So what is known about the genetics of speed? In the cases of horses and dogs, quite a lot. For humans, maybe not as much, since we don't deliberately breed people for speed. Still, there is quite a lot that is known. One big factor is limb conformation, specifically the lever arm of the attached muscles. This has to do with the relative lengths of the limbs and where exactly the muscles are attached. These things are mostly controlled by genetics and completely immune to training. The strength and composition of the muscles involved is also important, and in particular the types of muscle fibers composing the muscles. Muscle fiber type is specified by genetics, while training has the ability to strengthen muscles, but can't change the type. Recruitment, the degree and ease with which fibers are neurally activated is partially genetic but can be increased by training. Muscle training essentially works by increasing the size of individual fibers and improving their recruitment.
It's also known that good to excellent sprinters have much higher proportions of fast twitch fibers than average persons, and that even their type of fast twitch fiber is special. Studies of elite sprinters (and jumpers) have also shown that they were "born fast", or at least that they were always the fastest kid on the block and all showed exceptional speed at their first exposure to competition and training. Elite sprinters, as I've mentioned elsewhere, are almost all American or Caribbean of West African or European and West African descent. Independently, this population is known to have relatively high proportions of fast twitch muscle fibers.
Some of these facts are captured in some homely expressions long known to coaches. On the limits of training: "I can make you faster, but I can't make you fast." On the athletic benefits of speed: "Speed never has a bad day." On areas where training doesn't help: "You can't teach height."
In conclusion, for one crucial athletic ability, speed, genetics is the essential substrate of exceptional capability.