methods can identify early bone metastases that are not visible on CT images. In our analysis, KNN is considered the optimal classification method.

PET-based radiomic feature based on the cross-combination method for predicting the mid-term efficacy and prognosis in high-risk diffuse large B-cell lymphoma patients

Objectives

This study aims to develop 7×7 machine-learning cross-combinatorial methods for selecting and classifying radiomic features used to construct Radiomics Score (RadScore) of predicting the mid-term efficacy and prognosis in high-risk patients with diffuse large B-cell lymphoma (DLBCL).

Methods

Retrospectively, we recruited 177 high-risk DLBCL patients from two medical centers between October 2012 and September 2022 and randomly divided them into a training cohort (n=123) and a validation cohort (n=54). We finally extracted 110 radiomic features along with SUVmax, MTV, and TLG from the baseline PET. The 49 features selection-classification pairs were used to obtain the optimal LASSO-LASSO model with 11 key radiomic features for RadScore. Logistic regression was employed to identify independent RadScore, clinical and PET factors. These models were evaluated using receiver operating characteristic (ROC) curves and calibration curves. Decision curve analysis (DCA) was conducted to assess the predictive power of the models. The prognostic power of RadScore was assessed using cox regression (COX) and Kaplan-Meier plots (KM).

Results

177 patients (mean age, 63 ± 13 years,129 men) were evaluated. Multivariate analyses showed that gender (OR,2.760; 95%CI:1.196,6.368); p=0.017), B symptoms (OR,4.065; 95%CI:1.837,8.955; p=0.001), SUVmax (OR,2.619; 95%CI:1.107,6.194; p=0.028), and RadScore (OR,7.167; 95%CI:2.815,18.248; p<0.001) independently contributed to the risk factors for predicting mid-term outcome. The AUC values of the combined models in the training and validation groups were 0.846 and 0.724 respectively, outperformed the clinical model (0.714;0.556), PET based model (0.664; 0.589), NCCN-IPI model (0.523;0.406) and IPI model (0.510;0.412) in predicting mid-term treatment outcome. DCA showed that the combined model incorporating RadScore, clinical risk factors, and PET metabolic metrics has optimal net clinical benefit. COX indicated that the high RadScore group had worse prognosis and survival in progression-free survival (PFS) (HR, 2.1737,95%CI: 1.2983, 3.6392) and overall survival (OS) (HR,2.1356,95%CI: 1.2561, 3.6309) compared to the low RadScore group. KM survival analysis also showed the same prognosis prediction as Cox results.

Conclusion

The combined model incorporating RadScore, sex, B symptoms and SUVmax demonstrates a significant enhancement in predicting medium-term efficacy and prognosis in high-risk DLBCL patients. RadScore using 7×7 machine learning cross-combinatorial methods for selection and classification holds promise as a potential method for evaluating medium-term treatment outcome and prognosis in high-risk DLBCL patients.

Consequences of domestication in eastern oyster: Insights from whole genomic analyses

Selective breeding for production traits has yielded relatively rapid successes with high-fecundity aquaculture species. Discovering the genetic changes associated with selection is an important goal for understanding adaptation and can also facilitate better predictions about the likely fitness of selected strains if they escape aquaculture farms. Here, we hypothesize domestication as a genetic change induced by inadvertent selection in culture. Our premise is that standardized culture protocols generate parallel domestication effects across independent strains. Using eastern oyster as a model and a newly developed 600K SNP array, this study tested for parallel domestication effects in multiple independent selection lines compared with their progenitor wild populations. A single contrast was made between pooled selected strains (1-17 generations in culture) and all wild progenitor samples combined. Population structure analysis indicated rank order levels of differentiation as [wild - wild] < [wild - cultured] < [cultured - cultured]. A genome scan for parallel adaptation to the captive environment applied two methodologically distinct outlier tests to the wild versus selected strain contrast and identified a total of 1174 candidate SNPs. Contrasting wild versus selected strains revealed the early evolutionary consequences of domestication in terms of genomic differentiation, standing genetic diversity, effective population size, relatedness, runs of homozygosity profiles, and genome-wide linkage disequilibrium patterns. Random Forest was used to identify 37 outlier SNPs that had the greatest discriminatory power between bulked wild and selected oysters. The outlier SNPs were in genes enriched for cytoskeletal functions, hinting at possible traits under inadvertent selection during larval culture or pediveliger setting at high density. This study documents rapid genomic changes stemming from hatchery-based cultivation of eastern oysters, identifies candidate loci responding to domestication in parallel among independent aquaculture strains, and provides potentially useful genomic resources for monitoring interbreeding between farm and wild oysters.

Single-nucleus transcriptomic analysis reveals the relationship between gene expression in oligodendrocyte lineage and major depressive disorder

BACKGROUND

Major depressive disorder (MDD) is a common mental illness that affects millions of people worldwide and imposes a heavy burden on individuals, families and society. Previous studies on MDD predominantly focused on neurons and employed bulk homogenates of brain tissues. This paper aims to decipher the relationship between oligodendrocyte lineage (OL) development and MDD at the single-cell resolution level.

METHODS

Here, we present the use of a guided regularized random forest (GRRF) algorithm to explore single-nucleus RNA sequencing profiles (GSE144136) of the OL at four developmental stages, which contains dorsolateral prefrontal cortex of 17 healthy controls (HC) and 17 MDD cases, generated by Nagy C et al. We prioritized and ordered differentially expressed genes (DEGs) based on Nagy et al., which could predominantly discriminate cells in the four developmental stages and two adjacent developmental stages of the OL. We further screened top-ranked genes that distinguished between HC and MDD in four developmental stages. Moreover, we estimated the performance of the GRRF model via the area under the curve value. Additionally, we validated the pivotal candidate gene Malat1 in animal models.

RESULTS

We found that, among the four developmental stages, the onset development of OL (OPC2) possesses the best predictive power for distinguishing HC and MDD, and long noncoding RNA MALAT1 has top-ranked importance value in candidate genes of four developmental stages. In addition, results of fluorescence in situ hybridization assay showed that Malat1 plays a critical role in the occurrence of depression.

CONCLUSIONS

Our work elucidates the mechanism of MDD from the perspective of OL development at the single-cell resolution level and provides novel insight into the occurrence of depression.

