Regularized machine learning in the genetic prediction of complex traits.
Ontology highlight
ABSTRACT:
SUBMITTER: Okser S
PROVIDER: S-EPMC4230844 | biostudies-literature | 2014 Nov
REPOSITORIES: biostudies-literature
Publications
Regularized machine learning in the genetic prediction of complex traits.
Okser Sebastian S Pahikkala Tapio T Airola Antti A Salakoski Tapio T Ripatti Samuli S Aittokallio Tero T
PLoS genetics 20141113 11
Similar Datasets
Project description:The usefulness of genomic prediction in crop and livestock breeding programs has prompted efforts to develop new and improved genomic prediction algorithms, such as artificial neural networks and gradient tree boosting. However, the performance of these algorithms has not been compared in a systematic manner using a wide range of datasets and models. Using data of 18 traits across six plant species with different marker densities and training population sizes, we compared the performance of six linear and six non-linear algorithms. First, we found that hyperparameter selection was necessary for all non-linear algorithms and that feature selection prior to model training was critical for artificial neural networks when the markers greatly outnumbered the number of training lines. Across all species and trait combinations, no one algorithm performed best, however predictions based on a combination of results from multiple algorithms (i.e., ensemble predictions) performed consistently well. While linear and non-linear algorithms performed best for a similar number of traits, the performance of non-linear algorithms vary more between traits. Although artificial neural networks did not perform best for any trait, we identified strategies (i.e., feature selection, seeded starting weights) that boosted their performance to near the level of other algorithms. Our results highlight the importance of algorithm selection for the prediction of trait values.
| S-EPMC6829122 | biostudies-literature
Project description:Advances in high-throughput sequencing technologies have reduced the cost of genotyping dramatically and led to genomic prediction being widely used in animal and plant breeding, and increasingly in human genetics. Inspired by the efficient computing of linear mixed model and the accurate prediction of Bayesian methods, we propose a machine learning-based method incorporating cross-validation, multiple regression, grid search, and bisection algorithms named KAML that aims to combine the advantages of prediction accuracy with computing efficiency. KAML exhibits higher prediction accuracy than existing methods, and it is available at https://github.com/YinLiLin/KAML.
| S-EPMC7386246 | biostudies-literature
Project description:Gene expression profiles were generated from 199 primary breast cancer patients. Samples 1-176 were used in another study, GEO Series GSE22820, and form the training data set in this study. Sample numbers 200-222 form a validation set. This data is used to model a machine learning classifier for Estrogen Receptor Status. RNA was isolated from 199 primary breast cancer patients. A machine learning classifier was built to predict ER status using only three gene features.
2013-01-01 | E-GEOD-29210 | biostudies-arrayexpress
Project description:ObjectiveMetabolic syndrome (MetS) is a complex multifactorial disorder that considerably burdens healthcare systems. We aim to classify MetS using regularized machine learning models in the presence of the risk variants of GCKR, BUD13 and APOA5, and environmental risk factors.Materials and methodsA cohort study was conducted on 2,346 cases and 2,203 controls from eligible Tehran Cardiometabolic Genetic Study (TCGS) participants whose data were collected from 1999 to 2017. We used different regularization approaches [least absolute shrinkage and selection operator (LASSO), ridge regression (RR), elasticnet (ENET), adaptive LASSO (aLASSO), and adaptive ENET (aENET)] and a classical logistic regression (LR) model to classify MetS and select influential variables that predict MetS. Demographics, clinical features, and common polymorphisms in the GCKR, BUD13 and APOA5 genes of eligible participants were assessed to classify TCGS participant status in MetS development. The models' performance was evaluated by 10-repeated 10-fold crossvalidation. Various assessment measures of sensitivity, specificity, classification accuracy, and area under the receiver operating characteristic curve (AUC-ROC) and AUC-precision-recall (AUC-PR) curves were used to compare the models.ResultsDuring the follow-up period, 50.38% of participants developed MetS. The groups were not similar in terms of baseline characteristics and risk variants. MetS was significantly associated with age, gender, schooling years, body mass index (BMI), and alternate alleles in all the risk variants, as indicated by LR. A comparison of accuracy, AUCROC, and AUC-PR metrics indicated that the regularization models outperformed LR. Regularized machine learning models provided comparable classification performances, whereas the aLASSO model was more parsimonious and selected fewer predictors.ConclusionRegularized machine learning models provided more accurate and parsimonious MetS classifying models. These high-performing diagnostic models can lay the foundation for clinical decision support tools that use genetic and demographical variables to locate individuals at high risk for MetS.
