Single-Omic Models After Integration

Author

Philipp Sven Lars Schäfer

Published

November 28, 2024

1 Packages

Code

suppressPackageStartupMessages({
  library(tidyverse)
  library(flextable)
  library(ggdark)
  library(magick)
  source(file.path("..", "src", "read_data.R"))
  source(file.path("..", "src", "colors.R"))
  source(file.path("..", "src", "generate_targets.R"))
  source(file.path("..", "src", "model.R"))
})

2 Data

Code

input_dir = file.path("..", "data")

Code

celltype_meta <- read_celltype_meta(input_dir)
gene_meta <- read_gene_meta_plus(input_dir)
protein_meta <- read_protein_meta(input_dir)

meta_data <- read_harmonized_meta_data(input_dir)
specimen_per_day <- get_specimen_per_day(meta_data=meta_data)

RECOMPUTE <- TRUE
if (RECOMPUTE) {
  source(file.path("..", "src", "normalize_integrate.R"))
  
  raw_experimental_data <- read_raw_experimental_data(input_dir)
  
  filtered_experimental_data <- filter_experimental_data(
    meta_data=meta_data, 
    experimental_data=raw_experimental_data,
    gene_meta=gene_meta)
  
  write_rds(filtered_experimental_data, 
            file = file.path(input_dir, "prc_datasets", 
                             "filtered_experimental_data.RDS"))
  
  normalized_experimental_data <- normalize_experimental_data(
    meta_data=meta_data, 
    raw_experimental_data=filtered_experimental_data,
    gene_meta=gene_meta)
  
  write_rds(normalized_experimental_data, 
            file = file.path(input_dir, "prc_datasets", 
                             "normalized_experimental_data.RDS"))
  
  integrated_experimental_data <- integrate_experimental_data(
    meta_data=meta_data, 
    normalized_experimental_data=normalized_experimental_data)
  
  write_rds(integrated_experimental_data, 
            file = file.path(input_dir, "prc_datasets", 
                             "integrated_experimental_data.RDS"))

  # use raw/filtered experimental data for computation of targets
  target_list <- generate_all_targets(
    meta_data=meta_data, 
    experimental_data=filtered_experimental_data, 
    experimental_data_settings=experimental_data_settings, 
    gene_meta=gene_meta,
    protein_meta=protein_meta
    )
  
  write_rds(target_list, 
          file = file.path(input_dir, "prc_datasets", 
                           "target_list.RDS"))
  
  rm(raw_experimental_data, filtered_experimental_data, normalized_experimental_data)
  experimental_data <- integrated_experimental_data
} else {
  experimental_data <- read_rds(file = file.path(input_dir, "prc_datasets", 
                                                            "integrated_experimental_data.RDS"))
  target_list <- read_rds(file = file.path(input_dir, "prc_datasets", 
                                           "target_list.RDS"))
}

Loading required package: S4Vectors

Loading required package: stats4

Loading required package: BiocGenerics


Attaching package: 'BiocGenerics'

The following object is masked from 'package:flextable':

    width

The following objects are masked from 'package:lubridate':

    intersect, setdiff, union

The following objects are masked from 'package:dplyr':

    combine, intersect, setdiff, union

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, aperm, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which.max, which.min


Attaching package: 'S4Vectors'

The following objects are masked from 'package:lubridate':

    second, second<-

The following objects are masked from 'package:dplyr':

    first, rename

The following object is masked from 'package:tidyr':

    expand

The following objects are masked from 'package:base':

    I, expand.grid, unname

Loading required package: IRanges


Attaching package: 'IRanges'

The following object is masked from 'package:lubridate':

    %within%

The following objects are masked from 'package:dplyr':

    collapse, desc, slice

The following object is masked from 'package:purrr':

    reduce

Loading required package: GenomicRanges

Loading required package: GenomeInfoDb

Loading required package: SummarizedExperiment

Loading required package: MatrixGenerics

Loading required package: matrixStats


Attaching package: 'matrixStats'

The following object is masked from 'package:dplyr':

    count


Attaching package: 'MatrixGenerics'