Are rabid raccoons (Procyon lotor) ready for the rapture? Determining the geographic origin of rabies virus-infected raccoons using RADcapture and microhaplotypes

North America is recognized for the exceptional richness of rabies virus (RV) wildlife reservoir species. Management of RV is accomplished through vaccination targeting mesocarnivore reservoir populations, such as the raccoon (Procyon lotor) in Eastern North America. Raccoons are a common generalist species, and populations may reach high densities in developed areas, which can result in contact with humans and pets with potential exposures to the raccoon variant of RV throughout the eastern United States. Understanding the spatial movement of RV by raccoon populations is important for monitoring and refining strategies supporting the landscape-level control and local elimination of this lethal zoonosis. We developed a high-throughput genotyping panel for raccoons based on hundreds of microhaplotypes to identify population structure and genetic diversity relevant to rabies management programs. Throughout the eastern United States, we identified hierarchical population genetic structure with clusters that were connected through isolation-by-distance. We also illustrate that this genotyping approach can be used to support real-time management priorities by identifying the geographic origin of a rabid raccoon that was collected in an area of the United States that had been raccoon RV-free for 8 years. The results from this study and the utility of the microhaplotype panel and genotyping method will provide managers with information on raccoon ecology that can be incorporated into future management decisions.

Machine learning as a tool to identify inpatients who are not at risk of adverse drug events in a large dataset of a tertiary care hospital in the USA

AIMS

Adverse drug events (ADEs) are a major threat to inpatients in the United States of America (USA). It is unknown how well machine learning (ML) is able to predict whether or not a patient will suffer from an ADE during hospital stay based on data available at hospital admission for emergency department patients of all ages (binary classification task). It is further unknown whether ML is able to outperform logistic regression (LR) in doing so, and which variables are the most important predictors.

METHODS

In this study, 5 ML models- namely a random forest, gradient boosting machine (GBM), ridge regression, least absolute shrinkage and selection operator (LASSO) regression, and elastic net regression-as well as a LR were trained and tested for the prediction of inpatient ADEs identified using ICD-10-CM codes based on comprehensive previous work in a diverse population. In total, 210 181 observations from patients who were admitted to a large tertiary care hospital after emergency department stay between 2011 and 2019 were included. The area under the receiver operating characteristics curve (AUC) and AUC-precision-recall (AUC-PR) were used as primary performance indicators.

RESULTS

Tree-based models performed best with respect to AUC and AUC-PR. The gradient boosting machine (GBM) reached an AUC of 0.747 (95% confidence interval (CI): 0.735 to 0.759) and an AUC-PR of 0.134 (95% CI: 0.131 to 0.137) on unforeseen test data, while the random forest reached an AUC of 0.743 (95% CI: 0.731 to 0.755) and an AUC-PR of 0.139 (95% CI: 0.135 to 0.142), respectively. ML statistically significantly outperformed LR both on AUC and AUC-PR. Nonetheless, overall, models did not differ much with respect to their performance. Most important predictors were admission type, temperature and chief complaint for the best performing model (GBM).

CONCLUSIONS

The study demonstrated a first application of ML to predict inpatient ADEs based on ICD-10-CM codes, and a comparison with LR. Future research should address concerns arising from low precision and related problems.

FedScore: A privacy-preserving framework for federated scoring system development

OBJECTIVE

We propose FedScore, a privacy-preserving federated learning framework for scoring system generation across multiple sites to facilitate cross-institutional collaborations.

MATERIALS AND METHODS

The FedScore framework includes five modules: federated variable ranking, federated variable transformation, federated score derivation, federated model selection and federated model evaluation. To illustrate usage and assess FedScore's performance, we built a hypothetical global scoring system for mortality prediction within 30 days after a visit to an emergency department using 10 simulated sites divided from a tertiary hospital in Singapore. We employed a pre-existing score generator to construct 10 local scoring systems independently at each site and we also developed a scoring system using centralized data for comparison.

RESULTS

We compared the acquired FedScore model's performance with that of other scoring models using the receiver operating characteristic (ROC) analysis. The FedScore model achieved an average area under the curve (AUC) value of 0.763 across all sites, with a standard deviation (SD) of 0.020. We also calculated the average AUC values and SDs for each local model, and the FedScore model showed promising accuracy and stability with a high average AUC value which was closest to the one of the pooled model and SD which was lower than that of most local models.

CONCLUSION

This study demonstrates that FedScore is a privacy-preserving scoring system generator with potentially good generalizability.

Motor imagery classification using sparse representations: an exploratory study

The non-stationary nature of the EEG signal poses challenges for the classification of motor imagery. sparse representation classification (SRC) appears as an alternative for classification of untrained conditions and, therefore, useful in motor imagery. Empirical mode decomposition (EMD) deals with signals of this nature and appears at the rear of the classification, supporting the generation of features. In this work we evaluate the combination of these methods in a multiclass classification problem, comparing them with a conventional method in order to determine if their performance is regular. For comparison with SRC we use multilayer perceptron (MLP). We also evaluated a hybrid approach for classification of sparse representations with MLP (RSMLP). For comparison with EMD we used filtering by frequency bands. Feature selection methods were used to select the most significant ones, specifically Random Forest and Particle Swarm Optimization. Finally, we used data augmentation to get a more voluminous base. Regarding the first dataset, we observed that the classifiers that use sparse representation have results equivalent to each other, but they outperform the conventional MLP model. SRC and SRMLP achieve an average accuracy of [Formula: see text] and [Formula: see text] respectively while the MLP is [Formula: see text], representing a gain between [Formula: see text] and [Formula: see text]. The use of EMD in relation to other feature processing techniques is not superior. However, EMD does not influence negatively, there is an opportunity for improvement. Finally, the use of data augmentation proved to be important to obtain relevant results. In the second dataset, we did not observe the same results. Models based on sparse representation (SRC, SRMLP, etc.) have on average a performance close to other conventional models, but without surpassing them. The best sparse models achieve an average accuracy of [Formula: see text] among the subjects in the base, while other model reach [Formula: see text]. The improvement of self-adaptive mechanisms that respond efficiently to the user's context is a good way to achieve improvements in motor imagery applications. However, other scenarios should be investigated, since the advantage of these methods was not proven in all datasets studied. There is still room for improvement, such as optimizing the dictionary of sparse representation in the context of motor imagery. Investing efforts in synthetically increasing the training base has also proved important to reduce the costs of this group of applications.

