MRFcov: Markov Random Fields with additional covariates in R
Releases | Reporting
Issues | Blogpost
MRFcov (described by Clark et al, published in Ecology Statistical
Reports)
provides R functions for approximating interaction parameters of nodes
in undirected Markov Random Fields (MRF) graphical networks. Models can
incorporate covariates (a class of models known as Conditional Random
Fields;
CRFs;
following methods developed by Cheng et al 2014 and Lindberg 2016),
allowing users to estimate how interactions between nodes are predicted
to change across covariate gradients. Note, this is a development
version. For the stable version, please download from CRAN
Why Use Conditional Random Fields?
In principle, MRFcov models that use species’ occurrences or
abundances as outcome variables are similar to Joint Species
Distribution
models
in that variance can be partitioned among abiotic and biotic effects.
However, key differences are that MRFcov models can:
-
Produce directly interpretable coefficients that allow users to determine the relative importances (i.e. effect sizes) of biotic associations and environmental covariates in driving abundances or occurrence probabilities
-
Identify association strengths, rather than simply determining whether they are “significantly different from zero”
-
Estimate how associations are predicted to change across environmental gradients
Models such as these are also better at isolating true species ‘interactions’ using presence-absence occurrence data than are traditional null model co-occurrence methods (such as the all-too-common null model randomisation approaches). See this blogpost for a more detailed explanation and proof of this statement.
MRF and CRF interaction parameters are approximated using separate regressions for individual species within a joint modelling framework. Because all combinations of covariates and additional species are included as predictor variables in node-specific regressions, variable selection is required to reduce overfitting and add sparsity. This is accomplished through LASSO penalization using functions in the glmnet package.
Installation
You can install the stable version of the MRFcov package into R from
CRAN.
Alternatively, install the development version (updated features but no
gurantees of good functionality) from GitHub using:
# install.packages("devtools")
devtools::install_github("nicholasjclark/MRFcov")Brief Overview
We can explore the model’s primary functions using a test dataset that
is available with the package. Load the Bird.parasites dataset, which
contains binary occurrences of four avian blood parasites in New
Caledonian Zosterops species (available in its original form at
Dryad; Clark et al 2016). A
single continuous covariate is also included (scale.prop.zos), which
reflects the relative abundance of Zosterops species among different
sample sites
library(MRFcov)
data("Bird.parasites")Visualise the dataset to see how analysis data needs to be structured.
In short, when estimating co-occurrence probabilities, node variable
(i.e. species) occurrences should be included as binary variables (1s
and 0s) as the left-most variables in data. Any covariates can be
included as the right-most variables. Note, these covariates should
ideally be on a similar scale, using the scale function for continuous
covariates (or similar) so that covariates generally have mean = 0 and
sd = 1
help("Bird.parasites")
View(Bird.parasites)You can read more about specific requirements of data formats (for example, one-hot encoding of categorical covariates) in the supplied vignette
vignette("CRF_data_prep")Running MRFs and visualising interaction coefficients
Run an MRF model using the provided continuous covariate
(scale.prop.zos). Here, each species-specific regression will be
individually optimised through cross-validated LASSO variable selection.
Corresponding coefficients (e.g. the coefficient for effect of species A
on species B and the coefficient for effect of species B on species A)
will be symmetrised to form an undirected MRF
graph
MRF_mod <- MRFcov(data = Bird.parasites, n_nodes = 4, family = 'binomial')
#> Leave-one-out cv used for the following low-occurrence (rare) nodes:
#> Microfilaria ...
#> Fitting MRF models in sequence using 1 core ...Visualise the estimated species interaction coefficients as a heatmap.