The following objects are masked from 'package:matrixStats':

    colAlls, colAnyNAs, colAnys, colAvgsPerRowSet, colCollapse,
    colCounts, colCummaxs, colCummins, colCumprods, colCumsums,
    colDiffs, colIQRDiffs, colIQRs, colLogSumExps, colMadDiffs,
    colMads, colMaxs, colMeans2, colMedians, colMins, colOrderStats,
    colProds, colQuantiles, colRanges, colRanks, colSdDiffs, colSds,
    colSums2, colTabulates, colVarDiffs, colVars, colWeightedMads,
    colWeightedMeans, colWeightedMedians, colWeightedSds,
    colWeightedVars, rowAlls, rowAnyNAs, rowAnys, rowAvgsPerColSet,
    rowCollapse, rowCounts, rowCummaxs, rowCummins, rowCumprods,
    rowCumsums, rowDiffs, rowIQRDiffs, rowIQRs, rowLogSumExps,
    rowMadDiffs, rowMads, rowMaxs, rowMeans2, rowMedians, rowMins,
    rowOrderStats, rowProds, rowQuantiles, rowRanges, rowRanks,
    rowSdDiffs, rowSds, rowSums2, rowTabulates, rowVarDiffs, rowVars,
    rowWeightedMads, rowWeightedMeans, rowWeightedMedians,
    rowWeightedSds, rowWeightedVars

Loading required package: Biobase

Welcome to Bioconductor

    Vignettes contain introductory material; view with
    'browseVignettes()'. To cite Bioconductor, see
    'citation("Biobase")', and for packages 'citation("pkgname")'.


Attaching package: 'Biobase'

The following object is masked from 'package:MatrixGenerics':

    rowMedians

The following objects are masked from 'package:matrixStats':

    anyMissing, rowMedians

Loading required package: mgcv

Loading required package: nlme


Attaching package: 'nlme'

The following object is masked from 'package:IRanges':

    collapse

The following object is masked from 'package:dplyr':

    collapse

This is mgcv 1.9-1. For overview type 'help("mgcv-package")'.

Loading required package: genefilter


Attaching package: 'genefilter'

The following objects are masked from 'package:MatrixGenerics':

    rowSds, rowVars

The following objects are masked from 'package:matrixStats':

    rowSds, rowVars

The following object is masked from 'package:readr':

    spec

Loading required package: BiocParallel

pbmc_cell_frequency | Removed 550 specimens because missing in meta data

plasma_ab_titer | Removed 6931 specimens because missing in meta data

plasma_cytokine_concentration_by_olink | Removed 495 specimens because missing in meta data

pbmc_cell_frequency | Removed 56 features because not in feature subset

plasma_ab_titer | Removed 48 features because not in feature subset

plasma_cytokine_concentration_by_olink | Removed 234 features because not in feature subset

t_cell_polarization | Removed 3 features because not in feature subset

plasma_cytokine_concentration_by_olink | Removed 300 features because qc warning

plasma_ab_titer | Removed 10540 measurements because wrong unit used

plasma_cytokine_concentration_by_olink | Removed 2400 measurements because wrong unit used

plasma_cytokine_concentration_by_legendplex | Removed 8 because specimen is outlier

plasma_cytokine_concentration_by_olink | Removed specimen 750, 760, 824, 833, 894, 903 because fraction of measurements below LOQ > 50%

plasma_ab_titer | Removed specimen 674, 675, 676 because fraction of measurements below LOD > 50%

converting counts to integer mode

Found 4 batches
Using null model in ComBat-seq.
Adjusting for 0 covariate(s) or covariate level(s)
Estimating dispersions
Fitting the GLM model
Shrinkage off - using GLM estimates for parameters
Adjusting the data

Found3batches

Adjusting for0covariate(s) or covariate level(s)

Standardizing Data across genes

Fitting L/S model and finding priors

Finding parametric adjustments

Adjusting the Data

Code

experimental_data <- 
  experimental_data[-which(names(experimental_data) == "pbmc_gene_expression_counts")]

experimental_predictors <- generate_wide_experimental_data(experimental_data=experimental_data,
                                                            impute="median", 
                                                            verbose=TRUE)

plasma_cytokine_concentration_by_olink | NA Fraction: 0.00288417166589755 | Imputed with median imputation

t_cell_activation | NA Fraction: 0.0576923076923077 | Removed samples: 681

t_cell_activation | NA Fraction: 0.0553691275167785 | Imputed with median imputation

t_cell_polarization | NA Fraction: 0.0134099616858238 | Imputed with median imputation

Code

hvg <- gene_meta %>% 
  dplyr::slice_max(mean_rank, n=1000)
experimental_predictors$pbmc_gene_expression <- 
  experimental_predictors$pbmc_gene_expression[, hvg$versioned_ensembl_gene_id_clean]

3 Conclusions

Not sure whether this is a bug, but the performance of models using either normalized or batch corrected data is almost the same, except for olink features (and probably also gex data)

