Rob Bergsma

Rob Bergsma

Auto-generated profile, based on publicly available information

Rob Bergsma

Topigs Norsvin Research Center B.V., Beuningen, The Netherlands

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
22 February 2021 in Frontiers in Genetics
Verified
Journal Article
Open access
PDF file
Miriam PilesMiriam Piles
Miriam Piles
RBRob Bergsma
Rob Bergsma
DGDaniel Gianola
Daniel Gianola
HGHélène Gilbert
Hélène Gilbert
LTLlibertat Tusell
Llibertat Tusell
Feature selection (FS, i.e., selection of a subset of predictor variables) is essential in high-dimensional datasets to prevent overfitting of prediction/classification models and reduce computation time and resources. In genomics, FS allows identifying relevant markers and designing low-density SNP chips to evaluate selection candidates. In this research, several univariate and multivariate FS algorithms combined with various parametric and non-parametric learners were applied to the prediction of feed efficiency in growing pigs from high-dimensional genomic data. The objective was to find the best combination of feature selector, SNP subset size, and learner leading to accurate and stable (i.e., less sensitive to changes in the training data) prediction models. Genomic best linear unbiased prediction (GBLUP) without SNP pre-selection was the benchmark. Three types of FS methods were implemented: (i) filter methods: univariate (univ.dtree, spearcor) or multivariate (cforest, mrmr), with random selection as benchmark; (ii) embedded methods: elastic net and least absolute shrinkage and selection operator (LASSO) regression; (iii) combination of filter and embedded methods. Ridge regression, support vector machine (SVM), and gradient boosting (GB) were applied after pre-selection performed with the filter methods. Data represented 5,708 individual records of residual feed intake to be predicted from the animal’s own genotype. Accuracy (stability of results) was measured as the median (interquartile range) of the Spearman correlation between observed and predicted data in a 10-fold cross-validation. The best prediction in terms of accuracy and stability was obtained with SVM and GB using 500 or more SNPs [0.28 (0.02) and 0.27 (0.04) for SVM and GB with 1,000 SNPs, respectively]. With larger subset sizes (1,000–1,500 SNPs), the filter method had no influence on prediction quality, which was similar to that attained with a random selection. With 50–250 SNPs, the FS method had a huge impact on prediction quality: it was very poor for tree-based methods combined with any learner, but good and similar to what was obtained with larger SNP subsets when spearcor or mrmr were implemented with or without embedded methods. Those filters also led to very stable results, suggesting their potential use for designing low-density SNP chips for genome-based evaluation of feed efficiency.

23 October 2017 in Genetics Selection Evolution
Verified
Journal Article
Open access
PDF file
CSClaudia A. Sevillano
Claudia A. Sevillano
JVJeremie Vandenplas
Jeremie Vandenplas
JBJohn W. M. Bastiaansen
John W. M. Bastiaansen
RBRob Bergsma
Rob Bergsma
MCMario Pl L. Calus
Mario Pl L. Calus
Genomic prediction of purebred animals for crossbred performance can be based on a model that estimates effects of single nucleotide polymorphisms (SNPs) in purebreds on crossbred performance. For crossbred performance, SNP effects might be breed-specific due to differences between breeds in allele frequencies and linkage disequilibrium patterns between SNPs and quantitative trait loci. Accurately tracing the breed-of-origin of alleles (BOA) in three-way crosses is possible with a recently developed procedure called BOA. A model that accounts for breed-specific SNP effects (BOA model), has never been tested empirically on a three-way crossbreeding scheme. Therefore, the objectives of this study were to evaluate the estimates of variance components and the predictive accuracy of the BOA model compared to models in which SNP effects for crossbred performance were assumed to be the same across breeds, using either breed-specific allele frequencies ( $${\text{G}}_{\text{A}}$$ model) or allele frequencies averaged across breeds ( $${\text{G}}_{\text{B}}$$ model). In this study, we used data from purebred and three-way crossbred pigs on average daily gain (ADG), back fat thickness (BF), and loin depth (LD). Estimates of variance components for crossbred performance from the BOA model were mostly similar to estimates from models $${\text{G}}_{\text{A}}$$ and $${\text{G}}_{\text{B}}$$ . Heritabilities for crossbred performance ranged from 0.24 to 0.46 between traits. Genetic correlations between purebred and crossbred performance ( $${\text{r}}_{\text{pc}}$$ ) across breeds ranged from 0.30 to 0.62 for ADG and from 0.53 to 0.74 for BF and LD. For ADG, prediction accuracies of the BOA model were higher than those of the $${\text{G}}_{\text{A}}$$ and $${\text{G}}_{\text{B}}$$ models, with significantly higher accuracies only for one maternal breed. For BF and LD, prediction accuracies of models $${\text{G}}_{\text{A}}$$ and $${\text{G}}_{\text{B}}$$ were higher than those of the BOA model, with no significant differences. Across all traits, models $${\text{G}}_{\text{A}}$$ and $${\text{G}}_{\text{B}}$$ yielded similar predictions. The BOA model yielded a higher prediction accuracy for ADG in one maternal breed, which had the lowest $${\text{r}}_{\text{pc}}$$ (0.30). Using the BOA model was especially relevant for traits with a low $${\text{r}}_{\text{pc}}$$ . In all other cases, the use of crossbred information in models $${\text{G}}_{\text{A}}$$ and $${\text{G}}_{\text{B}}$$ , does not jeopardize predictions and these models are more easily implemented than the BOA model.

