Raw data and MCMC output underlying plots and results in the McQueen paper

The large tarballs are now also available on Dropbox here.

The file structure is identical to the github repository tarballs folder.

Reference datasets D_B

The combination of sequence data from the Los Alamos HIV sequence database, the Stanford drug resistance database and the HIV positive selection database, run through our quality control pipeline and used for our reference dataset at the time of performing inference is here:

• reference_datasets/HLA_inference/combined_{protease,reverse_transcriptase}_references.fasta.zip

The viral sequence data taken from the Stanford drug resistance database, run through our quality control pipeline and used for drug associated selection inference in the paper is here:

• reference_datasets/Drug_inference/{protease,reverse_transcriptase}_references.fasta.zip

Query datasets D

The viral sequence data taken from the Stanford drug resistance database, run through our quality control pipeline and used for drug associated selection inference in the paper is here:

• query_datasets/Drug_inference/{protease,reverse_transcriptase}_queries.fasta.zip

Associated host drug regime information for these patients is here:

• csv_files/Drug_regime_data/{protease,reverse_transcriptase}_drugs.csv

Main Text

Figures

• Figure 1: The underlying process and our approximation.
  • N/A
• Figure 2: Simulation results summary; Inference results for Simulation studies 1 and 2.
  • source_data_files/Figure_2.Rdata is an Rdata file containing: median, mean, and quantiles displayed in Figure 2a, b and c, as well as the 100 ROC curves for each independent run using the methods detailed in the paper and the average ROC curve for each method.
  • Raw output from MCMC runs: tarballs/simulation_study_1/dsim_closest_100_rec_{0,0.01}.tar.gz, tarballs/simulations_study_2/simulation_study_2.gz.
• Figure 3: Drug associated selection analysis.
  • source_data_files/Figure_3.Rdata contains the 2.5%, 10% and 50% quantiles and median omega for protease and reverse transriptase. These are used to define top-tier and second tier candidate sites, permutation of labels leads to the p-values and odds ratios observed in Figure 3b, and create the plot shown in Figure 3a for protease.
  • Raw output from MCMC runs: tarballs/Drug_inference/{protease,reverse_transcriptase}_drug_inference.tar.gz.
• Figure 4: Selected results summaries.
  • source_data_files/Figure_4.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in reverse transcriptase. We also include the number of individuals harbouring each of the HLA types.
  • Raw output from MCMC run: tarballs/HLA_inference/reverse_transcriptase_HLA_inference.tar.gz.
• Figure 5: HLA-B associated selection in reverse transcriptase.
  • source_data_files/Figure_5.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in reverse transcriptase. We also include the number of individuals harbouring each of the HLA types.
  • Raw output from MCMC run: tarballs/HLA_inference/reverse_transcriptase_HLA_inference.tar.gz.

Tables

• Table 1: Data used in this study: public databases.
  • N/A
• Table 2: Data used in this study: viral sequence data with associated host HLA information.
  • N/A
• Table 3: Overlap between sites under selection and known HLA epitopes.
  • source_data_files/Table_3.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA for protease and reverse transcriptase. These are used to define top-tier and second tier candidate sites, permutation of labels leads to the p-values and odds ratios observed in Table 3.
  • Raw output from MCMC run: tarballs/HLA_inference/{protease,reverse_transcriptase}_HLA_inference.tar.gz.
• Table 4: Codons in reverse transcriptase showing evidence for HLA associated selection.
  • source_data_files/Table_4.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA. These are used to define top-tier and second tier candidate sites in reverse transcriptase.
  • Raw output from MCMC run: tarballs/HLA_inference/reverse_transcriptase_HLA_inference.tar.gz.
• Table 5: Codons in protease showing evidence for HLA associated selection.
  • Raw output from MCMC run: tarballs/HLA_inference/protease_HLA_inference.tar.gz.
  • source_data_files/Table_5.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA. These are used to define top-tier and second tier candidate sites in protease.