4 Questions

5 Results

Code

task_meta <- list(
  task_11 = list(
    name = "task_11",
    header = "## Task 1.1",
    description = "Rank the individuals by IgG antibody levels against pertussis toxin (PT) that we detect in plasma 14 days post booster vaccinations."
  ),
  task_12 = list(
    name = "task_12",
    header = "## Task 1.2",
    description = "Rank the individuals by fold change of IgG antibody levels against pertussis toxin (PT) that we detect in plasma 14 days post booster vaccinations compared to titer values at day 0."
  ),
  task_21 = list(
    name = "task_21",
    header = "## Task 2.1",
    description = "Rank the individuals by predicted frequency of Monocytes on day 1 post boost after vaccination."
  ),
  task_22 = list(
    name = "task_22",
    header = "## Task 2.2",
    description = "Rank the individuals by fold change of predicted frequency of Monocytes on day 1 post booster vaccination compared to cell frequency values at day 0."
  ),
  task_31 = list(
    name = "task_31",
    header = "## Task 3.1",
    description = "Rank the individuals by predicted gene expression of CCL3 on day 3 post-booster vaccination."
  ),
  task_32 = list(
    name = "task_32",
    header = "## Task 3.2",
    description = "Rank the individuals by fold change of predicted gene expression of CCL3 on day 3 post booster vaccination compared to gene expression values at day 0."
  ),
  task_41 = list(
    name = "task_41",
    header = "## Task 4.1",
    description = "Rank the individuals based on their Th1/Th2 (IFN-g/IL-5) polarization ratio on Day 30 post-booster vaccination."
  )
)

Code

RENDER <- TRUE

make_flextable <- function(x) {
  if (RENDER) {
    x %>%
      flextable() %>% 
      bg(., bg = "#333333", part = "all") %>%
      color(., color = "white", part = "all") %>%
      set_table_properties(., align = "left") %>%
      flextable_to_rmd(ft) %>%
      return()
  } else {
    return(x)
  }
}

meta_data_covariates <- get_metadata_covariates(meta_data)

for (task in task_meta) {
  #task <- task_meta[[3]]
  
  cat(task$header)
  cat("\n\n")
  cat(task$description)
  cat("\n\n")
  
  cat(paste0("\n\n### ", "metadata model", "\n\n"))
  model_df <- target_list[[task$name]] %>%
    dplyr::left_join(meta_data_covariates, by="subject_id")
  set.seed(42)
  
  get_oob_perf(model_df=model_df) %>% 
    dplyr::mutate(mse = round(mse, 2), r2 = round(r2, 2), srho = round(srho, 2)) %>%
    make_flextable(.)
  
  get_loocv_perf(model_df=model_df) %>% 
    dplyr::mutate(mse = round(mse, 2), r2 = round(r2, 2), srho = round(srho, 2)) %>%
    make_flextable(.)
    
  get_cross_cohort_perf_single_repeated(model_df=model_df, meta_data=meta_data) %>%
    dplyr::mutate(srho_mean = round(srho_mean, 2), srho_sd = round(srho_sd, 2),
                  srho_baseline = round(srho_baseline, 2)) %>%
    make_flextable(.)
  
  cat("\n\n")
  
  for (modality in names(experimental_predictors)) {
    #modality <- names(experimental_predictors)[3]
    
    cat(paste0("\n\n### ", modality, "\n\n"))
    
    modality_data <- experimental_predictors[[modality]] %>%
      as.data.frame() %>% 
      tibble::rownames_to_column("specimen_id") %>%
      dplyr::mutate(specimen_id = as.numeric(specimen_id)) %>%
      dplyr::left_join((specimen_per_day$day_0 %>% dplyr::select(subject_id, specimen_id)),
                       by="specimen_id") %>%
      dplyr::filter(!is.na(subject_id)) %>%
      dplyr::select(-specimen_id)

    
    model_df <- target_list[[task$name]] %>%
      dplyr::left_join(modality_data, by="subject_id")
    
    if (sum(rowMeans(is.na(model_df)) > 0.1) > 0) {
      cat(paste0("\n\nRemoving ", sum(rowMeans(is.na(model_df)) > 0.1),
                 " targets because training data is missing\n\n"))
    }
    
    model_df <- model_df %>%
      tidyr::drop_na()
    
    set.seed(42)
    
    get_oob_perf(model_df=model_df) %>% 
      dplyr::mutate(mse = round(mse, 2), r2 = round(r2, 2), srho = round(srho, 2)) %>%
      make_flextable(.)
    
    get_loocv_perf(model_df=model_df) %>% 
      dplyr::mutate(mse = round(mse, 2), r2 = round(r2, 2), srho = round(srho, 2)) %>%
      make_flextable(.)
      