Genomic and machine learning-based screening of aquaculture-associated introgression into at-risk wild North American Atlantic salmon (Salmo salar) populations

The negative genetic impacts of gene flow from domestic to wild populations can be dependent on the degree of domestication and exacerbated by the magnitude of pre-existing genetic differences between wild populations and the domestication source. Recent evidence of European ancestry within North American aquaculture Atlantic salmon (Salmo salar) has elevated the potential impact of escaped farmed salmon on often at-risk wild North American salmon populations. Here, we compare the ability of single nucleotide polymorphism (SNP) and microsatellite (SSR) marker panels of different sizes (7-SSR, 100-SSR and 220K-SNP) to detect introgression of European genetic information into North American wild and aquaculture populations. Linear regression comparing admixture predictions for a set of individuals common to the three datasets showed that the 100-SSR panel and 7-SSR panels replicated the full 220K-SNP-based admixture estimates with low accuracy (r² of .64 and .49, respectively). Additional tests explored the effects of individual sample size and marker number, which revealed that ~300 randomly selected SNPs could replicate the 220K-SNP admixture predictions with greater than 95% fidelity. We designed a custom SNP panel (301-SNP) for European admixture detection in future monitoring work and then developed and tested a python package, salmoneuadmix (https://github.com/CNuge/SalmonEuAdmix), which uses a deep neural network to make de novo estimates of individuals' European admixture proportion without the need to conduct complete admixture analysis utilizing baseline samples. The results demonstrate the mobilization of targeted SNP panels and machine learning in support of at-risk species conservation and management.

RF-BiLSTM Neural Network Incorporating Attention Mechanism for Online Ride-Hailing Demand Forecasting

Accurately predicting online ride-hailing demand can help operators allocate vehicle resources on demand, avoid idle time, and improve traffic conditions. However, due to the randomness and complexity of online ride-hailing demand data, which are affected by many factors and mostly time-series in nature, it is difficult to forecast accurately and effectively based on traditional forecasting models. Therefore, this study proposes an online ride-hailing demand forecasting model based on the attention mechanism of a random forest (RF) combined with a symmetric bidirectional long short-term memory (BiLSTM) neural network (Att-RF-BiLSTM). The model optimizes the inputs and can use past and future data to forecast, improving the forecasting precision of online ride-hailing demand. The model utilizes a random forest to filter and optimize the input variables to reduce the neural network complexity, and then an attention mechanism was incorporated into the BiLSTM neural network to construct a demand forecasting model and validate it using actual Uber pickup data from New York City. Compared with other forecasting models (Att-XGBoost-BiLSTM, Att-BiLSTM, and pure LSTM), the results show that the proposed symmetrical Att-RF-BiLSTM online ride-hailing demand forecasting model has a higher forecasting precision and fitting degree, which indicates that the proposed model can be satisfactorily applied to the area of online ride-hailing demand.

Ability of machine learning models to identify preferred habitat traits of a small indigenous fish (Chanda nama) in a large river of peninsular India

Physical and chemical parameters of river influence the habitat of fish species in aquatic ecosystems. Fish showed a complex relationship with different aquatic factors in river. Machine learning modeling is a useful tool to identify relationships between components of a complex environmental system. We identified the preferred habitat indicators of Chanda nama (a small indigenous fish), in the Krishna River located in peninsular India, using machine learning modeling. Using data on Chanda nama fish distribution (presence/absence) and associated ten physical and chemical parameters of water at 22 sampling sites of the river collected during the year 2001-2002, machine learning models such as random forest, artificial neural network, support vector machine, and k-nearest neighbors were used for classification of Chanda nama distribution in the river. The machine learning model efficiency was evaluated using classification accuracy, Cohen's kappa coefficient, sensitivity, specificity, and receiver-operating-characteristics. Results showed that random forest is the best model with higher classification accuracy (82%), Cohen's kappa coefficient (0.55), sensitivity (0.57), specificity (0.76), and receiver-operating-characteristics (0.72) for prediction of the occurrence of Chanda nama in the Krishna River. Random forest model identified three preferred physicochemical habitat traits such as altitude, temperature, and depth for Chanda nama distribution in Krishna River. Our results will be helpful for researcher and policy maker to understand important physical and chemical variables for sustainable management of a small indigenous fish (Chanda nama) in a large tropical river.

An ensemble learning approach to map the genetic connectivity of the parasitoid Stethynium empoasca (Hymenoptera: Mymaridae) and identify the key influencing environmental and landscape factors

The effect of landscape patterns and environmental factors on the population structure and genetic diversity of organisms is well-documented. However, this effect is still unclear in the case of Mymaridae parasitoids. Despite recent advances in machine learning methods for landscape genetics, ensemble learning still needs further investigation. Here, we evaluated the performance of different boosting algorithms and analyzed the effects of landscape and environmental factors on the genetic variations in the tea green leafhopper parasitoid Stethynium empoasca (Hymenoptera: Mymaridae). The S. empoasca populations showed a distinct pattern of isolation by distance. The minimum temperature of the coldest month, annual precipitation, the coverage of evergreen/deciduous needleleaf trees per 1 km2, and the minimum precipitation of the warmest quarter were identified as the dominant factors affecting the genetic divergence of S. empoasca populations. Notably, compared to previous machine learning studies, our model showed an unprecedented accuracy (r = 0.87) for the prediction of genetic differentiation. These findings not only demonstrated how the landscape shaped S. empoasca genetics but also provided an essential basis for developing conservation strategies for this biocontrol agent. In a broader sense, this study demonstrated the importance and efficiency of ensemble learning in landscape genetics.

From STRs to SNPs via ddRAD-seq: Geographic assignment of confiscated tortoises at reduced costs

Assigning individuals to their source populations is crucial for conservation research, especially for endangered species threatened by illegal trade and translocations. Genetic assignment can be achieved with different types of molecular markers, but technical advantages and cost saving are recently promoting the shift from short tandem repeats (STRs) to single nucleotide polymorphisms (SNPs). Here, we designed, developed, and tested a small panel of SNPs for cost-effective geographic assignment of individuals with unknown origin of the endangered Mediterranean tortoise Testudo hermanni. We started by performing a ddRAD-seq experiment on 70 wild individuals of T. hermanni from 38 locations. Results obtained using 3182 SNPs are comparable to those previously obtained using STR markers in terms of genetic structure and power to identify the macro-area of origin. However, our SNPs revealed further insights into the substructure in Western populations, especially in Southern Italy. A small panel of highly informative SNPs was then selected and tested by genotyping 190 individuals using the KASP genotyping chemistry. All the samples from wild populations of known geographic origin were genetically re-assigned with high accuracy to the original population. This reduced SNPs panel represents an efficient molecular tool that enables individuals to be genotyped at low cost (less than €15 per sample) for geographical assignment and identification of hybrids. This information is crucial for the management in-situ of confiscated animals and their possible re-allocation in the wild. Our methodological pipeline can easily be extended to other species.

