Addendum to: Nature Communications; https://doi.org/10.1038/s41467-018-03819-3; published online 16 Apr 2018
In this article, a meta-analysis of the significant variants in the UK Biobank sample with a previously published study by 23andMe was included. Due to the individual study results being reported on different scales, the meta-analytic point estimates were not reliable. We have re-analysed the 17 significant variants in the UK Biobank using a comparable approach to 23andMe in Plink. We present an updated version of Table 1 associated with this Addendum that includes the newly calculated comparable effect size estimates for UK Biobank, the updated meta-analysis results and we have extended the table legend to provide information on the updated columns. As a consequence of the new analysis, the original statement in the first paragraph under the subheading “Genome-wide association study of depression” reading “All 17 variants remained significant (P < 5 × 10−8) in the meta-analysis” should be amended to “10 variants remained significant (P < 5 × 10−8) in the meta-analysis”. There was no change to the direction of allelic effect in either cohort.
The fourth sentence in Methods section under the subheading “Replication cohort and meta-analysis” on page 8 was incomplete and should read “Additionally, we used Metal36 to conduct an inverse variance-weighted meta-analysis, using LD score regression intercepts13 for genomic inflation control.”, adding “using LD score regression intercepts13 for genomic inflation control” at the end of the sentence.
We have also included transformed effect sizes and standard errors in our summary statistics deposited on the Edinburgh DataShare website and have updated the doi to https://doi.org/10.7488/ds/2350.
In a second related issue, we used the new BGENIE software package to generate the results with which to conduct gene, region and gene-set analyses to generate data presented in the “Gene and region-based analysis” and “Gene-set pathway analysis” sections, Table 2, Supplementary Table 6, Supplementary Data 6–12, and Supplementary Figures 7–9. BGENIE reported minor allele frequencies (MAF) across the whole of UK Biobank (n = 487,409) rather than based on those individuals that were included in each of the association analyses (broad depression n = 322,580; probable major depressive disorder n = 174,519; International Classification of Diseases-coded major depressive disorder n = 217,584). Therefore the results reported in the “Gene and region-based analysis” and “Gene-set pathway analysis” were based on the MAF across the whole of UK Biobank as opposed to what we described in the methods section. To correct these errors in our initial analysis, we have now re-analysed the data using the MAF based on only those individuals included in the respective association analysis. This has resulted in 73 rather than 78 significant genes for broad depression, three rather than two genes for probable MDD, and zero genes rather than one gene for ICD-coded MDD which was originally described in the “Gene and region-based analysis” section in the results. The gene-based results descri bed on page 3 under the subheading “Gene and region-based analyses” should read “We used the MAGMA18 package to identify genes with a significant effect (P < 2.77 × 10−6) on each phenotype. There were 73 genes significantly associated with broad depression (Supplementary Data 6, associated with this Addendum), and three genes that were associated with probable MDD (Supplementary Data 7, associated with this Addendum)”. Updated versions of Supplementary Note 1, Supplementary Figures 7, 8, and 9 in the Supplementary Information file, and Supplementary Data 6 and 7 are included here to reflect these changes in the data. The original Supplementary Data 8 should be disregarded as with the new analysis no genome-wide significant SNPs for ICD-coded MDD are detected. In the region-based analysis described on page 3 number of regions assessed increased from 8308 to 8345 requiring a slight change in the significance threshold from 6.02 × 10−8 to 5.99 × 10−6 so that the sentence in the second paragraph under the subheading “Gene and region-based analyses” now reads “We also used MAGMA to identify genomic regions, defined by recombination hotspots, with a statistically significant effect (P < 5.99 × 10−6) on each phenotype”. The number of significant regions remained the same and updated estimates of effect sizes of regions may be found in the updated Supplementary Data 8, 10, and 11 associated with this Addendum. The significant gene-sets described under the subheading “Gene-set pathway analysis” remained the same with slight differences in number of genes in each pathway, effect sizes and p-values. The first sentence in this section now reads: “We conducted gene-set enrichment analysis19,20 and identified five significant pathways for broad depression after applying correction for multiple testing; GO_EXCITATORY_SYNAPSE (beta = 0.342 ± 0.069, Pcorrected = 0.003), GO_MECHANOSENSORY_BEHAVIOR (beta = 1.270 ± 0.294, Pcorrected = 0.047), GO_POSTSYNAPSE (beta = 0.248 ± 0.050, Pcorrected = 0.003), GO_NEURON_SPINE (beta = 0.390 ± 0.089, Pcorrected = 0.019) and GO_DENDRITE (beta = 0.200 ± 0.044, Pcorrected = 0.021) (Table 2)”. Amended versions of Table 2, Supplementary Data 12 and Supplementary Table 6 in the Supplementary Information file are included in this Addendum.
The described errors have not been fixed in the original article. We reaffirm that the changes do not change the main conclusions of the manuscript.
*A full list of consortium members appears at the end of the paper.