| S-EPMC10542204 | biostudies-literature
Project description:The genetic analysis of complex traits does not escape the current excitement around artificial intelligence, including a renewed interest in "deep learning" (DL) techniques such as Multilayer Perceptrons (MLPs) and Convolutional Neural Networks (CNNs). However, the performance of DL for genomic prediction of complex human traits has not been comprehensively tested. To provide an evaluation of MLPs and CNNs, we used data from distantly related white Caucasian individuals (n ∼100k individuals, m ∼500k SNPs, and k = 1000) of the interim release of the UK Biobank. We analyzed a total of five phenotypes: height, bone heel mineral density, body mass index, systolic blood pressure, and waist-hip ratio, with genomic heritabilities ranging from ∼0.20 to 0.70. After hyperparameter optimization using a genetic algorithm, we considered several configurations, from shallow to deep learners, and compared the predictive performance of MLPs and CNNs with that of Bayesian linear regressions across sets of SNPs (from 10k to 50k) that were preselected using single-marker regression analyses. For height, a highly heritable phenotype, all methods performed similarly, although CNNs were slightly but consistently worse. For the rest of the phenotypes, the performance of some CNNs was comparable or slightly better than linear methods. Performance of MLPs was highly dependent on SNP set and phenotype. In all, over the range of traits evaluated in this study, CNN performance was competitive to linear models, but we did not find any case where DL outperformed the linear model by a sizable margin. We suggest that more research is needed to adapt CNN methodology, originally motivated by image analysis, to genetic-based problems in order for CNNs to be competitive with linear models.
| S-EPMC6218236 | biostudies-literature
Project description:Gene expression profiles were generated from 199 primary breast cancer patients. Samples 1-176 were used in another study, GEO Series GSE22820, and form the training data set in this study. Sample numbers 200-222 form a validation set. This data is used to model a machine learning classifier for Estrogen Receptor Status.
2013-01-01 | GSE29210 | GEO
Project description:Genomic wide selection (GWS) is one contributions of molecular genetics to breeding. Machine learning (ML) and artificial neural networks (ANN) methods are non-parameterized and can develop more accurate and parsimonious models for GWS analysis. Multivariate Adaptive Regression Splines (MARS) is considered one of the most flexible ML methods, automatically modeling nonlinearities and interactions of the predictor variables. This study aimed to evaluate and compare methods based on ANN, ML, including MARS, and G-BLUP through GWS. An F2 population formed by 1000 individuals and genotyped for 4010 SNP markers and twelve traits from a model considering epistatic effect, with QTL numbers ranging from eight to 480 and heritability ( h2 ) of 0.3, 0.5 or 0.8 were simulated. Variation in heritability and number of QTL impacts the performance of methods. About quantitative traits (40, 80, 120, 240, and 480 QTLs) was observed highest R2 to Radial Base Network (RBF) and G-BLUP, followed by Random Forest (RF), Bagging (BA), and Boosting (BO). RF and BA also showed better results for traits to h2 of 0.3 with R2 values 16.51% and 16.30%, respectively, while MARS methods showed better results for oligogenic traits with R2 values ranging from 39,12 % to 43,20 % in h2 of 0.5 and from 59.92% to 78,56% in h2 of 0.8. Non-additive MARS methods also showed high R2 for traits with high heritability and 240 QTLs or more. ANN and ML methods are powerful tools to predict genetic values in traits with epistatic effect, for different degrees of heritability and QTL numbers.