Comparing mixed models and Random Forest association tests using naturalGWAS and a Striped Bass SNP dataset

In this study, we used the phenotype simulation package naturalGWAS to test the performance of Zhao's Random Forest method in comparison to an uncorrected Random Forest test, latent factor mixed models (LFMM), genome-wide efficient mixed models (GEMMA), and confounder adjusted linear regression (CATE). We created 400 sets of phenotypes, corresponding to five effect sizes and 2, 5, 15, or 30 causal loci, simulated from two empirical datasets containing SNPs from Striped Bass representing three and 13 populations. All association methods were evaluated for their ability to detect genotype-phenotype associations based on power, false discovery rates, and number of false positives. Genomic inflation was highest for uncorrected Random Forest and LFMM tests and lowest for Gemma and Zhao's Random Forest. All association tests had similar power to detect causal loci, and Zhao's Random Forest had the lowest false discovery rate in all scenarios. To measure the performance of association tests in small datasets with few loci surrounding a causal gene we also ran analyses again after removing causal loci from each dataset. All association tests were only able to find true positives, defined as loci located within 30k bp of a causal locus, in 3%-18% of simulations. In contrast, at least one false positive was found in 17%-44% of simulations. Zhao's Random Forest again identified the fewest false positives of all association tests studied. The ability to test the power of association tests for individual empirical datasets can be an extremely useful first step when designing a GWAS study.

A feature selection strategy for improving software maintainability prediction

Software maintainability is a significant contributor while choosing particular software. It is helpful in estimation of the efforts required after delivering the software to the customer. However, issues like imbalanced distribution of datasets, and redundant and irrelevant occurrence of various features degrade the performance of maintainability prediction models. Therefore, current study applies ImpS algorithm to handle imbalanced data and extensively investigates several Feature Selection (FS) techniques including Symmetrical Uncertainty (SU), RandomForest filter, and Correlation-based FS using one open-source, three proprietaries and two commercial datasets. Eight different machine learning algorithms are utilized for developing prediction models. The performance of models is evaluated using Accuracy, G-Mean, Balance, & Area under the ROC Curve. Two statistical tests, Friedman Test and Wilcoxon Signed Ranks Test are conducted for assessing different FS techniques. The results substantiate that FS techniques significantly improve the performance of various prediction models with an overall improvement of 18.58%, 129.73%, 80.00%, and 45.76% in the median values of Accuracy, G-Mean, Balance, & AUC, respectively for all the datasets taken together. Friedman test advocates the supremacy of SU FS technique. Wilcoxon Signed Ranks test showcases that SU FS technique is significantly superior to the CFS technique for three out of six datasets.

Optimal PET-based radiomic signature construction based on the cross-combination method for predicting the survival of patients with diffuse large B-cell lymphoma

PURPOSE

To develop and externally validate models incorporating a PET radiomics signature (R-signature) obtained by the cross-combination method for predicting the survival of patients with diffuse large B-cell lymphoma (DLBCL).

METHODS

A total of 383 patients with DLBCL from two medical centres between 2011 and 2019 were included. The cross-combination method was used on three types of PET radiomics features from the training cohort to generate 49 feature selection-classification candidates based on 7 different machine learning models. The R-signature was then built by selecting the optimal candidates based on their progression-free survival (PFS) and overall survival (OS). Cox regression analysis was used to develop the survival prediction models. The calibration, discrimination, and clinical utility of the models were assessed and externally validated.

RESULTS

The R-signatures determined by 12 and 31 radiomics features were significantly associated with PFS and OS, respectively (P<0.05). The combined models that incorporated R-signatures, metabolic metrics, and clinical risk factors exhibited significant prognostic superiority over the clinical models, PET-based models, and the National Comprehensive Cancer Network International Prognostic Index in terms of both PFS (C-index: 0.801 vs. 0.732 vs. 0.785 vs. 0.720, respectively) and OS (C-index: 0.807 vs. 0.740 vs. 0.773 vs. 0.726, respectively). For external validation, the C-indices were 0.758 vs. 0.621 vs. 0.732 vs. 0.673 and 0.794 vs. 0.696 vs. 0.781 vs. 0.708 in the PFS and OS analyses, respectively. The calibration curves showed good consistency, and the decision curve analysis supported the clinical utility of the combined model.

CONCLUSION

The R-signature could be used as a survival predictor for DLBCL, and its combination with clinical factors may allow for accurate risk stratification.

Leaf Area Index Estimation of Pergola-Trained Vineyards in Arid Regions Based on UAV RGB and Multispectral Data Using Machine Learning Methods

The leaf area index (LAI), a valuable variable for assessing vine vigor, reflects nutrient concentrations in vineyards and assists in precise management, including fertilization, improving yield, quality, and vineyard uniformity. Although some vegetation indices (VIs) have been successfully used to assess LAI variations, they are unsuitable for vineyards of different types and structures. By calibrating the light extinction coefficient of a digital photography algorithm for proximal LAI measurements, this study aimed to develop VI-LAI models for pergola-trained vineyards based on high-resolution RGB and multispectral images captured by an unmanned aerial vehicle (UAV). The models were developed by comparing five machine learning (ML) methods, and a robust ensemble model was proposed using the five models as base learners. The results showed that the ensemble model outperformed the base models. The highest R2 and lowest RMSE values that were obtained using the best combination of VIs with multispectral data were 0.899 and 0.434, respectively; those obtained using the RGB data were 0.825 and 0.547, respectively. By improving the results by feature selection, ML methods performed better with multispectral data than with RGB images, and better with higher spatial resolution data than with lower resolution data. LAI variations can be monitored efficiently and accurately for large areas of pergola-trained vineyards using this framework.

Is it possible developing reliable prediction models considering only the pipe’s age for decision-making in sewer asset management?

Purpose

The purpose of this paper was exploring and comparing different deterioration models based on statistical and machine learning approaches. These models were chosen from their successful results in other case studies. The deterioration models were developing considering two scenarios: (i) only the age as covariate (Scenario 1); and (ii) the age together with other available sewer characteristics as covariates (Scenario 2). Both were evaluated to achieve two different management objectives related to the prediction of the critical condition of sewers: at the network and the sewer levels.

Design/methodology/approach

Six statistical and machine learning methods [logistic regression (LR), random forest (RF), multinomial logistic regression, ordinal logistic regression, linear discriminant analysis and support vector machine] were explored considering two kinds of predictor variables (independent variables in the model). The main propose of these models was predicting the structural condition at network and pipe level evaluated from deviation analysis and performance curve techniques. Further, the deterioration models were exploring for two case studies: the sewer systems of Bogota and Medellin. These case studies were considered because of both counts with their own assessment standards and low inspection rate.