Supplementary Materials

Figures

• Figure 1: Simulation study 1: summary of the selection profiles used.
  • N/A
• Figure 2: Simulation study 1: parameter estimates, no recombination (r= 0)
  • Rdata file containing median, mean, and quantiles displayed: source_data_files/Supplementary_Figure_2.Rdata.
  • Raw output from MCMC runs: tarballs/simulation_study_1/dsim_closest_100_rec_0.tar.gz.
• Figure 3: Simulation Study 1: comparison of estimates obtained using the closest 10 sequences by Hamming distance to the closest 100 sequences for r= 0 and r= 0.01.
  • Rdata file containing median, mean, and quantiles displayed: source_data_files/Supplementary_Figure_3.Rdata.
  • Raw output from MCMC runs tarballs/simulation_study_1/dsim_closest_{10,100}_rec_{0,0_01}.tar.gz
• Figure 4: Simulation Study 1: parameter estimates, r= 0.01.
  • Rdata file containing median, mean, and quantiles displayed: source_data_files/Supplementary_Figure_4.Rdata.
  • Raw output from MCMC runs tarballs/simulation_study_1/dsim_closest_100_rec_0_01.tar.gz
• Figure 5: Simulation Study 1: comparison of estimates obtained using the closest 100 sequences by Hamming distance to the sequences actually copied from in the simulation, r= 0.05.
  • Rdata file containing median, mean, and quantiles displayed: source_data_files/Supplementary_Figure_5.Rdata.
  • Raw output from MCMC runs tarballs/simulation_study_1/dsim_closest_100_rec_0_05.tar.gz, tarballs/simulation_study_1/dsim_closest_100_rec_0_05_mosaic.tar.gz.
• Figure 6: Simulation Study 2: parameter estimates.
  • Rdata file containing median, mean, and quantiles displayed: source_data_files/Supplementary_Figure_6.Rdata.
• Figure 7: Simulation Study 2: sample size of 3000, results summary.
  • Rdata files containing ROC curve inforamation for each method: source_data_files/Supplementary_Figure_7.Rdata.
  • Raw output from MCMC runs: tarballs/simulation_study_2/dgen_sim_HLA_coeff_1460_3000_comparison/3000_leaves/*.tar.gz.
• Figure 8: Simulation Study 2: impact of reduced reference sequence set, results summary.
  • Rdata files containing ROC curve inforamation for each method: source_data_files/Supplementary_Figure_8.Rdata.
  • Raw output from MCMC runs: tarballs/simulation_study_2/dgen_sim_HLA_coeff_1460_3000_comparison/1460_leaves/*.tar.gz.
• Figure 9: Simulation Study 3: example results for HLA-A alleles.
  • Rdata file containing median information, coverage and RMSE for each of the different reference datasets used: source_data_files/Supplementary_Figure_9.Rdata.
  • Raw output from MCMC runs: tarballs/simulation_study_3/*.tar.gz.
• Figure 10: Simulation Study 3: example results for HLA-B alleles.
  • Rdata file containing median information, coverage and RMSE for each of the different reference datasets used: source_data_files/Supplementary_Figure_10.Rdata.
  • Raw output from MCMC runs: tarballs/simulation_study_3/*.tar.gz.
• Figure 11: Simulation Study 3: example results for HLA-C alleles.
  • Rdata file containing median information, coverage and RMSE for each of the different reference datasets used: source_data_files/Supplementary_Figure_11.Rdata.
  • Raw output from MCMC runs: tarballs/simulation_study_3/*.tar.gz.
• Figure 12: Drug associated selection analysis: reverse transcriptase results summary.
  • source_data_files/Supplementary_Figure_12.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by drug.
  • Raw output from MCMC runs: tarballs/Drug_inference/reverse_transcriptase_drug_inference.tar.gz.
• Figure 13: Hierarchical clustering of drug selection profiles.
  • source_data_files/Supplementary_Figure_13.Rdata contains drug_selection_median; the median drug associated selection coefficients, and trees; native ape format (requires the R library ape) for the trees determined using hclust.
• Figure 14: Summary of HLA-A associated selection in protease.
  • source_data_files/Supplementary_Figure_14.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in protease. We also include the number of individuals harbouring each of the HLA types.
  • Raw output from MCMC run: tarballs/HLA_inference/protease_HLA_inference.tar.gz.
• Figure 15: Summary of HLA-B associated selection in protease.
  • source_data_files/Supplementary_Figure_15.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in protease. We also include the number of individuals harbouring each of the HLA types.
  • Raw output from MCMC run: tarballs/HLA_inference/protease_HLA_inference.tar.gz.
• Figure 16: Summary of HLA-C associated selection in protease.
  • source_data_files/Supplementary_Figure_16.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in protease. We also include the number of individuals harbouring each of the HLA types.
  • Raw output from MCMC run: tarballs/HLA_inference/protease_HLA_inference.tar.gz.
• Figure 17: Summary of HLA-A associated selection in reverse transcriptase.
  • source_data_files/Supplementary_Figure_17.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in reverse transcriptase. We also include the number of individuals harbouring each of the HLA types.
  • Raw output from MCMC run: tarballs/HLA_inference/reverse_transcriptase_HLA_inference.tar.gz.
• Figure 18: Summary of HLA-C associated selection in reverse transcriptase.
  • source_data_files/Supplementary_Figure_18.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in reverse transcriptase. We also include the number of individuals harbouring each of the HLA types.
  • Raw output from MCMC run: tarballs/HLA_inference/reverse_transcriptase_HLA_inference.tar.gz.
• Figure 19: Estimated recombination probabilities between sites across protease and reverse transcriptase.
  • source_data_files/Supplementary_Figure_19.Rdata contains the mean, median, 2.5%, 25%, 75%, and 97.5% quantiles for probability of recombination between adjacent sites in protease and reverse transcriptase, away from subtype B and subtype C consensus viral sequence.
  • Raw output from MCMC run: tarballs/HLA_inference/{protease,reverse_transcriptase}_HLA_inference.tar.gz.
• Figure 20: The impact of HLA alleles on differentiation between HIV-1 subtypes.
  • source_data_files/Supplementary_Figure_20.Rdata contains estimates of HLA frequency weighted selection away from subtype B and subtype C viral consensus.

