Proteomics • ckbtools

Introduction to CKB proteomics data

The China Kadoorie Biobank includes proteomics data from a nested case-cohort study of ~2k MI cases and ~2k subcohort participants. Plasma samples were analysed using two proteomics platforms: Olink Explore (~3k proteins) and SomaScan (~7k proteins).

Details of the CKB Data Sharing Policy, data release schedules and data request application procedures are available at www.ckbiobank.org/data-access.

Observational proteomics analyses

The functions in this package can be used to prepare analysis-ready data sets from raw CKB proteomics data and perform some observational proteomics analyses as shown in these articles:

Associations of proteins with participant characteristics demonstrates how to use this package to estimate associations between proteins and participant characteristics. The methods used are based on the papers “Associations of 2923 Olink proteins with demographic, lifestyle, environmental and health characteristics in middle-aged Chinese adults” and “Associations of 6600 SomaScan proteins with demographic, lifestyle, environmental and health characteristics in Chinese adults”.
Associations of proteins with incident disease demonstrates how to use this package to estimate associations between proteins and risk of incident disease using Cox proportional hazards models. The methods used are from “Proteome-wide assessment of ~9500 plasma proteins with major diseases and multi-morbidity in a Chinese population”.
Associations of proteins in a case-subcohort study demonstrates how to use this package to estimate associations between proteins and risk of incidence disease using Cox proportional hazards models with methods appropriate for a case-subcohort study design. The methods used are from “Plasma Proteomics to Identify Drug Targets for Ischemic Heart Disease”.

Preparing proteomics data

This package includes several functions to assist with preparing the CKB proteomics data for analysis.

Import data

First, all relevant protein data must be imported into R. For example, use readRDS() for RDS format data or read.csv() for csv data.

raw_olink_meta_data <- readRDS("olink_explore_meta.rds")
raw_olink_protein_data <- readRDS("data_baseline_olink_explore.rds")

raw_somascan_meta_data <- readRDS("somalogic_meta.rds")
raw_somascan_protein_data <- readRDS("data_baseline_somalogic.rds")
raw_somascan_sample_data  <- readRDS("data_baseline_somalogic_samples.rds")

For demonstration, the synthetic data sets available in this package can be used:

ckbtools_raw_olink_meta_data
ckbtools_raw_olink_protein_data
ckbtools_raw_somascan_meta_data
ckbtools_raw_somascan_protein_data
ckbtools_raw_somascan_sample_data

Prepare protein data

To use the proteomics data preparation functions, the CKB data release number must be set using options(). The release number for CKB data can be found in the release.txt file. This ensures the correct code is used for the format of the supplied data.

options(ckb.data.release = "19.02")

To prepare protein meta data, use the prepare_olink_meta_data() and prepare_somascan_meta_data() functions. These return data frames with one row per assay/aptamer and columns for protein details such as UniProt ID. The names argument specifies the naming system used for proteins: “protein” for names based on assay (for Olink) or target (for SomaScan); or “id” for names based on OlinkID (Olink ID) or SeqID (SomaLogic sequence identifier).

olink_meta_data <- prepare_olink_meta_data(ckbtools_raw_olink_meta_data,
                                           ckbtools_raw_olink_protein_data,
                                           names = "id")
somascan_meta_data <- prepare_somascan_meta_data(ckbtools_raw_somascan_meta_data,
                                                 names = "id")

To prepare protein data, use the prepare_olink_data() and prepare_somascan_data() functions. The functions each create a ‘wide’ data set with one row per sample/participant and columns for each protein measure.

olink_protein_data <- prepare_olink_data(ckbtools_raw_olink_protein_data,
                                         olink_meta_data)

# ✔ Protein values with QC warnings have been replaced with missing values.
# [Default]

The transform argument can be used to apply a transformation to all SomaScan protein values (for example "log2" will log₂ transform the RFU values).

somascan_protein_data <- prepare_somascan_data(ckbtools_raw_somascan_protein_data,
                                               ckbtools_raw_somascan_sample_data,
                                               somascan_meta_data,
                                               transform = "log2")

# ✔ RFU values have been log transformed (base 2). [Default]

# ✔ Participant samples with hybcontrolnormscale out of range (i.e. less than 0.4
# or greater than 2.5) have been removed. [Default]

# ✔ Data for the second sample (for participants who had a sample run twice) have
# been removed. [Default]

To merge participant data with protein data, use the prepare_proteomics_analysis_dataset() function. Only observations in both the participant data and proteomics data sets are retained. There must be a column in the participant data named “subcohort” that identifies observations that are part of the subcohort.

ckbtools_participant_data$subcohort <- ckbtools_participant_data$proteomics_subcohort

The plate_correction argument can be used to correct protein values for plate effects. Set this to "median-subcohort"/"mean-subcohort" to centre on medians/medians across plates based on subcohort values.