These represent mean interactions and are very useful for identifying
co-occurrence patterns, but they do not indicate how interactions change
across gradients. Note, for binary data such as this, we can also plot
the observed occurrences and co-occurrences using plot_observed_vals = TRUE
plotMRF_hm(MRF_mod, plot_observed_vals = TRUE, data = Bird.parasites)
Exploring regression coefficients and interpreting results
We can explore regression coefficients to get a better understanding of
just how important interactions are for predicting species’ occurrence
probabilities (in comparison to other covariates). This is perhaps the
strongest property of conditional MRFs, as competing methods (such as
Joint Species Distribution Models) do not provide interpretable
mechanisms for comparing the relative importances of interactions and
fixed covariates. MRF functions conveniently return a matrix of
important coefficients for each node in the graph, as well as their
relative importances (calculated using the formula B^2 / sum(B^2),
where the vector of Bs represents regression coefficients for
predictor variables). Variables with an underscore (_) indicate an
interaction between a covariate and another node, suggesting that
conditional dependencies of the two nodes vary across environmental
gradients
MRF_mod$key_coefs$Hzosteropis
#> Variable Rel_importance Standardised_coef Raw_coef
#> 1 Hkillangoi 0.64623474 -2.3087824 -2.3087824
#> 5 scale.prop.zos_Microfilaria 0.12980415 -1.0347421 -1.0347421
#> 3 Microfilaria 0.10143149 0.9146907 0.9146907
#> 4 scale.prop.zos 0.09788426 -0.8985542 -0.8985542
#> 2 Plas 0.01785290 -0.3837446 -0.3837446MRF_mod$key_coefs$Hkillangoi
#> Variable Rel_importance Standardised_coef Raw_coef
#> 1 Hzosteropis 0.79853150 -2.3087824 -2.3087824
#> 2 Microfilaria 0.11897509 -0.8911791 -0.8911791
#> 3 scale.prop.zos 0.08154704 -0.7378041 -0.7378041MRF_mod$key_coefs$Plas
#> Variable Rel_importance Standardised_coef Raw_coef
#> 2 Microfilaria 0.63590587 1.8658732 1.8658732
#> 3 scale.prop.zos 0.24611774 -1.1607994 -1.1607994
#> 5 scale.prop.zos_Microfilaria 0.07969128 0.6605278 0.6605278
#> 1 Hzosteropis 0.02689758 -0.3837446 -0.3837446
#> 4 scale.prop.zos_Hzosteropis 0.01023366 -0.2367016 -0.2367016MRF_mod$key_coefs$Microfilaria
#> Variable Rel_importance Standardised_coef Raw_coef
#> 3 Plas 0.4423652 1.8658732 1.8658732
#> 4 scale.prop.zos 0.1589327 -1.1184028 -1.1184028
#> 5 scale.prop.zos_Hzosteropis 0.1360445 -1.0347421 -1.0347421
#> 1 Hzosteropis 0.1063078 0.9146907 0.9146907
#> 2 Hkillangoi 0.1009129 -0.8911791 -0.8911791
#> 6 scale.prop.zos_Plas 0.0554369 0.6605278 0.6605278To work through more in-depth tutorials and examples, see the vignettes in the package and check out papers that have been published using the method
vignette("Bird_Parasite_CRF")
vignette("Gaussian_Poisson_CRFs")Clark et al 2018 Ecology | PDF
Peel et al 2019 Emerging Microbes & Infections
Fountain-Jones et al 2019 Journal of Animal Ecology
Clark et al 2020 Transboundary and Emerging Diseases
Clark et al 2020 Parasites & Vectors
Clark et al 2020 Nature Climate Change
References
Cheng, J., Levina, E., Wang, P. & Zhu, J. (2014). A sparse Ising model with covariates. Biometrics 70:943-953.
Clark, N.J., Wells, K., Lindberg, O. (2018). Unravelling changing interspecific interactions across environmental gradients using Markov random fields. Ecology DOI: https://doi.org/10.1002/ecy.2221
Clark, N.J., K. Wells, D. Dimitrov, and S.M. Clegg. (2016). Co-infections and environmental conditions drive the distributions of blood parasites in wild birds. Journal of Animal Ecology 85:1461-1470. Blogpost | PDF
Clark, N.J., S. Tozer, C. Wood, S.M. Firestone, M. Stevenson, C. Caraguel, A.L. Chaber, J. Heller, R.J. Soares Magalhães. 2020. Unravelling animal exposure profiles of human Q fever cases in Queensland, Australia using natural language processing. Transboundary and Emerging Diseases DOI: https://doi.org/10.1111/tbed.13565.
Clark, N.J., K. Owada, E. Ruberanziza, G. Ortu, I. Umulisa, U. Bayisenge, J.B. Mbonigaba, J.B. Mucaca, W. Lancaster, A. Fenwick, R.J. Soares Magalhães, A. Mbituyumuremyi. 2020. Parasite associations predict infection risk: incorporating co-infections in predictive models for neglected tropical diseases. Parasites & Vectors 13:1-16.
Clark, N.J., J.T. Kerry, C.I. Fraser. 2020. Rapid winter warming could disrupt coastal marine fish community structure. Nature Climate Change DOI: https://doi.org/10.1038/s41558-020-0838-5
Fountain‐Jones, N.M., N.J. Clark, A.C. Kinsley, M. Carstensen, J. Forester, T.J. Johnson, E. Miller, S. Moore, T.M. Wolf, M.E. Craft. 2019. Microbial associations and spatial proximity predict North American moose (Alces alces) gastrointestinal community composition. Journal of Animal Ecology 89:817-828.
Lindberg, O. (2016). Markov Random Fields in Cancer Mutation Dependencies. Master’s of Science Thesis. University of Turku, Turku, Finland.
Peel, A.J., K. Wells, J. Giles, V. Boyd, A. Burroughs, D. Edson, G. Crameri, M. L. Baker, H. Field, L-F. Wang, H. McCallum, R. K. Plowright, N. Clark. 2019. Synchronous shedding of multiple bat paramyxoviruses coincides with peak periods of Hendra virus spillover. Emerging Microbes & Infections 8:1314-1323
This project is licensed under the terms of the GNU General Public License (GNU GPLv3)