Tables

• Table 1: Priors on parameters of the model.
  • N/A
• Table 2: The collection of ‘false positive’ associations are interrogated.
  • N/A
• Table 3: The collection of apparent false negative associations are interrogated.
  • N/A
• Table 4: The collection of top-tier and second-tier sites in protease.
  • source_data_files/Supplementary_Figure_14.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA. These are used to define top-tier and second tier candidate sites in protease.
  • Raw output from MCMC run: tarballs/HLA_inference/protease_HLA_inference.tar.gz.
• Table 5: The collection of top-tier and second-tier sites in reverse transriptase.
  • source_data_files/Figure_4.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA. These are used to define top-tier and second tier candidate sites in reverse transcriptase.
  • Raw output from MCMC run: tarballs/HLA_inference/reverse_transcriptase_HLA_inference.tar.gz.
• Table 6: Overlap between studies, protease.
  • source_data_files/Supplementary_Figure_14.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA. These are used to define top-tier and second tier candidate sites in protease.
  • Raw output from MCMC run: tarballs/HLA_inference/protease_HLA_inference.tar.gz.
• Table 7: Overlap between studies, reverse transcriptase.
  • source_data_files/Figure_4.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA. These are used to define top-tier and second tier candidate sites in reverse transcriptase.
  • Raw output from MCMC run: tarballs/HLA_inference/reverse_transcriptase_HLA_inference.tar.gz.
• Table 8: Table of p-values for differences between H_B and H_C) associated selection at sites that distinguish subtype B and subtype C in protease and reverse transcriptase.
  • source_data_files/Supplementary_Figure_20.Rdata contains estimates of HLA frequency weighted selection away from subtype B and subtype C viral consensus. Labels are randomly permuted to generate the p-values ins Supplementary Table 8.
• Table 9: Summary of reference datasets for simulation study 3.
  • Rdata file containing median information, coverage and RMSE for each of the different reference datasets used: source_data_files/Supplementary_Figure_9.Rdata.
  • Raw output from MCMC runs: tarballs/simulation_study_3/*.tar.gz.