    # get_cross_cohort_perf_combinations(model_df=model_df, meta_data=meta_data) %>%
    #   dplyr::mutate(mse = round(mse, 2), r2 = round(r2, 2)) %>%
    # make_flextable(.)
      
    # get_cross_cohort_perf_single(model_df=model_df, meta_data=meta_data) %>%
    #   dplyr::mutate(mse = round(mse, 2), r2 = round(r2, 2),
    #                 srho = round(srho, 2), srho_baseline = round(srho_baseline, 2),
    #                 mse_tmean = round(mse_tmean, 2)) %>%
    #   make_flextable(.)
    
    get_cross_cohort_perf_single_repeated(model_df=model_df, meta_data=meta_data) %>%
      dplyr::mutate(srho_mean = round(srho_mean, 2), srho_sd = round(srho_sd, 2),
                    srho_baseline = round(srho_baseline, 2)) %>%
      make_flextable(.)
    
    cat("\n\n")
  }
}

5.1 Task 1.1

Rank the individuals by IgG antibody levels against pertussis toxin (PT) that we detect in plasma 14 days post booster vaccinations.

5.1.1 metadata model

mode	mse	r2	srho
oob	21,930,271	0.81	0.68

mode	mse	r2	srho
loocv	20,912,097	0.82	0.74

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.31	0.01	0.44	32	20
2022	2021	-0.34	0.04	0.60	20	32

5.1.2 pbmc_cell_frequency

Removing 3 targets because training data is missing

mode	mse	r2	srho
oob	28,569,686	0.76	0.73

mode	mse	r2	srho
loocv	29,632,541	0.75	0.71

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.33	0.04	0.44	29	20
2022	2021	-0.42	0.05	0.59	20	29

5.1.3 pbmc_gene_expression

mode	mse	r2	srho
oob	88,883,136	0.25	0.46

mode	mse	r2	srho
loocv	84,975,227	0.28	0.56

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.11	0.10	0.44	32	20
2022	2021	-0.09	0.05	0.60	20	32

5.1.4 plasma_ab_titer

mode	mse	r2	srho
oob	30,973,495	0.74	0.62

mode	mse	r2	srho
loocv	29,795,572	0.75	0.64

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.09	0.05	0.44	32	20
2022	2021	-0.08	0.05	0.60	20	32

5.1.5 plasma_cytokine_concentration_by_legendplex

Removing 3 targets because training data is missing

mode	mse	r2	srho
oob	16,199,037	0.86	0.76

mode	mse	r2	srho
loocv	16,019,496	0.86	0.74

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.30	0.06	0.48	31	18
2022	2021	0.03	0.03	0.61	18	31

5.1.6 plasma_cytokine_concentration_by_olink

Removing 2 targets because training data is missing

mode	mse	r2	srho
oob	48,451,390	0.59	0.74

mode	mse	r2	srho
loocv	46,729,315	0.61	0.74

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.42	0.05	0.52	32	18
2022	2021	-0.33	0.04	0.60	18	32

5.1.7 t_cell_activation

Removing 1 targets because training data is missing

mode	mse	r2	srho
oob	20,527,029	0.83	0.7

mode	mse	r2	srho
loocv	19,009,398	0.84	0.7

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.22	0.05	0.44	31	20
2022	2021	-0.13	0.03	0.57	20	31

5.1.8 t_cell_polarization

Removing 12 targets because training data is missing

mode	mse	r2	srho
oob	26,095,972	0.76	0.66

mode	mse	r2	srho
loocv	26,164,472	0.76	0.64

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.48	0.07	0.34	25	15
2022	2021	-0.24	0.03	0.62	15	25

5.2 Task 1.2

Rank the individuals by fold change of IgG antibody levels against pertussis toxin (PT) that we detect in plasma 14 days post booster vaccinations compared to titer values at day 0.

