It's quite clear to me that a great deal of medicine innovations can come from computer simulation. If a computer can fully simulate a medium composed by human cells, water, nutrients and other molecules, and also offenders such as viruses and bacterias, we should be able to test for new medicines and treatments in a much faster, safer and fair way than current methods. There would be less need for testing on animals and human voluntaries, because computer simulations can quickly filter out many faulty alternatives. Hence, the rate of success on the initial real-world tests would be greatly improved.
The computer required to do this must be very powerful even for current standards. It needs to simulate molecular or perhaps atomic interactions in three or "four" space dimensions, plus the time, taking account pressure, temperature and momentum. A human cell should contain some 100 trillion atoms. If each atom can be modeled with about 1024 bytes of data (1kb), then a full human cell would require about 90 petabytes to be simulated. While this sounds prohibitively huge, there are predictions saying that personal devices will have this capacity by 2020. So, why can't we start development of systems and algorithms right now, so that when the metal is ready, we just load the hacks and run them?