Source datafile - to be available for download with manuscript.

These will be small in filesize so that they may be downloaded from the journal website and provide sufficient data to recreate the various plots.

Main text

• Figure 2 source_data_files/Figure_2.Rdata is a list containing all the information to create Figure 2. The list has a collection of sublists:

  • simulation_study_1_closest_100_R_0: information for simulation study 1 with the recombination rate between neighbouring sites set to 0 and using the closest 100 sequences by Hamming distance as references.
  • simulation_study_1_closest_100_R_0_01: information for simulation study 1 with the recombination rate between neighbouring sites set to 0.01 and using the closest 100 sequences by Hamming distance as references.
  • simulation_study_2: information for simulation study 2 estimations after simulating under a birth-death process.
  • simulation_study_2_ROC: ROC curve information.

  Each of the first three lists consists of the following entries:

  • truth: a list consisting of the true underlying host associated selection parameters (HLA_esc_truth) and reversion parameters rev_truth.
  • omega_truth: the true underlying dN/dS ratio for the codon model, for one of the 100 independent runs.
  • omega_mean: the mean dN/dS estimates for one of the 100 independent runs.
  • omega_median: the median dN/dS estimates for one of the 100 independent runs.
  • omega_credible_95: the 95% credible interval for dN/dS estimates for one of the 100 independent runs.
  • omega_credible_50: the 50% credible interval for dN/dS estimates for one of the 100 independent runs.
  • R_truth: the true underlying recombination probability between adjacent sites for the simulation study.
  • mean_R: the mean estimate of recombination between adjacent sites.
  • median_R: the median estimate of recombination between adjacent sites.
  • credible_95_R: the average 95% credible interval across the 100 independent runs.
  • credible_50_R: the average 50% credible interval across the 100 independent runs.
  • mean_escape: the mean host associated selection estimates.
  • median_escape: the median host associated selection estimates.
  • coverage_95_escape: the coverage of the 95% credible interval across the 100 independent runs.
  • coverage_50_escape: the coverage of the 50% credible interval across the 100 independent runs.
  • credible_95_escape: the average 95% credible interval across the 100 independent runs.
  • credible_50_escape: the average 50% credible interval across the 100 independent runs.
  • mean_reversion: the mean reversion estimates at each site.
  • median_reversion: the median reversion estimates at each site.
  • coverage_95_reversion: the coverage of the 95% credible interval across the 100 independent runs.
  • coverage_50_reversion: the coverage of the 50% credible interval across the 100 independent runs.
  • credible_95_reversion: the average 95% credible interval across the 100 independent runs.
  • credible_50_reversion: the average 50% credible interval across the 100 independent runs.

  The final list simulation_study_2_ROC consists of the following entries:

  • bhattacharya_average: average ROC curve using the methods of Bhattacharya et_al..
  • phyloD_average: average ROC curve using the methods of Carlson et al..
  • phyloD_false_positive: rolling false positive rate as threshold is decreased.
  • phyloD_true_positive: rolling true positive rate as threshold is decreased.
  • phyloD_OR_average: average ROC curve using the methods of Carlson et al..
  • phyloD_OR_false_positive: rolling false positive rate as threshold is decreased.
  • phyloD_OR_true_positive: rolling true positive rate as threshold is decreased.
  • fishers_exact_average: average ROC curve using the methods of Moore et al..
  • fryer_average: average ROC curve using the methods of Fryer et al..
  • fryer_false_positive: rolling false positive rate as threshold is decreased.
  • fryer_true_positive: rolling true positive rate as threshold is decreased.
  • our_method_average: average ROC curve using our method, McQueen.
  • our_false_positive: rolling false positive rate as threshold is decreased.
  • our_true_positive: rolling true positive rate as threshold is decreased.