17 September 2019 in Journal of Animal Breeding and Genetics
Verified
Journal Article
Open access
PDF file
LVLisanne M. G. Verschuren
Lisanne M. G. Verschuren
DSDirkjan Schokker
Dirkjan Schokker
RBRob Bergsma
Rob Bergsma
AJAlfons J. M. Jansman
Alfons J. M. Jansman
and Francesc Molistand Francesc Molist
and Francesc Molist
MCMario Pl L. Calus
Mario Pl L. Calus
Microbiota play an important role in total tract nutrient digestion, especially when fibrous diets are fed to pigs. This study aimed to use metagenomics to predict faecal nutrient digestibility in grower‐finisher pigs. The study design consisted of 160 three‐way crossbreed grower‐finisher pigs (80 female and 80 male) which were either fed a diet based on corn/soybean meal or a more fibrous diet based on wheat/barley/by‐products. On the day before slaughter, faecal samples were collected and used to determine faecal digestibility of dry matter, ash, organic matter, crude protein, crude fat, crude fibre and non‐starch polysaccharides. The faecal samples were also sequenced for the 16S hypervariable region of bacteria (V3/V4) to profile the faecal microbiome. With these data, we calculated the between‐animal variation in faecal nutrient digestibility associated with variation in the faecal microbiome, that is the “microbiability”. The microbiability values were significantly greater than zero for dry matter, organic matter, crude protein, crude fibre and non‐starch polysaccharides, ranging from 0.58 to 0.93, as well as for crude fat with a value of 0.37, but not significantly different from zero for ash. Using leave‐one‐out cross‐validation, we estimated the accuracy of predicting digestibility values of individual pigs based on their faecal microbiota composition. The accuracies of prediction for crude fat and ash digestibility were virtually 0, and for the other nutrients, the accuracies ranged from 0.42 to 0.63. In conclusion, the faecal microbiota composition gave high microbiability values for faecal digestibility of dry matter, organic matter, crude protein, crude fibre and non‐starch polysaccharides. The accuracies of prediction are relatively low if the interest is in precisely predicting faecal nutrient digestibility of individual pigs, but are promising from the perspective of ranking animals in a genetic selection context.

18 December 2020 in Frontiers in Genetics
Verified
Journal Article
Open access
PDF file
LTLlibertat Tusell
Llibertat Tusell
RBRob Bergsma
Rob Bergsma
HGHélène Gilbert
Hélène Gilbert
DGDaniel Gianola
Daniel Gianola
MPMiriam Piles
Miriam Piles
This research assessed the ability of a Support Vector Machine (SVM) regression model to predict pig crossbred (CB) performance from various sources of phenotypic and genotypic information for improving crossbreeding performance at reduced genotyping cost. Data consisted of average daily gain (ADG) and residual feed intake (RFI) records and genotypes of 5,708 purebred (PB) boars and 5,007 CB pigs. Prediction models were fitted using individual PB genotypes and phenotypes (trn.1); genotypes of PB sires and average of CB records per PB sire (trn.2); and individual CB genotypes and phenotypes (trn.3). The average of CB offspring records was the trait to be predicted from PB sire’s genotype using cross-validation. Single nucleotide polymorphisms (SNPs) were ranked based on the Spearman Rank correlation with the trait. Subsets with an increasing number (from 50 to 2,000) of the most informative SNPs were used as predictor variables in SVM. Prediction performance was the median of the Spearman correlation (SC, interquartile range in brackets) between observed and predicted phenotypes in the testing set. The best predictive performances were obtained when sire phenotypic information was included in trn.1 (0.22 [0.03] for RFI with SVM and 250 SNPs, and 0.12 [0.05] for ADG with SVM and 500–1,000 SNPs) or when trn.3 was used (0.29 [0.16] with Genomic best linear unbiased prediction (GBLUP) for RFI, and 0.15 [0.09] for ADG with just 50 SNPs). Animals from the last two generations were assigned to the testing set and remaining animals to the training set. Individual’s PB own phenotype and genotype improved the prediction ability of CB offspring of young animals for ADG but not for RFI. The highest SC was 0.34 [0.21] and 0.36 [0.22] for RFI and ADG, respectively, with SVM and 50 SNPs. Predictive performance using CB data for training leads to a SC of 0.34 [0.19] with GBLUP and 0.28 [0.18] with SVM and 250 SNPs for RFI and 0.34 [0.15] with SVM and 500 SNPs for ADG. Results suggest that PB candidates could be evaluated for CB performance with SVM and low-density SNP chip panels after collecting their own RFI or ADG performances or even earlier, after being genotyped using a reference population of CB animals.

