Erratum for the Research Article “Patterns and ecological drivers of ocean viral communities” by J. R. Brum, J. Cesar Ignacio-Espinoza, S. Roux, G. Doulcier, S. G. Acinas, A. Alberti, S. Chaffron, C. Cruaud, C. de Vargas, J. M. Gasol, G. Gorsky, A. C. Gregory, L. Guidi, P. Hingamp, D. Iudicone, F. Not, H. Ogata, S. Pesant, B. T. Poulos, S. M. Schwenck, S. Speich, C. Dimier, S. Kandels-Lewis, M. Picheral, S. Searson, Tara Oceans Coordinators, P. Bork, C. Bowler, S. Sunagawa, P. Wincker, E. Karsenti, M. B. Sullivan

+ See all authors and affiliations

Science  12 Aug 2016:
Vol. 353, Issue 6300, aah6502
DOI: 10.1126/science.aah6502

Correction to Brum et al., Science 348, 1261498 (2015)

In the Research Article “Patterns and ecological drivers of ocean viral communities,” the Tara Oceans coordinators provided numbers of protein clusters (PCs). The authors have since found that the PC numbers are not reproducible from the published data set. This does not alter the Research Article’s conclusions.

To establish PCs and the numbers of core genes in each depth layer, the authors first used a graph-based clustering algorithm for the Tara Ocean Virome (TOV) data set. Later, they established the PCs with the use of a greedy incremental clustering algorithm to mirror more closely PCs defined in previous data sets, such as Pacific Ocean Viromes (POV). When they switched algorithms, they failed to update the numbers of core genes, which caused the error.

The total number of PCs and the hierarchical set of core genes for TOV stated on page 1261498-2 in the two paragraphs below the heading in the first column have been updated with the correct numbers. To clarify, the changes are as listed in the Fig. 2 legend on page 1261498-4, where 1,075,763; 710; 424; 220; and 65 have been replaced with 1,075,761; 318; 181; 81; and 24, respectively. Figure 2 has also been revised and replaced.

In the corrected Materials and Methods section, the authors have clarified that individual open reading frames (ORFs) from the Tara TOV data set were mapped to individual ORFs (not PCs) from the POV data set and only from the POV photic zone. Also, only PCs with two or more ORFs were considered bona fide and were used for subsequent analyses. Finally, the determination of surface, DCM, photic, and absolute core numbers were based on the origin of PC members, not reads mapping to PCs. This means that, if a PC had membership from all SUR samples, then it would be core SUR PC, but a PC could be core from more than just one category or all categories if it had membership from all samples of that category.

An updated version of the data set has been uploaded to the data commons Cyverse:

In the Materials and Methods, under the heading Protein Clustering (p. 1261498-6), read counts per PC for statistical analyses were mapped back to predicted ORFs with Mosiak. The correct URL is

Navigate This Article