Findings

The results indicate that LR models for both case studies show higher prediction capacity under Scenario 1 (considering only the age) for the management objective related to the network, such as annual budget plans; and RF shows the highest success percentage of sewers in critical condition (sewer level) considering Scenario 2 for both case studies.

Practical implications

There is not a deterioration method whose predictions are adaptable for achieving different management objectives; it is important to explore different approaches to find which one could support a sewer asset management objective for a specific case study.

Originality/value

The originality of this paper consists of there is not a paper in which the prediction of several statistical and machine learning-based deterioration models has been compared for case studies with different local assessment standard. The above to find which is adaptable for each one and which model is adaptable for each management objective.

Forecast of the COVID-19 Epidemic Based on RF-BOA-LightGBM

In this paper, we utilize the Internet big data tool, namely Baidu Index, to predict the development trend of the new coronavirus pneumonia epidemic to obtain further data. By selecting appropriate keywords, we can collect the data of COVID-19 cases in China between 1 January 2020 and 1 April 2020. After preprocessing the data set, the optimal sub-data set can be obtained by using random forest feature selection method. The optimization results of the seven hyperparameters of the LightGBM model by grid search, random search and Bayesian optimization algorithms are compared. The experimental results show that applying the data set obtained from the Baidu Index to the Bayesian-optimized LightGBM model can better predict the growth of the number of patients with new coronary pneumonias, and also help people to make accurate judgments to the development trend of the new coronary pneumonia.

Low-coverage whole-genome sequencing reveals molecular markers for spawning season and sex identification in Gulf of Maine Atlantic cod (Gadus morhua, Linnaeus 1758)

Atlantic cod (Gadus morhua, Linnaeus 1758) in the western Gulf of Maine are managed as a single stock despite several lines of evidence supporting two spawning groups (spring and winter) that overlap spatially, while exhibiting seasonal spawning isolation. Low-coverage whole-genome sequencing was used to evaluate the genomic population structure of Atlantic cod spawning groups in the western Gulf of Maine and Georges Bank using 222 individuals collected over multiple years. Results indicated low total genomic differentiation, while also showing strong differentiation between spring and winter-spawning groups at specific regions of the genome. Guided regularized random forest and ranked F ST methods were used to select panels of single nucleotide polymorphisms (SNPs) that could reliably distinguish spring and winter-spawning Atlantic cod (88.5% assignment rate), as well as males and females (95.0% assignment rate) collected in the western Gulf of Maine. These SNP panels represent a valuable tool for fisheries research and management of Atlantic cod in the western Gulf of Maine that will aid investigations of stock production and support accuracy of future assessments.

RIFS2D: A two-dimensional version of a randomly restarted incremental feature selection algorithm with an application for detecting low-ranked biomarkers

The era of big data introduces both opportunities and challenges for biomedical researchers. One of the inherent difficulties in the biomedical research field is to recruit large cohorts of samples, while high-throughput biotechnologies may produce thousands or even millions of features for each sample. Researchers tend to evaluate the individual correlation of each feature with the class label and use the incremental feature selection (IFS) strategy to select the top-ranked features with the best prediction performance. Recent experimental data showed that a subset of continuously ranked features randomly restarted from a low-ranked feature (an RIFS block) may outperform the subset of top-ranked features. This study proposed a feature selection Algorithm RIFS2D by integrating multiple RIFS blocks. A comprehensive comparative experiment was conducted with the IFS, RIFS and existing feature selection algorithms and demonstrated that a subset of low-ranked features may also achieve promising prediction performance. This study suggested that a prediction model with promising performance may be trained by low-ranked features, even when top-ranked features did not achieve satisfying prediction performance. Further comparative experiments were conducted between RIFS2D and t-tests for the detection of early-stage breast cancer. The data showed that the RIFS2D-recommended features achieved better prediction accuracy and were targeted by more drugs than the t-test top-ranked features.

Use of radiomics based on 18F-FDG PET/CT and machine learning methods to aid clinical decision-making in the classification of solitary pulmonary lesions: an innovative approach

PURPOSE

This study was designed and performed to assess the ability of ¹⁸F-fluorodeoxyglucose (FDG) positron emission tomography (PET) and computed tomography (CT) radiomics features combined with machine learning methods to differentiate between primary and metastatic lung lesions and to classify histological subtypes. Moreover, we identified the optimal machine learning method.

METHODS

A total of 769 patients pathologically diagnosed with primary or metastatic lung cancers were enrolled. We used the LIFEx package to extract radiological features from semiautomatically segmented PET and CT images within the same volume of interest. Patients were randomly distributed in training and validation sets. Through the evaluation of five feature selection methods and nine classification methods, discriminant models were established. The robustness of the procedure was controlled by tenfold cross-validation. The model's performance was evaluated using the area under the receiver operating characteristic curve (AUC).

RESULTS

Based on the radiomics features extracted from PET and CT images, forty-five discriminative models were established. Combined with appropriate feature selection methods, most classifiers showed excellent discriminative ability with AUCs greater than 0.75. In the differentiation between primary and metastatic lung lesions, the feature selection method gradient boosting decision tree (GBDT) combined with the classifier GBDT achieved the highest classification AUC of 0.983 in the PET dataset. In contrast, the feature selection method eXtreme gradient boosting combined with the classifier random forest (RF) achieved the highest AUC of 0.828 in the CT dataset. In the discrimination between squamous cell carcinoma and adenocarcinoma, the combination of GBDT feature selection method with GBDT classification had the highest AUC of 0.897 in the PET dataset. In contrast, the combination of the GBDT feature selection method with the RF classification had the highest AUC of 0.839 in the CT dataset. Most of the decision tree (DT)-based models were overfitted, suggesting that the classification method was not appropriate for practical application.

CONCLUSION

¹⁸F-FDG PET/CT radiomics features combined with machine learning methods can distinguish between primary and metastatic lung lesions and identify histological subtypes in lung cancer. GBDT and RF were considered optimal classification methods for the PET and CT datasets, respectively, and GBDT was considered the optimal feature selection method in our analysis.

GLIMPSE: a glioblastoma prognostication model using ensemble learning-a surveillance, epidemiology, and end results study

Purpose

Glioblastoma is one of the most common and aggressive brain tumors in the world with a poor prognosis. A glioblastoma prognostication model has the potential to improve the cancer's standard of care. No other paper has looked at using ensemble learning with a population database to predict multiple binary glioblastoma survival outcomes.

