A few years ago, I was fascinated by the news of the discovery of a new Carbon allotrope based on first-principle simulations by using USPEX software described here. Or particularly this article:
Zhu Q., Oganov A.R., Salvado M., Pertierra P., Lyakhov A.O. (2011). Denser than diamond:ab initio search for superdense carbon allotropes. Phys. Rev. B83, 193410, download here.
Update:
I promised to elaborate a bit more on my answer here. As I said in my comments after a few years later, this new Carbon allotrope is synthesized in the lab, which is truly fascinating how a theoretically predicted material structure, is synthesized by experimentalists (see here: https://elementy.ru/novosti_nauki/431105 and sorry for the Russian language only).
There are several other examples of discovering new materials for specific applications of batteries or fuel cells are provided here, that shows how quantum mechanical simulations combined with machine-learning methods could be used to tailor materials structure for a new application and developing new materials properties based on that.