Skip to contents

Score a study against the 10-point QC checklist

Source: R/01_ingest.R
mx_qc_study.Rd

Applies 10 data-level QC criteria. Each criterion is recorded as "pass", "fail" or "not evaluable" (the information needed is absent, e.g. alignment rate is not deposited in GEO). qc_score is the number of criteria passed. Criteria that could not be evaluated are not counted as passed and the score is not rescaled, so a score is only comparable between studies together with the number of evaluable criteria. Decide on inclusion with mx_filter_studies, which works on failures rather than the score.

Usage

mx_qc_study(study, min_genes = 15000)

Arguments

study

A metaXpressStudy object.

min_genes

Integer. Threshold for criterion 6. Default 15000, appropriate for an unfiltered human or mouse gene-level matrix; lower it for matrices already restricted (e.g. to protein-coding genes).

Value

The input study with qc_score filled (criteria passed, 0–10). The per-criterion table (criterion, name, status, score, details) is stored in S4Vectors::metadata(study)$qc and in attr(study, "qc_details").

Details

The 10 criteria are:

  1. Minimum sample size: at least 3 samples in every condition group (Schurch et al. 2016).

  2. Sequencing depth: median library size >= 10 million assigned reads.

  3. Alignment rate: mean >= 70% (column alignment_rate in colData, fraction or percent).

  4. rRNA contamination: mean < 10% (column rrna_fraction).

  5. Duplicate rate: mean < 50% (column duplicate_rate).

  6. Gene detection: at least min_genes genes with a non-zero count in >= 50% of samples.

  7. Condition labels complete: every sample has a non-missing, non-empty condition.

  8. Clear contrast: exactly two condition levels. Studies with more groups fail until they are subset to the two compared groups.

  9. No batch-condition confounding: batch and condition are not perfectly aliased (Leek et al. 2010). Not evaluable without a batch column.

  10. Raw counts: non-negative integers with maximum > 1,000, i.e. untransformed counts suitable for count models.

References

Schurch, N.J. et al. (2016) How many biological replicates are needed in an RNA-seq experiment and which differential expression tool should you use? RNA, 22(6), 839–851. doi:10.1261/rna.053959.115

Leek, J.T. et al. (2010) Tackling the widespread and critical impact of batch effects in high-throughput data. Nature Reviews Genetics, 11(10), 733–739. doi:10.1038/nrg2825

Heberle, B.A. et al. (2024) Systematic review and meta-analysis of bulk RNAseq studies in human Alzheimer's disease brain tissue. bioRxiv. doi:10.1101/2024.11.07.622520 (complementary study-level reporting quality assessment).

Examples

if (requireNamespace("airway", quietly = TRUE)) {
  data("airway", package = "airway")
  study <- mx_from_se(airway, condition_col = "dex", accession = "GSE52778")
  study <- mx_qc_study(study)
  study@qc_score
  S4Vectors::metadata(study)$qc
}
#>    criterion              name        status score
#> 1          1       sample size          pass     1
#> 2          2  sequencing depth          pass     1
#> 3          3    alignment rate not evaluable    NA
#> 4          4     rRNA fraction not evaluable    NA
#> 5          5    duplicate rate not evaluable    NA
#> 6          6    gene detection          pass     1
#> 7          7  condition labels          pass     1
#> 8          8    clear contrast          pass     1
#> 9          9 batch confounding not evaluable    NA
#> 10        10        raw counts          pass     1
#>                                                       details
#> 1                        Min samples per group: 4 (need >= 3)
#> 2                  Median assigned reads: 20.9M (need >= 10M)
#> 3                             'alignment_rate' not in colData
#> 4                              'rrna_fraction' not in colData
#> 5                             'duplicate_rate' not in colData
#> 6  Genes detected in >= 50% of samples: 24706 (need >= 15000)
#> 7                       Samples with a condition label: 8 / 8
#> 8                        Condition levels: 2 (need exactly 2)
#> 9                    No multi-level 'batch' column in colData
#> 10                                    Raw integer counts: yes