The scaling argument can be used to scale protein values. For example, "scale-subcohort" will scale values (as in the R function scale()) using the mean and standard deviation of subcohort values. "rint" will apply a rank inverse normal transformation to all values.

olink_analysis_data <- prepare_proteomics_analysis_dataset(
  participant_data  = ckbtools_participant_data,
  protein_data      = olink_protein_data,
  protein_meta_data = olink_meta_data,
  plate_correction  = "median-subcohort",
  scaling           = "none"
)

# ✔ Plate correction 'median-subcohort' has been applied to protein values.

# ✔ No scaling of protein values has been applied. [Default]

somascan_analysis_data <- prepare_proteomics_analysis_dataset(
  participant_data  = ckbtools_participant_data,
  protein_data      = somascan_protein_data,
  protein_meta_data = somascan_meta_data,
  plate_correction  = "none",
  scaling           = "none"
)

# ✔ No plate correction has been applied to protein values. [Default]
# ✔ No scaling of protein values has been applied. [Default]

In summary, the proteomics data preparation steps in order are:

Log transformation (using the transform argument in prepare_somascan_data()).
Removing values/samples with QC warnings (using the qc argument in prepare_olink_data() and prepare_somascan_data()).
Removing repeated samples (using the repeat_samples argument in prepare_somascan_data()).
Merging proteomics and participant data (using the participant_data and protein_data arguments in prepare_proteomics_analysis_dataset()). Note that this is an “inner join” so that only matching observations in both data sets are kept.
Plate correction (using the plate_correction argument in prepare_proteomics_analysis_dataset()).
Scaling protein values (using the scaling argument in prepare_proteomics_analysis_dataset()).

Descriptive summary statistics

Use summarise_continuous() to create basic descriptive summary statistics for proteins.

protein_summary_stats <- summarise_continuous(olink_analysis_data,
                                              cols = olink_meta_data$name,
                                              by = "sex")

protein_summary_stats[1:12, 1:7]

#         var    sex n_nonmissing n_missing percent_missing       mean        sd
# 1  OID13950   <NA>          987        13       1.3000000 0.17825730 0.7761187
# 2  OID13950   Male          413         6       1.4319809 0.18644455 0.7509019
# 3  OID13950 Female          574         7       1.2048193 0.17236647 0.7943648
# 4  OID01935   <NA>          988        12       1.2000000 0.16706032 0.8281963
# 5  OID01935   Male          413         6       1.4319809 0.19642842 0.8623041
# 6  OID01935 Female          575         6       1.0327022 0.14596636 0.8029020
# 7  OID12648   <NA>          993         7       0.7000000 0.13701380 0.8096633
# 8  OID12648   Male          416         3       0.7159905 0.10770473 0.7944150
# 9  OID12648 Female          577         4       0.6884682 0.15814478 0.8205168
# 10 OID91452   <NA>          994         6       0.6000000 0.07626006 0.8198431
# 11 OID91452   Male          415         4       0.9546539 0.09757931 0.8470904
# 12 OID91452 Female          579         2       0.3442341 0.06097942 0.8001332

Olink: Limit of detection

The number and percentages of Olink protein values below the limit of detection can be obtained as follows:

olink_meta_data <- prepare_olink_meta_data(ckbtools_raw_olink_meta_data,
                                           ckbtools_raw_olink_protein_data)
olink_protein_data <- prepare_olink_data(ckbtools_raw_olink_protein_data,
                                         olink_meta_data,
                                         columns = "lod")

# ✔ Protein values with QC warnings have been replaced with missing values.
# [Default]

olink_meta_data %>%
  dplyr::arrange(name) %>%
  dplyr::select(name) %>%
  dplyr::rowwise() %>%
  dplyr::mutate(n = sum(!is.na(olink_protein_data[name])),
                below_lod = sum(olink_protein_data[name] <
                                  olink_protein_data[paste0("lod_", name)], na.rm = TRUE)) %>%
  dplyr::ungroup() %>%
  dplyr::mutate(p = 100 * below_lod / n)

# # A tibble: 48 × 4
#    name       n below_lod     p
#    <chr>  <int>     <int> <dbl>
#  1 ACAA1    991       205  20.7
#  2 BAMBI    994       187  18.8
#  3 BCR      994       206  20.7
#  4 CHEK2    988       213  21.6
#  5 CLASP1   991       188  19.0
#  6 CNGB3    986       205  20.8
#  7 CNTF     988       223  22.6
#  8 COMT     988       199  20.1
#  9 DAND5    983       204  20.8
# 10 DAPP1    984       199  20.2
# # ℹ 38 more rows