Benchmarking phyloregion with comparable packages

In this vignette, we benchmark phyloregion against other similar R packages in analyses of standard alpha diversity metrics commonly used in conservation, such as phylogenetic diversity and phylogenetic endemism as well as metrics for analyzing compositional turnover (e.g., beta diversity and phylogenetic beta diversity). Specifically, we compare phyloregion’s functions with available packages for efficiency in memory allocation and computation speed in various biogeographic analyses.

First, load the packages for the benchmarking:

library(ape)
library(Matrix)
library(bench)
library(ggplot2)
# packages we benchmark
library(phyloregion)
library(betapart)
library(picante)
library(vegan)
library(hilldiv)
library(BAT)
library(pez)

We will use a small data set which comes with phyloregion. This dataset consists of a dated phylogeny of the woody plant species of southern Africa along with their geographical distributions. The dataset comes from a study that maps tree diversity hotspots in southern Africa.1

data(africa)
# subset matrix
X_sparse <- africa$comm[1:30, ]
X_sparse <- X_sparse[, colSums(X_sparse)>0]
X_dense <- as.matrix(X_sparse)
Xt_dense <- t(X_dense)

object.size(X_sparse)
## 76504 bytes
object.size(X_dense)
## 134752 bytes
dim(X_sparse)
## [1]  30 401

To make results comparable, it is often desirable to make sure that the taxa in different datasets match each other.2 For example, the community matrix in the hilldiv package3 needs to be transposed. These transformations can influence the execution times of the function, often only marginally. Thus, to benchmark phyloregion against other packages, we here use the package bench4 because it returns execution times and provides estimates of memory allocations for each computation.

1. Analysis of alpha diversity

1.1. Benchmarking phyloregion for analysis of phylogenetic diversity

For analysis of alpha diversity commonly used in conservation such as phylogenetic diversity - the sum of all phylogenetic branch lengths within an area5 - phyloregion is 31 to 284 times faster and 67 to 192 times memory efficient, compared to other packages!

tree <- africa$phylo
tree <- keep.tip(tree, colnames(X_sparse))

pd_picante <- function(x, tree){
    res <- picante::pd(x, tree)[,1]
    names(res) <- row.names(x)
    res
}

pd_pez <- function(x, tree){
    dat <- pez::comparative.comm(tree, x)
    res <- pez::.pd(dat)[,1]
    names(res) <- row.names(x)
    res
}

pd_hilldiv <- function(x, tree) hilldiv::index_div(x, tree, index="faith")
pd_phyloregion <- function(x, tree) phyloregion::PD(x, tree)

res1 <- bench::mark(picante=pd_picante(X_dense, tree),
          hilldiv=pd_hilldiv(Xt_dense,tree=tree),
          pez=pd_pez(X_dense, tree),
          phyloregion=pd_phyloregion(X_sparse, tree))
summary(res1)
## # A tibble: 4 x 6
##   expression       min   median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
## 1 picante     118.88ms 125.72ms     7.34    59.62MB    12.8 
## 2 hilldiv         1.1s     1.1s     0.911  170.22MB     4.56
## 3 pez         125.34ms 139.49ms     7.27    60.89MB    14.5 
## 4 phyloregion   3.21ms   4.61ms   142.       1.86MB     4.00
autoplot(res1)

plot of chunk phylo_diversity

1.2. Benchmarking phyloregion for analysis of phylogenetic endemism

Another benchmark for phyloregion is in the analysis of phylogenetic endemism, the degree to which phylogenetic diversity is restricted to any given area.6 Here, we found that phyloregion is 160 times faster and 489 times efficient in memory allocation.

tree <- africa$phylo
tree <- keep.tip(tree, colnames(X_sparse))

pe_pez <- function(x, tree){
    dat <- pez::comparative.comm(tree, x)
    res <- pez::pez.endemism(dat)[,1]
    names(res) <- row.names(x)
    res
}

pe_phyloregion <- function(x, tree) phyloregion::phylo_endemism(x, tree)

res2 <- bench::mark(pez=pe_pez(X_dense, tree),
          phyloregion=pe_phyloregion(X_sparse, tree))
summary(res2)
## # A tibble: 2 x 6
##   expression       min   median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
## 1 pez            1.38s    1.38s     0.725  493.84MB     4.35
## 2 phyloregion   3.24ms   3.47ms   265.       1.07MB     1.99
autoplot(res2)

plot of chunk phylo_endemism

2. Analysis of compositional turnover (beta diversity)

2.1. Benchmarking phyloregion for analysis of taxonomic beta diversity

For analysis of taxonomic beta diversity, which compares diversity between communities,7 phyloregion has marginal advantage over other packages. Nonetheless, it is 1-39 times faster and allocates 2 to 110 times less memory than other packages.

