Document is current

Article History

Received: 22 August 2015

Accepted: 16 October 2015

First Online: 16 November 2015

Change Date: 31 August 2017

Change Type: Corrigendum

Change Details: In the version of this article initially published, the third and fourth paragraphs of Online Methods section “Differential stability in cortex and resting state network analysis” read as follows: The next step was to map the Allen Human Brain Atlas (AHBA) tissue samples to the HCP 52 region parcellation so that comparison could be made. Using the MNI centroid coordinate of the AHBA samples, and by manually examining each of the AHBA tissue samples using the online tools, one can assign a set of HCP space voxels to each AHBA tissue sample. As each of the 52 parcels is composed of a set of voxels, we now have potentially one-to-many map from AHBA tissue to HCP parcels. If all ABHA tissue samples belong to a common HCP parcel, we average the gene expression of that tissue in the corresponding parcel. However, some of the 52 parcels represent smaller regions of the brain and therefore there is no unique assignment of AHBA gene expression tissue samples to that region. Therefore, if a collection of AHBA tissue samples intersects more than one region, we average the gene expression values as before but fractionally weight the expression contribution to each of the interesting HCP parcels. This has the effect of allowing some assignment of expression without overweighting non-unique samples. Supplementary Table 12 gives the sample distribution by parcels as well as the uniquely assigned samples. To obtain the expression correlation matrix for a given gene (Fig. 7c, right panel), we transformed the expression values of that gene into <i>z</i>-scores over all the sampled brain regions (averaging sample data for those samples contained in the same parcel) and calculated the coexpression as the outer product of this <i>z</i>-score vector. Thus, if two regions both show high expression or low expression of the gene of interest, they will have a high positive coexpression value for that gene, whereas if they show opposite expression patterns, they will have a large negative value for that gene. After generating these matrices, we compared each of the 17,348 gene coexpression matrices to the parcellated connectome matrix by calculating the Pearson's correlation between the vectorized elements above the diagonal of the matrices (Fig. 7d). We also obtained a significance value for each gene-connectivity comparison using the randomized gene coexpression matrices. Supplementary Table 12 gives the complete distribution of tissue samples by HCP parcel for the 52 regions and the functional genetic correlations and <i>P</i>-values. In the current version, these paragraphs have been rewritten to unambiguously explain how each RSN parcel was mapped to the AHBA samples. The original version did not clearly delineate the approach for each of the three possible cases in which RSN parcels could overlap the AHBA samples. The new text also has an additional paragraph describing the rationale behind the two sets of <i>P</i>-values included in Supplementary Table 12. The error has been corrected in the HTML and PDF versions of the article.

Competing interests

: The authors declare no competing financial interests.