Methods

We utilized ensemble learning to design, build, and test a prognostication system for glioblastoma for short-, intermediate- and long-term survival, based on various clinical features. We used the population database SEER which covers 17 different registries. The most important prognostic features were identified and used as a clinical feature set. The statistical feature set was determined using Random Forests. The accuracy, sensitivity, specificity, area under the receiver operating characteristic (AUROC), positive predictive value (PPV), and negative predictive value (NPV) were reported.

Results

Statistically-determined feature sets had the best performance. All the top models for short, intermediate, and long-term survival were random forests. With regards to short-term survival, top model had metrics AUROC = 0.937, accuracy = 86%, specificity = 88%, sensitivity = 85%, NPV = 85%, and PPV = 87%. For long-term survival, the top model had AUROC = 0.893, accuracy = 81%, specificity = 79%, sensitivity = 83%, NPV = 82%, and PPV = 79%. The top intermediate-term survival prediction had AUROC ≥ 0.780 and the other metrics were at least 70%.

Conclusions

Our ensemble models were high-performing and achieved AUROCs as high as 0.94, highlighting the importance of balancing, using ensemble techniques and statistical feature selection. Our models can potentially be used by clinicians after external validation.

Power of a dual-use SNP panel for pedigree reconstruction and population assignment

The use of high-throughput, low-density sequencing approaches has dramatically increased in recent years in studies of eco-evolutionary processes in wild populations and domestication in commercial aquaculture. Most of these studies focus on identifying panels of SNP loci for a single downstream application, whereas there have been few studies examining the trade-offs for selecting panels of markers for use in multiple applications. Here, we detail the use of a bioinformatic workflow for the development of a dual-purpose SNP panel for parentage and population assignment, which included identifying putative SNP loci, filtering for the most informative loci for the two tasks, designing effective multiplex PCR primers, optimizing the SNP panel for performance, and performing quality control steps for downstream applications. We applied this workflow to two adjacent Alaskan Sockeye Salmon populations and identified a GTseq panel of 142 SNP loci for parentage and 35 SNP loci for population assignment. Only 50-75 panel loci were necessary for >95% accurate parentage, whereas population assignment success, with all 172 panel loci, ranged from 93.9% to 96.2%. Finally, we discuss the trade-offs and complexities of the decision-making process that drives SNP panel development, optimization, and testing.

Leaf Area Index Estimation Algorithm for GF-5 Hyperspectral Data Based on Different Feature Selection and Machine Learning Methods

Leaf area index (LAI) is an essential vegetation parameter that represents the light energy utilization and vegetation canopy structure. As the only in-operation hyperspectral satellite launched by China, GF-5 is potentially useful for accurate LAI estimation. However, there is no research focus on evaluating GF-5 data for LAI estimation. Hyperspectral remote sensing data contains abundant information about the reflective characteristics of vegetation canopies, but these abound data also easily result in a dimensionality curse. Therefore, feature selection (FS) is necessary to reduce data redundancy to achieve more reliable estimations. Currently, machine learning (ML) algorithms have been widely used for FS. Moreover, the same ML algorithm is usually conducted for both FS and regression in LAI estimation. However, no evidence suggests that this is the optimal solution. Therefore, this study focuses on evaluating the capacity of GF-5 spectral reflectance for estimating LAI and the performances of different combination of FS and ML algorithms. Firstly, the PROSAIL model, which coupled leaf optical properties model PROSPECT and the scattering by arbitrarily inclined leaves (SAIL) model, was used to generate simulated GF-5 reflectance data under different vegetation and soil conditions, and then three FS methods, including random forest (RF), K-means clustering (K-means) and mean impact value (MIV), and three ML algorithms, including random forest regression (RFR), back propagation neural network (BPNN) and K-nearest neighbor (KNN) were used to develop nine LAI estimation models. The FS process was conducted twice using different strategies: Firstly, three FS methods were conducted to search the lowest dimension number, which maintained the estimation accuracy of all bands. Then, the sequential backward selection (SBS) method was used to eliminate the bands having minimal impact on LAI estimation accuracy. Finally, three best estimation models were selected and evaluated using reference LAI. The results showed that although the RF_RFR model (RF used for feature selection and RFR used for regression) achieved reliable LAI estimates (coefficient of determination (R2) = 0.828, root mean square error (RMSE) = 0.839), the poor performance (R2 = 0.763, RMSE = 0.987) of the MIV_BPNN model (MIV used for feature selection and BPNN used for regression) suggested using feature selection and regression conducted by the same ML algorithm could not always ensure an optimal estimation. Moreover, RF selection preserved the most informative bands for LAI estimation so that each ML regression method could achieve satisfactory estimation results. Finally, the results indicated that the RF_KNN model (RF used as feature selection and KNN used for regression) with seven GF-5 spectral band reflectance achieved the better estimation results than others when validated by simulated data (R2 = 0.834, RMSE = 0.824) and actual reference LAI (R2 = 0.659, RMSE = 0.697).

Divergent and linked selection shape patterns of genomic differentiation between European and North American Atlantic salmon (Salmo salar)

As populations diverge many processes can shape genomic patterns of differentiation. Regions of high differentiation can arise due to divergent selection acting on selected loci, genetic hitchhiking of nearby loci, or through repeated selection against deleterious alleles (linked background selection); this divergence may then be further elevated in regions of reduced recombination. Atlantic salmon (Salmo salar) from Europe and North America diverged >600,000 years ago and despite some evidence of secondary contact, the majority of genetic data indicate substantial divergence between lineages. This deep divergence with potential gene flow provides an opportunity to investigate the role of different mechanisms that shape the genomic landscape during early speciation. Here, using 184,295 single nucleotide polymorphisms (SNPs) and 80 populations, we investigate the genomic landscape of differentiation across the Atlantic Ocean with a focus on highly differentiated regions and the processes shaping them. We found evidence of high (mean F_ST  = 0.26) and heterogeneous genomic differentiation between continents. Genomic regions associated with high trans-Atlantic differentiation ranged in size from single loci (SNPs) within important genes to large regions (1-3 Mbp) on four chromosomes (Ssa06, Ssa13, Ssa16 and Ssa19). These regions showed signatures consistent with selection, including high linkage disequilibrium, despite no significant reduction in recombination. Genes and functional enrichment of processes associated with differentiated regions may highlight continental differences in ocean navigation and parasite resistance. Our results provide insight into potential mechanisms underlying differences between continents, and evidence of near-fixed and potentially adaptive trans-Atlantic differences concurrent with a background of high genome-wide differentiation supports subspecies designation in Atlantic salmon.