• Figure 3 source_data_files/Figure_3.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by drug, and the median omega.

• Figure 4 source_data_files/Figure_4.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in reverse transcriptase. Information is contained in named lists with 2 elements, one from subtype B consensus and one from subtype C consensus. We also include the number of individuals harbouring each of the HLA types.

• Figure 5 source_data_files/Figure_5.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in reverse transcriptase. Information is contained in named lists with 2 elements, one from subtype B consensus and one from subtype C consensus. We also include the number of individuals harbouring each of the HLA types.

Supplementary Materials

• Figure 2 source_data_files/Supplementary_Figure_2.Rdata contains the following information for simulation study 1 with recombination probability set to 0 between neighbouring sites and using the closest 100 reference samples by Hamming distance:

  • truth: a list consisting of the true underlying host associated selection parameters (HLA_esc_truth) and reversion parameters rev_truth.
  • omega_truth: the true underlying dN/dS ratio for the codon model, for one of the 100 independent runs.
  • omega_mean: the mean dN/dS estimates for one of the 100 independent runs.
  • omega_median: the median dN/dS estimates for one of the 100 independent runs.
  • omega_credible_95: the 95% credible interval for dN/dS estimates for one of the 100 independent runs.
  • omega_credible_50: the 50% credible interval for dN/dS estimates for one of the 100 independent runs.
  • R_truth: the true underlying recombination probability between adjacent sites for the simulation study.
  • mean_R: the mean estimate of recombination between adjacent sites.
  • median_R: the median estimate of recombination between adjacent sites.
  • credible_95_R: the average 95% credible interval across the 100 independent runs.
  • credible_50_R: the average 50% credible interval across the 100 independent runs.
  • mean_escape: the mean host associated selection estimates.
  • median_escape: the median host associated selection estimates.
  • coverage_95_escape: the coverage of the 95% credible interval across the 100 independent runs.
  • coverage_50_escape: the coverage of the 50% credible interval across the 100 independent runs.
  • credible_95_escape: the average 95% credible interval across the 100 independent runs.
  • credible_50_escape: the average 50% credible interval across the 100 independent runs.
  • mean_reversion: the mean reversion estimates at each site.
  • median_reversion: the median reversion estimates at each site.
  • coverage_95_reversion: the coverage of the 95% credible interval across the 100 independent runs.
  • coverage_50_reversion: the coverage of the 50% credible interval across the 100 independent runs.
  • credible_95_reversion: the average 95% credible interval across the 100 independent runs.
  • credible_50_reversion: the average 50% credible interval across the 100 independent runs.

• Figure 3 source_data_files/Supplementary_Figure_3.Rdata contains lists of information for output from simulation study 1:

  • closest_10_R_0: simulation study 1 with recombination probability set to 0 between neighbouring sites and using the closest 10 reference samples by Hamming distance.
  • closest_100_R_0: simulation study 1 with recombination probability set to 0 between neighbouring sites and using the closest 100 reference samples by Hamming distance
  • closest_10_R_0_01: simulation study 1 with recombination probability set to 0.01 between neighbouring sites and using the closest 100 reference samples by Hamming distance
  • closest_100_R_0_01: simulation study 1 with recombination probability set to 0.01 between neighbouring sites and using the closest 100 reference samples by Hamming distance

  Each contains the information as in source_data_files/Supplementary_Figure_2.Rdata.

• Figure 4 source_data_files/Supplementary_Figure_4.Rdata contains the information as in source_data_files/Supplementary_Figure_2.Rdata for simulation study 1 with recombination probability set to 0.01 between neighbouring sites and using the closest 100 reference samples by Hamming distance.

