One of the great frustrations with scholarly publishing today is that primary research data is an afterthought. At best, you can download a dataset (probably after searching five minutes for the link). At worst, there is no dataset to be found and you have to resort to bugging the author for more info. And it is not just datasets that are sorely abused in publishing - the in-line tables, figures, and text are ripe areas for data play with today’s technologies, but are ignored.
The second goal of PeerJ is to make all content reusable. Since the paper, and specifically just one version of the paper, is only one facet of the research, this means that PeerJ will provide a venue to easily make the underlying data available. Moreover, authors will have the ability to seamlessly add new datasets related to the paper. After all, research is a living, breathing process that is never set in stone. The ability to consume that data in multiple formats is something that PeerJ is also striving toward.