5.2.1 metadata model

mode	mse	r2	srho
oob	0.38	0.61	0.81

mode	mse	r2	srho
loocv	0.39	0.6	0.81

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.36	0.03	-0.89	32	20
2022	2021	0.04	0.02	-0.71	20	32

5.2.2 pbmc_cell_frequency

Removing 3 targets because training data is missing

mode	mse	r2	srho
oob	0.49	0.5	0.74

mode	mse	r2	srho
loocv	0.49	0.5	0.76

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.14	0.07	-0.89	29	20
2022	2021	0.16	0.04	-0.74	20	29

5.2.3 pbmc_gene_expression

mode	mse	r2	srho
oob	0.93	0.04	0.2

mode	mse	r2	srho
loocv	0.94	0.03	0.22

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.28	0.05	-0.89	32	20
2022	2021	0.23	0.04	-0.71	20	32

5.2.4 plasma_ab_titer

mode	mse	r2	srho
oob	0.47	0.51	0.78

mode	mse	r2	srho
loocv	0.48	0.5	0.78

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.33	0.05	-0.89	32	20
2022	2021	0.32	0.06	-0.71	20	32

5.2.5 plasma_cytokine_concentration_by_legendplex

Removing 3 targets because training data is missing

mode	mse	r2	srho
oob	0.55	0.41	0.64

mode	mse	r2	srho
loocv	0.55	0.41	0.65

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.01	0.06	-0.89	31	18
2022	2021	-0.20	0.02	-0.69	18	31

5.2.6 plasma_cytokine_concentration_by_olink

Removing 2 targets because training data is missing

mode	mse	r2	srho
oob	0.65	0.32	0.72

mode	mse	r2	srho
loocv	0.64	0.33	0.75

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.22	0.06	-0.93	32	18
2022	2021	0.09	0.06	-0.71	18	32

5.2.7 t_cell_activation

Removing 1 targets because training data is missing

mode	mse	r2	srho
oob	0.47	0.52	0.74

mode	mse	r2	srho
loocv	0.5	0.49	0.71

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.03	0.05	-0.89	31	20
2022	2021	-0.01	0.06	-0.72	20	31

5.2.8 t_cell_polarization

Removing 12 targets because training data is missing

mode	mse	r2	srho
oob	0.52	0.47	0.71

mode	mse	r2	srho
loocv	0.51	0.49	0.74

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.20	0.06	-0.91	25	15
2022	2021	0.29	0.03	-0.70	15	25

5.3 Task 2.1

Rank the individuals by predicted frequency of Monocytes on day 1 post boost after vaccination.

5.3.1 metadata model

mode	mse	r2	srho
oob	64.45	0.34	0.59

mode	mse	r2	srho
loocv	63.73	0.34	0.59

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.53	0.04	0.88	12	33
2020	2022	0.13	0.06	0.55	12	21
2021	2020	0.72	0.04	0.81	33	12
2021	2022	0.49	0.02	0.55	33	21
2022	2020	0.20	0.04	0.81	21	12
2022	2021	0.60	0.01	0.88	21	33

5.3.2 pbmc_cell_frequency

mode	mse	r2	srho
oob	50.79	0.48	0.7

mode	mse	r2	srho
loocv	50.61	0.48	0.7

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.81	0.02	0.88	12	33
2020	2022	0.54	0.02	0.55	12	21
2021	2020	0.72	0.02	0.81	33	12
2021	2022	0.58	0.02	0.55	33	21
2022	2020	0.61	0.04	0.81	21	12
2022	2021	0.82	0.01	0.88	21	33

5.3.3 pbmc_gene_expression

Removing 1 targets because training data is missing

mode	mse	r2	srho
oob	84.34	0.14	0.32

mode	mse	r2	srho
loocv	84.4	0.14	0.35

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.34	0.03	0.88	11	33
2020	2022	0.11	0.05	0.55	11	21
2021	2020	0.67	0.08	0.84	33	11
2021	2022	0.58	0.05	0.55	33	21
2022	2020	0.41	0.03	0.84	21	11
2022	2021	0.49	0.03	0.88	21	33

5.3.4 plasma_ab_titer

Removing 16 targets because training data is missing

mode	mse	r2	srho
oob	75.6	0.23	0.5

mode	mse	r2	srho
loocv	76.96	0.21	0.5

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.33	0.04	0.55	29	21
2022	2021	0.37	0.05	0.88	21	29

5.3.5 plasma_cytokine_concentration_by_legendplex

Removing 20 targets because training data is missing

mode	mse	r2	srho
oob	63.89	0.34	0.56

mode	mse	r2	srho
loocv	66.66	0.31	0.53

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.25	0.06	0.53	28	18
2022	2021	0.81	0.02	0.87	18	28

5.3.6 plasma_cytokine_concentration_by_olink

Removing 14 targets because training data is missing

mode	mse	r2	srho
oob	66.35	0.3	0.49

mode	mse	r2	srho
loocv	65.97	0.3	0.51

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.13	0.05	0.55	33	19
2022	2021	-0.10	0.06	0.88	19	33