Resolving fine-scale population structure and fishery exploitation using sequenced microsatellites in a northern fish

The resiliency of populations and species to environmental change is dependent on the maintenance of genetic diversity, and as such, quantifying diversity is central to combating ongoing widespread reductions in biodiversity. With the advent of next-generation sequencing, several methods now exist for resolving fine-scale population structure, but the comparative performance of these methods for genetic assignment has rarely been tested. Here, we evaluate the performance of sequenced microsatellites and a single nucleotide polymorphism (SNP) array to resolve fine-scale population structure in a critically important salmonid in north eastern Canada, Arctic Charr (Salvelinus alpinus). We also assess the utility of sequenced microsatellites for fisheries applications by quantifying the spatial scales of movement and exploitation through genetic assignment of fishery samples to rivers of origin and comparing these results with a 29-year tagging dataset. Self-assignment and simulation-based analyses of 111 genome-wide microsatellite loci and 500 informative SNPs from 28 populations of Arctic Charr in north-eastern Canada identified largely river-specific genetic structure. Despite large differences (~4X) in the number of loci surveyed between panels, mean self-assignment accuracy was similar with the microsatellite loci and the SNP panel (>90%). Subsequent analysis of 996 fishery-collected samples using the microsatellite panel revealed that larger rivers contribute greater numbers of individuals to the fishery and that coastal fisheries largely exploit individuals originating from nearby rivers, corroborating results from traditional tagging experiments. Our results demonstrate the efficacy of sequence-based microsatellite genotyping to advance understanding of fine-scale population structure and harvest composition in northern and understudied species.

Development of heart failure risk prediction models based on a multi-marker approach using random forest algorithms

Background:

The early identification of heart failure (HF) risk may favorably affect outcomes, and the combination of multiple biomarkers may provide a more comprehensive and valuable means for improving the risk of stratification. This study was conducted to assess the importance of individual cardiac biomarkers creatine kinase MB isoenzyme (CK-MB), B-type natriuretic peptide (BNP), galectin-3 (Gal-3) and soluble suppression of tumorigenicity-2 (sST2) for HF diagnosis, and the predictive performance of the combination of these four biomarkers was analyzed using random forest algorithms.

Methods:

A total of 193 participants (80 patients with HF and 113 age- and gender-matched healthy controls) were included from June 2017 to December 2017. The correlation and regression analysis were conducted between cardiac biomarkers and echocardiographic parameters. The accuracy and importance of these predictor variables were assessed using random forest algorithms.

Results:

Patients with HF exhibited significantly higher levels of CK-MB, BNP, Gal-3, and sST2. BNP exhibited a good independent predictive capacity for HF (AUC 0.956). However, CK-MB, sST2, and Gal-3 exhibited a modest diagnostic performance for HF, with an AUC of 0.709, 0.711, and 0.777, respectively. BNP was the most important variable, with a remarkably higher mean decrease accuracy and Gini. Furthermore, there was a general increase in predictive performance using the multi-marker model, and the sensitivity, specificity was 91.5% and 96.7%, respectively.

Conclusion:

The random forest algorithm provides a robust method to assess the accuracy and importance of predictor variables. The combination of CK-MB, BNP, Gal-3, and sST2 achieves improvement in prediction accuracy for HF.

Genomic signatures and correlates of widespread population declines in salmon

Global losses of biodiversity are occurring at an unprecedented rate, but causes are often unidentified. Genomic data provide an opportunity to isolate drivers of change and even predict future vulnerabilities. Atlantic salmon (Salmo salar) populations have declined range-wide, but factors responsible are poorly understood. Here, we reconstruct changes in effective population size (Ne) in recent decades for 172 range-wide populations using a linkage-based method. Across the North Atlantic, Ne has significantly declined in >60% of populations and declines are consistently temperature-associated. We identify significant polygenic associations with decline, involving genomic regions related to metabolic, developmental, and physiological processes. These regions exhibit changes in presumably adaptive diversity in declining populations consistent with contemporary shifts in body size and phenology. Genomic signatures of widespread population decline and associated risk scores allow direct and potentially predictive links between population fitness and genotype, highlighting the power of genomic resources to assess population vulnerability.

Genetic characterization of worldwide Prunus domestica (plum) germplasm using sequence-based genotyping

Prunus domestica commonly known as European plum is a hexaploid fruit tree species cultivated around the world. Locally it is used for fresh consumption, in jams or jellies, and the production of spirits while commercially the fruit is primarily sold dried (prunes). Despite its agricultural importance and long history of cultivation, many questions remain about the origin of this species, the relationships among its many pomological types, and its underlying genetics. Here, we used a sequence-based genotyping approach to characterize worldwide plum germplasm including the potential progenitor Eurasian plum species. Analysis of 405 DNA samples established a set of four clades consistent with the pomological groups Greengages, Mirabelles, European plums, and d'Agen (French) prune plums. A number of cultivars from each clade were identified as likely clonal selections, particularly among the "French" type prune germplasm that is widely cultivated today. Overall, there was relatively low genetic diversity across all cultivated plums suggesting they have been largely inbred and/or derived from a limited number of founders. The results agree with P. domestica having originated as an interspecific hybrid of a diploid P. cerasifera and a tetraploid P. spinosa that itself may have been an interspecific hybrid of P. cerasifera and an unknown Eurasian plum species. The low genetic diversity and lack of true wild-types coupled with the known cultivation history of Eurasian plums imply that P. domestica may have been a product of inter-specific cross breeding and artificial selection by early agrarian Eurasian societies.

Environmental extremes drive population structure at the northern range limit of Atlantic salmon in North America

Conservation of exploited species requires an understanding of both genetic diversity and the dominant structuring forces, particularly near range limits, where climatic variation can drive rapid expansions or contractions of geographic range. Here, we examine population structure and landscape associations in Atlantic salmon (Salmo salar) across a heterogeneous landscape near the northern range limit in Labrador, Canada. Analysis of two amplicon-based data sets containing 101 microsatellites and 376 single nucleotide polymorphisms (SNPs) from 35 locations revealed clear differentiation between populations spawning in rivers flowing into a large marine embayment (Lake Melville) compared to coastal populations. The mechanisms influencing the differentiation of embayment populations were investigated using both multivariate and machine-learning landscape genetic approaches. We identified temperature as the strongest correlate with genetic structure, particularly warm temperature extremes and wider annual temperature ranges. The genomic basis of this divergence was further explored using a subset of locations (n = 17) and a 220K SNP array. SNPs associated with spatial structuring and temperature mapped to a diverse set of genes and molecular pathways, including regulation of gene expression, immune response, and cell development and differentiation. The results spanning molecular marker types and both novel and established methods clearly show climate-associated, fine-scale population structure across an environmental gradient in Atlantic salmon near its range limit in North America, highlighting valuable approaches for predicting population responses to climate change and managing species sustainability.