| S-EPMC9547190 | biostudies-literature
Project description:Machine-learning techniques have helped solve a broad range of prediction problems, yet are not widely used to build polygenic risk scores for the prediction of complex traits. We propose a novel heuristic based on machine-learning techniques (GraBLD) to boost the predictive performance of polygenic risk scores. Gradient boosted regression trees were first used to optimize the weights of SNPs included in the score, followed by a novel regional adjustment for linkage disequilibrium. A calibration set with sample size of ~200 individuals was sufficient for optimal performance. GraBLD yielded prediction R 2 of 0.239 and 0.082 using GIANT summary association statistics for height and BMI in the UK Biobank study (N = 130 K; 1.98 M SNPs), explaining 46.9% and 32.7% of the overall polygenic variance, respectively. For diabetes status, the area under the receiver operating characteristic curve was 0.602 in the UK Biobank study using summary-level association statistics from the DIAGRAM consortium. GraBLD outperformed other polygenic score heuristics for the prediction of height (p < 2.2 × 10-16) and BMI (p < 1.57 × 10-4), and was equivalent to LDpred for diabetes. Results were independently validated in the Health and Retirement Study (N = 8,292; 688,398 SNPs). Our report demonstrates the use of machine-learning techniques, coupled with summary-level data from large genome-wide meta-analyses to improve the prediction of polygenic traits.
| S-EPMC5627249 | biostudies-literature
Project description:Biological Ecosystem Networks (BENs) are webs of biological species (nodes) establishing trophic relationships (links). Experimental confirmation of all possible links is difficult and generates a huge volume of information. Consequently, computational prediction becomes an important goal. Artificial Neural Networks (ANNs) are Machine Learning (ML) algorithms that may be used to predict BENs, using as input Shannon entropy information measures (Shk) of known ecosystems to train them. However, it is difficult to select a priori which ANN topology will have a higher accuracy. Interestingly, Auto Machine Learning (AutoML) methods focus on the automatic selection of the more efficient ML algorithms for specific problems. In this work, a preliminary study of a new approach to AutoML selection of ANNs is proposed for the prediction of BENs. We call it the Net-Net AutoML approach, because it uses for the first time Shk values of both networks involving BENs (networks to be predicted) and ANN topologies (networks to be tested). Twelve types of classifiers have been tested for the Net-Net model including linear, Bayesian, trees-based methods, multilayer perceptrons and deep neuronal networks. The best Net-Net AutoML model for 338,050 outputs of 10 ANN topologies for links of 69 BENs was obtained with a deep fully connected neuronal network, characterized by a test accuracy of 0.866 and a test AUROC of 0.935. This work paves the way for the application of Net-Net AutoML to other systems or ML algorithms.
| S-EPMC6098100 | biostudies-literature
Project description:The structure of genetic interaction networks predicts that, analogous to synthetic lethal interactions between non-essential genes, combinations of compounds with latent activities may exhibit potent synergism. To test this hypothesis, we generated a chemical-genetic matrix of 195 diverse yeast deletion strains treated with 4,915 compounds. This approach uncovered 1,221 genotype-specific inhibitors, which we termed cryptagens. Synergism between 8,128 structurally disparate cryptagen pairs was assessed experimentally and used to benchmark predictive algorithms. A model based on the chemical-genetic matrix and the genetic interaction network failed to accurately predict synergism. However, a combined random forest and Naive Bayesian learner that associated chemical structural features with genotype-specific growth inhibition had strong predictive power. This approach identified previously unknown compound combinations that exhibited species-selective toxicity toward human fungal pathogens. This work demonstrates that machine learning methods trained on unbiased chemical-genetic interaction data may be widely applicable for the discovery of synergistic combinations in different species.
| S-EPMC5998823 | biostudies-literature