5.3.7 t_cell_activation

Removing 16 targets because training data is missing

mode	mse	r2	srho
oob	58.13	0.4	0.6

mode	mse	r2	srho
loocv	59.5	0.39	0.62

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.48	0.02	0.60	30	20
2022	2021	0.75	0.01	0.88	20	30

5.3.8 t_cell_polarization

Removing 25 targets because training data is missing

mode	mse	r2	srho
oob	68.76	0.31	0.6

mode	mse	r2	srho
loocv	69.47	0.31	0.58

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.33	0.03	0.54	25	16
2022	2021	0.72	0.03	0.86	16	25

5.4 Task 2.2

Rank the individuals by fold change of predicted frequency of Monocytes on day 1 post booster vaccination compared to cell frequency values at day 0.

5.4.1 metadata model

mode	mse	r2	srho
oob	0.11	-0.04	-0.04

mode	mse	r2	srho
loocv	0.11	-0.03	0

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	-0.11	0.03	-0.22	12	33
2020	2022	-0.31	0.02	-0.21	12	21
2021	2020	0.52	0.03	-0.43	33	12
2021	2022	-0.31	0.04	-0.21	33	21
2022	2020	-0.44	0.04	-0.43	21	12
2022	2021	0.00	0.03	-0.22	21	33

5.4.2 pbmc_cell_frequency

mode	mse	r2	srho
oob	0.11	-0.01	0.1

mode	mse	r2	srho
loocv	0.11	-0.01	0.12

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.39	0.03	-0.22	12	33
2020	2022	0.13	0.04	-0.21	12	21
2021	2020	0.16	0.06	-0.43	33	12
2021	2022	0.15	0.04	-0.21	33	21
2022	2020	0.18	0.04	-0.43	21	12
2022	2021	0.09	0.02	-0.22	21	33

5.4.3 pbmc_gene_expression

Removing 1 targets because training data is missing

mode	mse	r2	srho
oob	0.12	-0.09	-0.08

mode	mse	r2	srho
loocv	0.12	-0.11	-0.14

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.05	0.05	-0.22	11	33
2020	2022	-0.36	0.07	-0.21	11	21
2021	2020	0.03	0.11	-0.31	33	11
2021	2022	0.33	0.06	-0.21	33	21
2022	2020	-0.68	0.06	-0.31	21	11
2022	2021	0.09	0.06	-0.22	21	33

5.4.4 plasma_ab_titer

Removing 16 targets because training data is missing

mode	mse	r2	srho
oob	0.08	-0.03	0.19

mode	mse	r2	srho
loocv	0.08	-0.01	0.14

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.24	0.05	-0.21	29	21
2022	2021	0.11	0.05	-0.25	21	29

5.4.5 plasma_cytokine_concentration_by_legendplex

Removing 20 targets because training data is missing

mode	mse	r2	srho
oob	0.06	0.14	0.41

mode	mse	r2	srho
loocv	0.06	0.16	0.43

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.05	0.07	-0.27	28	18
2022	2021	0.04	0.05	-0.25	18	28

5.4.6 plasma_cytokine_concentration_by_olink

Removing 14 targets because training data is missing

mode	mse	r2	srho
oob	0.07	0.09	0.1

mode	mse	r2	srho
loocv	0.07	0.08	0.02

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.17	0.06	-0.28	33	19
2022	2021	-0.12	0.03	-0.22	19	33

5.4.7 t_cell_activation

Removing 16 targets because training data is missing

mode	mse	r2	srho
oob	0.08	-0.11	0.09

mode	mse	r2	srho
loocv	0.08	-0.1	0.09

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.2	0.01	-0.16	30	20
2022	2021	0.0	0.04	-0.23	20	30

5.4.8 t_cell_polarization

Removing 25 targets because training data is missing

mode	mse	r2	srho
oob	0.08	-0.12	0.06

mode	mse	r2	srho
loocv	0.08	-0.12	0.07

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.00	0.07	-0.29	25	16
2022	2021	0.13	0.03	-0.24	16	25

5.5 Task 3.1

Rank the individuals by predicted gene expression of CCL3 on day 3 post-booster vaccination.