chk_fun <- function(target, current)
    all.equal(target, current, check.attributes = FALSE)

fun_phyloregion <- function(x) as.matrix(phyloregion::beta_diss(x)[[3]])
fun_betapart <- function(x) as.matrix(betapart::beta.pair(x)[[3]])
fun_vegan  <- function(x) as.matrix(vegan::vegdist(x, binary=TRUE))
fun_BAT <- function(x) as.matrix(BAT::beta(x, func = "Soerensen")[[1]])
res3 <- bench::mark(phyloregion=fun_phyloregion(X_sparse),
                    betapart=fun_betapart(X_dense),
                    vegan=fun_vegan(X_dense),
                    BAT=fun_BAT(X_dense), check=chk_fun)
summary(res3)
## # A tibble: 4 x 6
##   expression       min   median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
## 1 phyloregion  753.1µs  929.9µs     903.    404.1KB     2.10
## 2 betapart     863.1µs    924µs     998.    594.1KB     4.38
## 3 vegan        945.8µs      1ms     946.    993.4KB     7.05
## 4 BAT           38.3ms   44.3ms      23.6    31.8MB    15.7
autoplot(res3)

plot of chunk beta_diversity

2.2. Benchmarking phyloregion for analysis of phylogenetic beta diversity

For analysis of phylogenetic turnover (beta-diversity) among communities - the proportion of shared phylogenetic branch lengths between communities8 - phyloregion is 300-400 times faster and allocates 100-600 times less memory!

fun_phyloregion <- function(x, tree) phyloregion::phylobeta(x, tree)[[3]]
fun_betapart <- function(x, tree) betapart::phylo.beta.pair(x, tree)[[3]]
fun_picante <- function(x, tree) 1 - picante::phylosor(x, tree)
fun_BAT <- function(x, tree) BAT::beta(x, tree, func = "Soerensen")[[1]]

chk_fun <- function(target, current)
    all.equal(target, current, check.attributes = FALSE)

res4 <- bench::mark(picante=fun_picante(X_dense, tree),
                       betapart=fun_betapart(X_dense, tree),
                       BAT=fun_BAT(X_dense, tree),
                       phyloregion=fun_phyloregion(X_sparse, tree), check=chk_fun)
summary(res4)
## # A tibble: 4 x 6
##   expression       min   median `itr/sec` mem_alloc `gc/sec`
##   <bch:expr>  <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
## 1 picante        1.86s    1.86s     0.539    1.24GB    1.62 
## 2 betapart       1.83s    1.83s     0.545    1.24GB    2.18 
## 3 BAT            1.42s    1.42s     0.703  208.99MB    0.703
## 4 phyloregion   4.34ms   4.52ms   215.       1.15MB    0
autoplot(res4)

plot of chunk phylobeta

Note that for this test, picante returns a similarity matrix while betapart, and phyloregion return a dissimilarity matrix.