5 July 2023 in Journal of Animal Breeding and Genetics
Verified
Journal Article
Open access
MFMónica Mora Fenoll
Mónica Mora Fenoll
PGPablo González
Pablo González
JQJose Ramon Quevedo-Perez
Jose Ramon Quevedo-Perez
EMElena Montañés
Elena Montañés
LTLlibertat Tusell
Llibertat Tusell
Miriam PilesMiriam Piles
Miriam Piles
Feeding represents the largest economic cost in meat production; therefore, selection to improve traits related to feed efficiency is a goal in most livestock breeding programs. Residual feed intake (RFI), that is, the difference between the actual and the expected feed intake based on animal's requirements, has been used as the selection criteria to improve feed efficiency since it was proposed by Kotch in 1963. In growing pigs, it is computed as the residual of the multiple regression model of daily feed intake (DFI), on average daily gain (ADG), backfat thickness (BFT), and metabolic body weight (MW). Recently, prediction using single‐output machine learning algorithms and information from SNPs as predictor variables have been proposed for genomic selection in growing pigs, but like in other species, the prediction quality achieved for RFI has been generally poor. However, it has been suggested that it could be improved through multi‐output or stacking methods. For this purpose, four strategies were implemented to predict RFI. Two of them correspond to the computation of RFI in an indirect way using the predicted values of its components obtained from (i) individual (multiple single‐output strategy) or (ii) simultaneous predictions (multi‐output strategy). The other two correspond to the direct prediction of RFI using (iii) the individual predictions of its components as predictor variables jointly with the genotype (stacking strategy), or (iv) using only the genotypes as predictors of RFI (single‐output strategy). The single‐output strategy was considered the benchmark. This research aimed to test the former three hypotheses using data recorded from 5828 growing pigs and 45,610 SNPs. For all the strategies two different learning methods were fitted: random forest (RF) and support vector regression (SVR). A nested cross‐validation (CV) with an outer 10‐folds CV and an inner threefold CV for hyperparameter tuning was implemented to test all strategies. This scheme was repeated using as predictor variables different subsets with an increasing number (from 200 to 3000) of the most informative SNPs identified with RF. Results showed that the highest prediction performance was achieved with 1000 SNPs, although the stability of feature selection was poor (0.13 points out of 1). For all SNP subsets, the benchmark showed the best prediction performance. Using the RF as a learner and the 1000 most informative SNPs as predictors, the mean (SD) of the 10 values obtained in the test sets were: 0.23 (0.04) for the Spearman correlation, 0.83 (0.04) for the zero–one loss, and 0.33 (0.03) for the rank distance loss. We conclude that the information on predicted components of RFI (DFI, ADG, MW, and BFT) does not contribute to improve the quality of the prediction of this trait in relation to the one obtained with the single‐output strategy.

28 June 2022 in Genetics Selection Evolution
Verified
Journal Article
Open access
PDF file
IJIbrahim Jibrila
Ibrahim Jibrila
JVJeremie Vandenplas
Jeremie Vandenplas
JNJan Ten Napel
Jan Ten Napel
RBRob Bergsma
Rob Bergsma
RVRoel F. Veerkamp
Roel F. Veerkamp
MCMario Pl L. Calus
Mario Pl L. Calus
Background: Empirically assessing the impact of preselection on genetic evaluation of preselected animals requires comparing scenarios that take different approaches into account, including scenarios without preselection. However, preselection is almost always performed in animal breeding programs, so it is difficult to have a dataset without preselection. Hence, most studies on preselection have used simulated datasets, and have concluded that genomic estimated breeding values (GEBV) from subsequent single-step genomic best linear unbiased prediction (ssGBLUP) evaluations are unbiased. The aim of this study was to investigate the impact of genomic preselection (GPS) on accuracy and bias in subsequent ssGBLUP evaluations, using data from a commercial pig breeding program. Methods: We used data on average daily gain during performance testing, average daily gain throughout life, backfat thickness, and loin depth from one sire line and one dam line of pigs. As these traits have different weights in the breeding goals of the two lines, we analyzed the lines separately. For each line, we implemented a reference GPS scenario that kept all available data, against which the next two scenarios were compared. We then implemented two other scenarios with additional layers of GPS by removing all animals without progeny either (i) only in the validation generation, or (ii) in all generations. We conducted subsequent ssGBLUP evaluations for each GPS scenario, using all the data remaining after implementing the GPS scenario. Accuracy and bias were computed by comparing GEBV against progeny yield deviations of validation animals. Results: Results for all traits and in both lines showed a marginal loss in accuracy due to the additional layers of GPS. Average accuracies across all GPS scenarios in the two lines were 0.39, 0.47, 0.56, and 0.60, for average daily gain during performance testing and throughout life, backfat thickness, and loin depth, respectively. Biases were largely absent, and when present, did not differ greatly between the GPS scenarios. Conclusions: We conclude that the impact of preselection on accuracy and bias in subsequent ssGBLUP evaluations of selection candidates in pigs is generally minimal. We expect this conclusion to apply for other animal breeding programs as well, since preselection of any type or intensity generally has the same effect in animal breeding programs.