Bayesian MCMC Latent-class Analysis of Parasitic Diagnostic Tests in the Philippines

This repository contains the code and figures generated by the Latent-Class Analysis (LCA) discussed in the paper Reassessing The Use of Microscopy-Based Techniques, by Vachel Gay V. Paller¹, Frederick T. A. Freeth², Allen Jethro I. Alonte¹, Vicente Y. Belizario Jr.³, Billy P. Divina⁴, Sheina Macy P. Manalo⁴, Martha E. Betson², and Joaquin M. Prada². Paper currently submitted and under review.

¹ Institute of Biological Sciences, College of Arts and Sciences, University of the Philippines Los Baños, Laguna, the Philippines,
² School of Veterinary Medicine, University of Surrey, Guildford, Surrey, United Kingdom,
³ Department of Parasitology, College of Public Health, University of the Philippines Manila, Manila, the Philippines,
⁴ Department of Veterinary Paraclinical Sciences, College of Veterinary Medicine, University of the Philippines Los Baños, Laguna, the Philippines.

Data

The files Data_Humans.csv and Data_Animals.csv contain the neccessary data used for the analysis. It contains the data for the status of a host. In the animals, 1 denotes the host being infected with a 0 implying undetectable levels of infection. In humans, raw fecal egg counts have been given.

Algorithms and Supplementary Figures

The file LatentClassAnalysisFTAF.R contains the source code that implements the Markov Chain Monte Carlo analysis and produces the results. It also contains the JAGS models for the humans and animals to be run with the data.

Informative Prior Calibration Data

Informative priors were used to ensure model convergence and minimal autocorrelation between samples. We least-squares fit the 25%, 50% and 75% quantiles of raw egg counts to the shape sh and rate rt parameters of a gamma distribution using R's optim() function in the stats package. This gamma distribution represents the infection intensity experienced by each host and is the parameter lambda in the JAGS model. The values of these quantiles were sourced from an independent unpublished study. In cases where suitable quantiles for the egg counts were not available, e.g. S. Japonicum across all hosts, the shape and rate parameters for the gamma distribution were supplied as values predetermined from an unpublished, independent study. The table below summarises the parameters used for each host and diagnostic method.

Species	Ascaris	Trichuris	Hookworm / Strongyles	Schistosomiasis
Humans	25%: 0, 50%: 30.49, 75%: 0	Same as in Cats and Dogs	Same as in Cats and Dogs	sh = 0.5775932 rt = 0.07099037
Cats and Dogs	Same as in Humans	25%: 0, 50%: 4.417, 75%: 1.250	25%: 1.75, 50%: 3.5, 75%: 5.25	Same as in Humans
Pigs	25%: 0, 50%: 0.5, 75%: 1	Same as in Water Buffalo and Cattle	25%: 3.75, 50%: 3.5, 75%: 5.25	Same as in Humans
Water Buffalo and Cattle	25%: 0, 50%: 0.1892, 75%: 0	25%: 0, 50%: 0.1351, 75%: 0	25%: 0, 50%: 1.216, 75%: 0	Same as in Humans

Main Results and Figures

• Analysis_Human.csv, Analysis_Cats_and_Dogs.csv, Analysis_Pigs.csv, and Analysis_Water_Buffalo_and_Cattle.csv contain the diagnostic sensitivity, specificity and prevalence data for each parasite generated by the JAGS model. They will have not been uploaded to this repository, however, they can be reproduced locally by running LatentClassAnalysisFTAF.R as they are large .csv files. In place are the 95% Bayesian credible intervals seen in Quantile_Data_All_Hosts.csv.
• Prevalence_Real_Sim_All_Hosts.svg is the plot of the simulated prevalence of each parasite compared with the prevalence recorded in the real-world data, with error bars highlighting the 95% confidence interval. Each host is arranged in a two-by-two grid with all diagnostics plotted together.
• Sensitivity_Prevalence_All_Hosts.svg is the plot of the simulated sensitivites and prevalences of the JAGS model. Each host is arranged in a two-by-two grid, with all diagnostics plotted together. This has not been uploaded to this repository since it is a large file (~650 MB) however, it can be reproduced locally by running LatentClassAnalysisFTAF.R.
• Sensitivity_Prevalence_All_Hosts_Simplified.svg is the plot of the simulated sensitivites and prevalences of the JAGS model. Unlike the plots in the file Sensitivity_Prevalence_All_Hosts.svg, the scatterpoints are replaced by error bars corresponding to the 95% confidence interval of the data, and the mean of the sensitivity and the mean of the prevalence is plotted point.
• p1_p0_All_Hosts_Simplified.svg is the the Reciever Operating Characteristic (ROC) plot of the diagnostic sensitivity and 1 minus the diagnostic specificity. Like Sensitivity_Prevalence_All_Hosts_Simplified, it plots each host for all parasites and diagnostics in a two-by-two grid. The mean of the diagnostic sensitivity and 1 minus the diagnostic specificity are plotted, with error bars corresponding to the 95% confidence interval for the data.

Note: When running the code generating the main figures Prevalence_Real_Sim_All_Hosts.svg, Sensitivity_Prevalence_All_Hosts.svg, Sensitivity_Prevalence_All_Hosts_Simplified.svg, and p1_p0_All_Hosts_Simplified.svg, please ensure the RStudio plotting window is very large and wide, especially on small screens such as a laptop.

Supplementary Figures

These figures have not been included in the main publication. For each host and parasite, <Parasite>_<Host>_Supplementary.svg in /Supplementary_Figures/ contains a histogram of the estimated probabilities of infection, a histogram of the estimated prevalence of the model, and frequency curves of the sensitivity and specificity of the diagnostics of the model each the parasite.

Session Info

The R version and used packages are displayed below:

R version 4.2.2 (2022-10-31)
Platform: aarch64-apple-darwin20 (64-bit)
Running under: macOS Ventura 13.4

Matrix products: default
LAPACK: /Library/Frameworks/R.framework/Versions/4.2-arm64/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] compiler  stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] rstudioapi_0.14   bayestestR_0.13.0 MCMCpack_1.6-3    MASS_7.3-58.1     ggplot2_3.4.0     cowplot_1.1.1    
[7] runjags_2.2.1-7   rjags_4-13        coda_0.19-4      

loaded via a namespace (and not attached):
 [1] cellranger_1.1.0   pillar_1.8.1       iterators_1.0.14   tools_4.2.2        viridisLite_0.4.1 
 [6] lifecycle_1.0.3    tibble_3.1.8       gtable_0.3.1       lattice_0.20-45    pkgconfig_2.0.3   
[11] rlang_1.0.6        foreach_1.5.2      igraph_1.3.5       Matrix_1.5-1       cli_3.6.0         
[16] parallel_4.2.2     SparseM_1.81       withr_2.5.0        dplyr_1.1.0        MatrixModels_0.5-1
[21] generics_0.1.3     vctrs_0.5.2        grid_4.2.2         tidyselect_1.2.0   glue_1.6.2        
[26] R6_2.5.1           fansi_1.0.4        survival_3.4-0     readxl_1.4.1       farver_2.1.1      
[31] magrittr_2.0.3     codetools_0.2-18   splines_4.2.2      scales_1.2.1       mcmc_0.9-7        
[36] datawizard_0.6.5   insight_0.19.0     colorspace_2.1-0   quantreg_5.94      utf8_1.2.2        
[41] munsell_0.5.0