• Figure 5 source_data_files/Supplementary_Figure_5.Rdata contains two lists of information for output from simulation study 1 with recombination probability set to 0.05 between neighbouring sites:

  • closest_100_R_0_05: simulation study 1 with recombination probability set to 0.05 between neighbouring sites and using the closest 100 reference samples by Hamming distance
  • closest_100_R_0_05_actually_copied_from: simulation study 1 with recombination probability set to 0 between neighbouring sites and using the reference sequences actually copied from to generate the query sequences.

  Each contains the information as in source_data_files/Supplementary_Figure_2.Rdata.

• Figure 6 source_data_files/Supplementary_Figure_6.Rdata contains the information as in source_data_files/Supplementary_Figure_2.Rdata for simulation study 2 (simulating data under a birth-death process) and using the closest 100 reference samples by Hamming distance.

• Figure 7 source_data_files/Supplementary_Figure_7.Rdata contains the collection of ROC curves for the compared methods when simulating trees with 3000 leaves.

• Figure 8 source_data_files/Supplementary_Figure_8.Rdata contains the collection of ROC curves for the compared methods when simulating trees with 1460 leaves.

• Figure 9 source_data_files/Supplementary_Figure_9.Rdata contains a collection of lists:

  • our_reference_set: All available sequence data from the Los Alamos HIV sequence database, the Stanford drug resistance database and the HIV positive selection database.
  • our_reference_set_LOO: All available sequence data from the Los Alamos HIV sequence database, the Stanford drug resistance database and the HIV positive selection database, plus all query sequences using a leave one out approach.
  • los_alamos_reference_set: Restrict to only Los Alamos sequence data in the reference dataset.
  • los_alamos_reference_set_LOO: Restrict to only Los Alamos sequence data in the reference dataset, plus all query sequences using a leave one out approach.
  • los_alamos_reference_set_remove_botswana: Restrict to only Los Alamos sequence data in the reference dataset and remove all sequences sampled from Botswana.
  • los_alamos_reference_set_remove_botswana_LOO: Restrict to only Los Alamos sequence data in the reference dataset and remove all sequences sampled from Botswana and include all query sequences using a leave one out approach.
  • only_queries: Use query sequences as the reference dataset using a leave one out approach.
  • gold_standard_reference_set: Use the sequences actually copied from as the reference dataset. Each list contains the following data:
  • median_escape: The median escape estimates across the runs site by site for each HLA.
  • coverage_50_escape: The coverage of the 50% credible invervals site by site for each HLA.
  • coverage_95_escape: The coverage of the 95% credible intervals site by site for each HLA.
  • RMSE: The root mean squared error site by site for each HLA. In addition, there is also HLA_truth which contains the true underlying selection coefficients for the parametric bootstrap.

• Figure 10 source_data_files/Supplementary_Figure_10.Rdata contains the same information as in source_data_files/Supplementary_Figure_9.Rdata.

• Figure 11 source_data_files/Supplementary_Figure_11.Rdata contains the same information as in source_data_files/Supplementary_Figure_9.Rdata.

• Figure 12 source_data_files/Supplementary_Figure_12.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by drug.

• Figure 13 source_data_files/Supplementary_Figure_13.Rdata contains drug_selection_median; the median drug associated selection coefficients, and trees; native ape format (requires the R library ape) for the trees determined using hclust.

• Figure 14 source_data_files/Supplementary_Figure_14.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in protease. Information is contained in named lists with 2 elements, one from subtype B consensus and one from subtype C consensus. We also include the number of individuals harbouring each of the HLA types.

• Figure 15 source_data_files/Supplementary_Figure_15.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in protease. Information is contained in named lists with 2 elements, one from subtype B consensus and one from subtype C consensus. We also include the number of individuals harbouring each of the HLA types.

• Figure 16 source_data_files/Supplementary_Figure_16.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in protease. Information is contained in named lists with 2 elements, one from subtype B consensus and one from subtype C consensus. We also include the number of individuals harbouring each of the HLA types.