Session Information

## R version 4.0.2 (2020-06-22)
## Platform: x86_64-apple-darwin17.0 (64-bit)
## Running under: macOS Catalina 10.15.6
## 
## Matrix products: default
## BLAS:   /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
## LAPACK: /Library/Frameworks/R.framework/Versions/4.0/Resources/lib/libRlapack.dylib
## 
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
##  [1] pez_1.2-0         BAT_2.0.1         hilldiv_1.5.1     picante_1.8.2    
##  [5] nlme_3.1-148      vegan_2.5-6       lattice_0.20-41   permute_0.9-5    
##  [9] betapart_1.5.1    phyloregion_1.0.4 ggplot2_3.3.2     bench_1.1.1      
## [13] Matrix_1.2-18     ape_5.4           knitr_1.29       
## 
## loaded via a namespace (and not attached):
##   [1] utf8_1.1.4              ks_1.11.7               tidyselect_1.1.0       
##   [4] htmlwidgets_1.5.1       grid_4.0.2              combinat_0.0-8         
##   [7] munsell_0.5.0           animation_2.6           codetools_0.2-16       
##  [10] miniUI_0.1.1.1          withr_2.2.0             profmem_0.5.0          
##  [13] colorspace_1.4-1        highr_0.8               rstudioapi_0.11        
##  [16] geometry_0.4.5          stats4_4.0.2            tensor_1.5             
##  [19] ggsignif_0.6.0          huge_1.3.4.1            FD_1.0-12              
##  [22] nls2_0.2                polyclip_1.10-0         mnormt_2.0.1           
##  [25] farver_2.0.3            rprojroot_1.3-2         coda_0.19-3            
##  [28] vctrs_0.3.1             generics_0.0.2          clusterGeneration_1.3.4
##  [31] xfun_0.15               fastcluster_1.1.25      R6_2.4.1               
##  [34] ggbeeswarm_0.6.0        pdist_1.2               manipulateWidget_0.10.1
##  [37] spatstat.utils_1.17-0   assertthat_0.2.1        promises_1.1.1         
##  [40] scales_1.1.1            nnet_7.3-14             beeswarm_0.2.3         
##  [43] rgeos_0.5-3             gtable_0.3.0            caper_1.0.1            
##  [46] goftest_1.2-2           processx_3.4.3          phangorn_2.5.5         
##  [49] MatrixModels_0.4-1      rlang_0.4.6             FSA_0.8.30             
##  [52] scatterplot3d_0.3-41    splines_4.0.2           rstatix_0.6.0          
##  [55] acepack_1.4.1           broom_0.7.0             checkmate_2.0.0        
##  [58] rgl_0.100.54            yaml_2.2.1              reshape2_1.4.4         
##  [61] abind_1.4-5             d3Network_0.5.2.1       crosstalk_1.1.0.1      
##  [64] backports_1.1.8         httpuv_1.5.4            Hmisc_4.4-0            
##  [67] tools_4.0.2             psych_1.9.12.31         lavaan_0.6-6           
##  [70] cubature_2.0.4.1        ellipsis_0.3.1          raster_3.3-7           
##  [73] RColorBrewer_1.1-2      Rcpp_1.0.5              plyr_1.8.6             
##  [76] base64enc_0.1-3         progress_1.2.2          purrr_0.3.4            
##  [79] ps_1.3.3                prettyunits_1.1.1       deldir_0.1-25          
##  [82] ggpubr_0.4.0            rpart_4.1-15            pbapply_1.4-2          
##  [85] deSolve_1.28            qgraph_1.6.5            haven_2.3.1            
##  [88] cluster_2.1.0           fs_1.4.2                magrittr_1.5           
##  [91] data.table_1.12.8       openxlsx_4.1.5          SparseM_1.78           
##  [94] tmvnsim_1.0-2           mvtnorm_1.1-1           whisker_0.4            
##  [97] hms_0.5.3               mime_0.9                evaluate_0.14          
## [100] xtable_1.8-4            rio_0.5.16              jpeg_0.1-8.1           
## [103] mclust_5.4.6            readxl_1.3.1            gridExtra_2.3          
## [106] compiler_4.0.2          tibble_3.0.2            maps_3.3.0             
## [109] KernSmooth_2.23-17      crayon_1.3.4            hypervolume_2.0.12     
## [112] htmltools_0.5.0         mgcv_1.8-31             corpcor_1.6.9          
## [115] later_1.1.0.1           Formula_1.2-3           tidyr_1.1.0            
## [118] expm_0.999-4            apTreeshape_1.5-0       magic_1.5-9            
## [121] subplex_1.6             MASS_7.3-51.6           ade4_1.7-15            
## [124] car_3.0-8               cli_2.0.2               quadprog_1.5-8         
## [127] parallel_4.0.2          igraph_1.2.5            BDgraph_2.62           
## [130] forcats_0.5.0           pkgconfig_2.0.3         pkgdown_1.5.1          
## [133] numDeriv_2016.8-1.1     foreign_0.8-80          sp_1.4-2               
## [136] pbivnorm_0.6.0          vipor_0.4.5             webshot_0.5.2          
## [139] stringr_1.4.0           callr_3.4.3             digest_0.6.25          
## [142] phytools_0.7-47         rcdd_1.2-2              spatstat.data_1.4-3    
## [145] rmarkdown_2.3           cellranger_1.1.0        fastmatch_1.1-0        
## [148] htmlTable_2.0.1         curl_4.3                quantreg_5.55          
## [151] shiny_1.5.0             gtools_3.8.2            rjson_0.2.20           
## [154] geiger_2.0.7            lifecycle_0.2.0         glasso_1.11            
## [157] jsonlite_1.7.0          carData_3.0-4           desc_1.2.0             
## [160] fansi_0.4.1             pillar_1.4.6            fastmap_1.0.1          
## [163] httr_1.4.1              plotrix_3.7-8           survival_3.2-3         
## [166] glue_1.4.1              spatstat_1.64-1         zip_2.0.4              
## [169] fdrtool_1.2.15          png_0.1-7               class_7.3-17           
## [172] stringi_1.4.6           rematch2_2.1.2          latticeExtra_0.6-29    
## [175] memoise_1.1.0           dplyr_1.0.0             e1071_1.7-3

REFERENCES

1. Daru, B. H., Bank, M. van der & Davies, T. J. Spatial incongruence among hotspots and complementary areas of tree diversity in southern africa. Diversity and Distributions 21, 769–780 (2015).

2. Kembel, S. W. et al. Picante: R tools for integrating phylogenies and ecology. Bioinformatics 26, 1463–1464 (2010).

3. Alberdi, A. Hilldiv: An r package for the integral analysis of diversity based on hill numbers. bioRxiv (2019).

4. Hester, J. Bench: High precision timing of r expressions. (2020).

5. Faith, D. P. Conservation evaluation and phylogenetic diversity. Biological Conservation 61, 1–10 (1992).

6. Rosauer, D., Laffan, S. W., Crisp, M. D., Donnellan, S. C. & Cook, L. G. Phylogenetic endemism: A new approach for identifying geographical concentrations of evolutionary history. Molecular Ecology 18, 4061–4072 (2009).

7. Koleff, P., Gaston, K. J. & Lennon, J. J. Measuring beta diversity for presence–absence data. Journal of Animal Ecology 72, 367–382 (2003).

8. Graham, C. H. & Fine, P. V. A. Phylogenetic beta diversity: Linking ecological and evolutionary processes across space in time. Ecology Letters 11, 1265–1277 (2008).