Genotyping-by-sequencing of genome-wide microsatellite loci reveals fine-scale harvest composition in a coastal Atlantic salmon fishery

Individual assignment and genetic mixture analysis are commonly utilized in contemporary wildlife and fisheries management. Although microsatellite loci provide unparalleled numbers of alleles per locus, their use in assignment applications is increasingly limited. However, next-generation sequencing, in conjunction with novel bioinformatic tools, allows large numbers of microsatellite loci to be simultaneously genotyped, presenting new opportunities for individual assignment and genetic mixture analysis. Here, we scanned the published Atlantic salmon genome to identify 706 microsatellite loci, from which we developed a final panel of 101 microsatellites distributed across the genome (average 3.4 loci per chromosome). Using samples from 35 Atlantic salmon populations (n = 1,485 individuals) from coastal Labrador, Canada, a region characterized by low levels of differentiation in this species, this panel identified 844 alleles (average of 8.4 alleles per locus). Simulation-based evaluations of assignment and mixture identification accuracy revealed unprecedented resolution, clearly identifying 26 rivers or groups of rivers spanning 500 km of coastline. This baseline was used to examine the stock composition of 696 individuals harvested in the Labrador Atlantic salmon fishery and revealed that coastal fisheries largely targeted regional groups (<300 km). This work suggests that the development and application of large sequenced microsatellite panels presents great potential for stock resolution in Atlantic salmon and more broadly in other exploited anadromous and marine species.

A Machine Learning Application Based in Random Forest for Integrating Mass Spectrometry-Based Metabolomic Data: A Simple Screening Method for Patients With Zika Virus

Recent Zika outbreaks in South America, accompanied by unexpectedly severe clinical complications have brought much interest in fast and reliable screening methods for ZIKV (Zika virus) identification. Reverse-transcriptase polymerase chain reaction (RT-PCR) is currently the method of choice to detect ZIKV in biological samples. This approach, nonetheless, demands a considerable amount of time and resources such as kits and reagents that, in endemic areas, may result in a substantial financial burden over affected individuals and health services veering away from RT-PCR analysis. This study presents a powerful combination of high-resolution mass spectrometry and a machine-learning prediction model for data analysis to assess the existence of ZIKV infection across a series of patients that bear similar symptomatic conditions, but not necessarily are infected with the disease. By using mass spectrometric data that are inputted with the developed decision-making algorithm, we were able to provide a set of features that work as a "fingerprint" for this specific pathophysiological condition, even after the acute phase of infection. Since both mass spectrometry and machine learning approaches are well-established and have largely utilized tools within their respective fields, this combination of methods emerges as a distinct alternative for clinical applications, providing a diagnostic screening-faster and more accurate-with improved cost-effectiveness when compared to existing technologies.

A practical introduction to Random Forest for genetic association studies in ecology and evolution

Large genomic studies are becoming increasingly common with advances in sequencing technology, and our ability to understand how genomic variation influences phenotypic variation between individuals has never been greater. The exploration of such relationships first requires the identification of associations between molecular markers and phenotypes. Here, we explore the use of Random Forest (RF), a powerful machine-learning algorithm, in genomic studies to discern loci underlying both discrete and quantitative traits, particularly when studying wild or nonmodel organisms. RF is becoming increasingly used in ecological and population genetics because, unlike traditional methods, it can efficiently analyse thousands of loci simultaneously and account for nonadditive interactions. However, understanding both the power and limitations of Random Forest is important for its proper implementation and the interpretation of results. We therefore provide a practical introduction to the algorithm and its use for identifying associations between molecular markers and phenotypes, discussing such topics as data limitations, algorithm initiation and optimization, as well as interpretation. We also provide short R tutorials as examples, with the aim of providing a guide to the implementation of the algorithm. Topics discussed here are intended to serve as an entry point for molecular ecologists interested in employing Random Forest to identify trait associations in genomic data sets.

Range-wide parallel climate-associated genomic clines in Atlantic salmon

Clinal variation across replicated environmental gradients can reveal evidence of local adaptation, providing insight into the demographic and evolutionary processes that shape intraspecific diversity. Using 1773 genome-wide single nucleotide polymorphisms we evaluated latitudinal variation in allele frequency for 134 populations of North American and European Atlantic salmon (Salmo salar). We detected 84 (4.74%) and 195 (11%) loci showing clinal patterns in North America and Europe, respectively, with 12 clinal loci in common between continents. Clinal single nucleotide polymorphisms were evenly distributed across the salmon genome and logistic regression revealed significant associations with latitude and seasonal temperatures, particularly average spring temperature in both continents. Loci displaying parallel clines were associated with several metabolic and immune functions, suggesting a potential basis for climate-associated adaptive differentiation. These climate-based clines collectively suggest evidence of large-scale environmental associated differences on either side of the North Atlantic. Our results support patterns of parallel evolution on both sides of the North Atlantic, with evidence of both similar and divergent underlying genetic architecture. The identification of climate-associated genomic clines illuminates the role of selection and demographic processes on intraspecific diversity in this species and provides a context in which to evaluate the impacts of climate change.

Applications of random forest feature selection for fine-scale genetic population assignment

Genetic population assignment used to inform wildlife management and conservation efforts requires panels of highly informative genetic markers and sensitive assignment tests. We explored the utility of machine‐learning algorithms (random forest, regularized random forest and guided regularized random forest) compared with F_ST ranking for selection of single nucleotide polymorphisms (SNP) for fine‐scale population assignment. We applied these methods to an unpublished SNP data set for Atlantic salmon (Salmo salar) and a published SNP data set for Alaskan Chinook salmon (Oncorhynchus tshawytscha). In each species, we identified the minimum panel size required to obtain a self‐assignment accuracy of at least 90% using each method to create panels of 50–700 markers Panels of SNPs identified using random forest‐based methods performed up to 7.8 and 11.2 percentage points better than F_ST‐selected panels of similar size for the Atlantic salmon and Chinook salmon data, respectively. Self‐assignment accuracy ≥90% was obtained with panels of 670 and 384 SNPs for each data set, respectively, a level of accuracy never reached for these species using F_ST‐selected panels. Our results demonstrate a role for machine‐learning approaches in marker selection across large genomic data sets to improve assignment for management and conservation of exploited populations.