5.5.1 metadata model

mode	mse	r2	srho
oob	1	0.18	0.46

mode	mse	r2	srho
loocv	1	0.18	0.46

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.35	0.02	0.57	26	36
2020	2022	-0.06	0.04	0.53	26	21
2021	2020	0.01	0.03	0.27	36	26
2021	2022	0.38	0.05	0.53	36	21
2022	2020	-0.01	0.02	0.27	21	26
2022	2021	0.40	0.02	0.57	21	36

5.5.2 pbmc_cell_frequency

Removing 18 targets because training data is missing

mode	mse	r2	srho
oob	1	0.15	0.33

mode	mse	r2	srho
loocv	1.03	0.13	0.32

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.40	0.03	0.59	11	33
2020	2022	-0.13	0.07	0.53	11	21
2021	2020	0.05	0.08	0.22	33	11
2021	2022	0.12	0.03	0.53	33	21
2022	2020	0.04	0.04	0.22	21	11
2022	2021	-0.03	0.05	0.59	21	33

5.5.3 pbmc_gene_expression

mode	mse	r2	srho
oob	0.96	0.21	0.5

mode	mse	r2	srho
loocv	0.94	0.23	0.51

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.03	0.04	0.57	26	36
2020	2022	0.31	0.05	0.53	26	21
2021	2020	-0.37	0.04	0.27	36	26
2021	2022	0.51	0.06	0.53	36	21
2022	2020	0.03	0.06	0.27	21	26
2022	2021	0.32	0.03	0.57	21	36

5.5.4 plasma_ab_titer

Removing 30 targets because training data is missing

mode	mse	r2	srho
oob	0.95	0.12	0.33

mode	mse	r2	srho
loocv	0.94	0.13	0.36

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.10	0.05	0.53	32	21
2022	2021	0.26	0.05	0.60	21	32

5.5.5 plasma_cytokine_concentration_by_legendplex

Removing 34 targets because training data is missing

mode	mse	r2	srho
oob	1.05	0.05	0.23

mode	mse	r2	srho
loocv	1.06	0.04	0.24

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.22	0.08	0.53	31	18
2022	2021	0.30	0.04	0.60	18	31

5.5.6 plasma_cytokine_concentration_by_olink

Removing 28 targets because training data is missing

mode	mse	r2	srho
oob	1.23	0.06	0.21

mode	mse	r2	srho
loocv	1.27	0.04	0.12

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.01	0.05	0.60	36	19
2022	2021	-0.09	0.04	0.57	19	36

5.5.7 t_cell_activation

Removing 30 targets because training data is missing

mode	mse	r2	srho
oob	1.02	0.18	0.46

mode	mse	r2	srho
loocv	1.05	0.15	0.42

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.54	0.02	0.53	33	20
2022	2021	0.20	0.03	0.62	20	33

5.5.8 t_cell_polarization

Removing 40 targets because training data is missing

mode	mse	r2	srho
oob	1.04	0.17	0.39

mode	mse	r2	srho
loocv	1.03	0.17	0.37

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.48	0.03	0.46	27	16
2022	2021	0.16	0.04	0.72	16	27

5.6 Task 3.2

Rank the individuals by fold change of predicted gene expression of CCL3 on day 3 post booster vaccination compared to gene expression values at day 0.

5.6.1 metadata model

mode	mse	r2	srho
oob	1.14	0.01	0.1

mode	mse	r2	srho
loocv	1.15	0	0.08

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.21	0.01	-0.43	26	36
2020	2022	-0.26	0.03	-0.19	26	21
2021	2020	-0.16	0.02	-0.33	36	26
2021	2022	-0.08	0.05	-0.19	36	21
2022	2020	-0.35	0.02	-0.33	21	26
2022	2021	0.16	0.05	-0.43	21	36

5.6.2 pbmc_cell_frequency

Removing 18 targets because training data is missing

mode	mse	r2	srho
oob	0.96	0	0

mode	mse	r2	srho
loocv	0.97	-0.01	-0.05

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.26	0.03	-0.37	11	33
2020	2022	-0.13	0.04	-0.19	11	21
2021	2020	-0.18	0.10	-0.43	33	11
2021	2022	-0.17	0.04	-0.19	33	21
2022	2020	-0.27	0.03	-0.43	21	11
2022	2021	0.04	0.04	-0.37	21	33

5.6.3 pbmc_gene_expression

mode	mse	r2	srho
oob	1.2	-0.04	0.04

mode	mse	r2	srho
loocv	1.17	-0.02	0.11

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2020	2021	0.11	0.04	-0.43	26	36
2020	2022	-0.28	0.04	-0.19	26	21
2021	2020	0.18	0.05	-0.33	36	26
2021	2022	-0.34	0.05	-0.19	36	21
2022	2020	-0.45	0.05	-0.33	21	26
2022	2021	-0.13	0.05	-0.43	21	36

