This directory contains the code to produce the results for Leavitt et al. 2020, published in IJE (https://doi.org/10.1093/ije/dyaa031). It contains code to run the simulations assessing the performance of the naive Bayes transmission method including estimating the reproductive number and the application to a TB outbreak in Hamburg, Germany. These programs use the nbTransmission R package available on CRAN or GitHub (https://github.com/sarahleavitt/nbTransmission) as well as the simulation programs in the nbSimulation directory (https://github.com/sarahleavitt/nbSimulation).
The following programs were used to run a main simulation to test the performance of the naive Bayes transmission method and then perform various sensitivity analyses.
This program contains a function "simRun" which performs one iteration of a simulation which simulates an outbreak with pathogen genetics and covariates, estimates the relative transmission probability using naive Bayes, assesses the performance, and estimates the reproductive number.
This program is a wrapper for "simRun" which runs the simulation coded in that function nSim times. It saves the raw pair-level results for one iteration and the effect estimates, reproductive number, and performance for all iterations. This is run by "SimQsub.qsub" with sample size and observation date as inputs.
This program has a function called "sampleSizeRun" which is analogous to "simRun" but is specifically designed to assess the affect of the sample size and the training dataset proportion on performance and so it varies the training proportion from 0.1 to 1.
This program is analogous to PerformSimulation.R but runs multiple iterations of "sampleSizeRun". This is run by "SampleSizeQsub.qsub" with sample size as an input.
This program is used to run a sensitivity analysis of the pathogen genome length. It simulates one outbreak and then simulates pathogen genomes of various lengths nSim times. Then it calculates various summaries of the SNP distances across the outbreaks. This is run by "GenomeQsub.qsub" with no inputs.
This a qsub file to run PerformSimulation.R as a batch job. It runs t parallel jobs and therefore the total number of simulations is t*nSim times. It needs sample size and observation date (1 for infection date, 2 for sampling date) as inputs and gets them from the shell script "Simulation.sh".
This a qsub file to run SampleSizePerform.R as a batch job. It runs t parallel jobs and therefore the total number of simulations is t*nSim times. It needs sample size as an input and gets it from the shell script "Simulation.sh".
This a qsub file to run GenomeTesting.R as a batch job, it does not have any inputs. It runs t parallel jobs and therefore the total number of simulations is t*nSim times.
This is a shell script that runs SimQsub.qsub with infection date for the main simulation, SimQsub.qsub with sampling date for the date sensitivity analysis, SampleSizeQsub.qsub with different sample sizes for the training proportion analysis, and GenomeQsub.qsub for the pathogen genome length sensitivity analysis.
This program reads in the results from all of the simulation programs and then calculates all results used in Leavitt et al. 2020 including values in the text, tables and figures.
The following programs were used to prepare, analyze, and evaluate applying the naive Bayes transmission method to a TB outbreak in Hamburg Germany anaylzed by Roetzer et al. 2013 (https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1001387) The full dataset used is not publically available.
This program reads in the demographic, contact, and pathogen genetic data, creates cleaned individual-level and pair-level datasets to be used for analysis.
This program calculates relative transmission probabilities using naive Bayes, training with contact investigation, training with SNP distance, using various serial intervals, and randomly. It then estimates the monthly and over all average reproductive numbers with 95% confidence intervals. It also runs a sensitivity analysis for the small sample size correction.
This program takes the results saved in HamburgAnalysis.R and creates all results used in Leavitt et al. 2020 including values in the text, tables and figures.