• Figure 17 source_data_files/Supplementary_Figure_17.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in reverse transcriptase. Information is contained in named lists with 2 elements, one from subtype B consensus and one from subtype C consensus. We also include the number of individuals harbouring each of the HLA types.

• Figure 18 source_data_files/Supplementary_Figure_18.Rdata contains the 2.5%, 10% and 50% quantiles for HLA associated selection, and median omega for inference in reverse transcriptase. Information is contained in named lists with 2 elements, one from subtype B consensus and one from subtype C consensus. We also include the number of individuals harbouring each of the HLA types.

• Figure 19 source_data_files/Supplementary_Figure_19.Rdata contains the mean, median, 2.5%, 25%, 75%, and 97.5% quantiles for probability of recombination between adjacent sites in protease and reverse transcriptase, away from subtype B and subtype C consensus viral sequence.

• Figure 20 source_data_files/Supplementary_Figure_20.Rdata contains estimates of HLA frequency weighted selection away from subtype B and subtype C viral consensus:

  • log_protease_from_B_by_subtype_B: log base 10 of selection away from subtype B protease consensus by HLA frequency distribution of hosts harbouring subtype B viruses.
  • log_protease_from_B_by_subtype_C: log base 10 of selection away from subtype B protease consensus by HLA frequency distribution of hosts harbouring subtype C viruses.
  • log_protease_from_C_by_subtype_B: log base 10 of selection away from subtype C protease consensus by HLA frequency distribution of hosts harbouring subtype B viruses.
  • log_protease_from_C_by_subtype_C: log base 10 of selection away from subtype C protease consensus by HLA frequency distribution of hosts harbouring subtype C viruses.
  • log_RT_from_B_by_subtype_B: log base 10 of selection away from subtype B reverse transcriptase consensus by HLA frequency distribution of hosts harbouring subtype B viruses.
  • log_RT_from_B_by_subtype_C: log base 10 of selection away from subtype B reverse transcriptase consensus by HLA frequency distribution of hosts harbouring subtype C viruses.
  • log_RT_from_C_by_subtype_B: log base 10 of selection away from subtype C reverse transcriptase consensus by HLA frequency distribution of hosts harbouring subtype B viruses.
  • log_RT_from_C_by_subtype_C: log base 10 of selection away from subtype C reverse transcriptase consensus by HLA frequency distribution of hosts harbouring subtype C viruses. We also included the collection of sites that differ between subtype B and subtype C in reverse transcriptase: B_C_diffs_protease, B_C_diffs_RT.

• Table 4: The collection of top-tier and second-tier sites in protease.

  • source_data_files/Supplementary_Figure_14.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA. These are used to define top-tier and second tier candidate sites in protease.

• Table 5: The collection of top-tier and second-tier sites in reverse transriptase.

  • source_data_files/Figure_4.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA. These are used to define top-tier and second tier candidate sites in reverse transcriptase.

• Table 6: Overlap between studies, protease.

  • source_data_files/Supplementary_Figure_14.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA. These are used to define top-tier and second tier candidate sites in protease.

• Table 7: Overlap between studies, reverse transcriptase.

  • source_data_files/Figure_4.Rdata contains the 2.5%, 10% and 50% quantiles with rows labelled by HLA. These are used to define top-tier and second tier candidate sites in reverse transcriptase.

• Table 8: Table of p-values for differences between H_B and H_C) associated selection at sites that distinguish subtype B and subtype C in protease and reverse transcriptase.

  • source_data_files/Supplementary_Figure_20.Rdata contains estimates of HLA frequency weighted selection away from subtype B and subtype C viral consensus. Labels are randomly permuted to generate the p-values ins Supplementary Table 8.

• Table 9: Summary of reference datasets for simulation study 3.

  • Rdata file containing median information, coverage and RMSE for each of the different reference datasets used: source_data_files/Supplementary_Figure_9.Rdata.