There are many promising research directions, which are however grossly underexplored, either because there is a low chance it might lead to a positive result, or because it cannot be patented even if it's successful. Examples in the latter category include off-label use of established procedures, or of already known drugs. Conversely, the majority of money and energy is being spent on highly predictable research and which can result in a patent; almost always the downside is low impact. So as you can see, profit optimisation is not correlated with impact optimisation.