5.6.4 plasma_ab_titer

Removing 30 targets because training data is missing

mode	mse	r2	srho
oob	1.26	0.09	0.43

mode	mse	r2	srho
loocv	1.22	0.12	0.42

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.42	0.04	-0.19	32	21
2022	2021	0.49	0.02	-0.52	21	32

5.6.5 plasma_cytokine_concentration_by_legendplex

Removing 34 targets because training data is missing

mode	mse	r2	srho
oob	1.37	0.04	-0.01

mode	mse	r2	srho
loocv	1.36	0.05	0.01

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.21	0.04	-0.18	31	18
2022	2021	0.32	0.04	-0.54	18	31

5.6.6 plasma_cytokine_concentration_by_olink

Removing 28 targets because training data is missing

mode	mse	r2	srho
oob	1.28	0.05	0.21

mode	mse	r2	srho
loocv	1.27	0.06	0.22

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.29	0.10	-0.12	36	19
2022	2021	0.15	0.05	-0.43	19	36

5.6.7 t_cell_activation

Removing 30 targets because training data is missing

mode	mse	r2	srho
oob	1.25	0.08	0.33

mode	mse	r2	srho
loocv	1.3	0.04	0.33

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.20	0.04	-0.24	33	20
2022	2021	0.39	0.01	-0.45	20	33

5.6.8 t_cell_polarization

Removing 40 targets because training data is missing

mode	mse	r2	srho
oob	1.51	-0.15	0.04

mode	mse	r2	srho
loocv	1.49	-0.13	0.04

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.24	0.05	-0.28	27	16
2022	2021	0.09	0.04	-0.37	16	27

5.7 Task 4.1

Rank the individuals based on their Th1/Th2 (IFN-g/IL-5) polarization ratio on Day 30 post-booster vaccination.

5.7.1 metadata model

mode	mse	r2	srho
oob	4.52	0.03	0.21

mode	mse	r2	srho
loocv	4.53	0.03	0.18

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.47	0.05	0.74	27	16
2022	2021	0.53	0.02	0.55	16	27

5.7.2 pbmc_cell_frequency

Removing 2 targets because training data is missing

mode	mse	r2	srho
oob	4.34	0.08	0.39

mode	mse	r2	srho
loocv	4.25	0.1	0.42

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.48	0.03	0.74	25	16
2022	2021	0.67	0.01	0.62	16	25

5.7.3 pbmc_gene_expression

mode	mse	r2	srho
oob	4.92	-0.05	0.16

mode	mse	r2	srho
loocv	4.91	-0.05	0.14

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.24	0.07	0.74	27	16
2022	2021	0.39	0.03	0.55	16	27

5.7.4 plasma_ab_titer

Removing 2 targets because training data is missing

mode	mse	r2	srho
oob	5.22	-0.07	-0.09

mode	mse	r2	srho
loocv	5.3	-0.09	-0.14

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.44	0.04	0.74	25	16
2022	2021	0.19	0.03	0.56	16	25

5.7.5 plasma_cytokine_concentration_by_legendplex

Removing 5 targets because training data is missing

mode	mse	r2	srho
oob	5.59	-0.1	-0.03

mode	mse	r2	srho
loocv	5.57	-0.09	-0.02

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.06	0.05	0.79	24	14
2022	2021	0.53	0.02	0.52	14	24

5.7.6 plasma_cytokine_concentration_by_olink

Removing 2 targets because training data is missing

mode	mse	r2	srho
oob	5.37	-0.11	-0.26

mode	mse	r2	srho
loocv	5.48	-0.13	-0.28

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	-0.15	0.06	0.73	27	14
2022	2021	0.10	0.05	0.55	14	27

5.7.7 t_cell_activation

Removing 2 targets because training data is missing

mode	mse	r2	srho
oob	4.45	0.07	0.33

mode	mse	r2	srho
loocv	4.49	0.07	0.28

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.42	0.05	0.83	26	15
2022	2021	0.55	0.02	0.59	15	26

5.7.8 t_cell_polarization

mode	mse	r2	srho
oob	4.25	0.09	0.26

mode	mse	r2	srho
loocv	4.16	0.11	0.3

trainset	testset	srho_mean	srho_sd	srho_baseline	train_n	test_n
2021	2022	0.08	0.03	0.74	27	16
2022	2021	0.29	0.02	0.55	16	27