Professor Christian Keysers discusses the hypothesis linking autism, mirror neurons, and synaptic proteins.
Now one of the interesting links of autism to the mirror [neuron] system is this idea that I presented before that maybe we train up our mirror [neuron] system by observing our own actions. To do this, what you need is a brain that can optimally learn to associate things that happened together at the synaptic level. Now there is more and more evidence that in certain cases of autism at least show this function in synaptic proteins, and these synaptic proteins are essential in learning. So it could be that part of why autism may have problems in the mirror [neuron] system when social cognition actually relies on much more fundamental dysfunctions at the synaptic level that make it harder for them to understand what belongs together or what happens together. And these kinds of general dysfunctions would have big consequences in the social domain, because you would find it a lot more difficult to link the sight of an action with the motor program for an action.