ASIC's do not have only advantages they are less flexible.
In the WCG particular case, at one side you should have an ASIC supporting for example VINA and at the other side the science project must be rewritten for fitting with the new environment.
Based on the computation complexity including floating point calculation, I doubt that the efforts and the investment required for supporting such an approach would show a positive ROI.
Additionally, you should not forget that sciences and algorithms are updated from time to time. This flexibility is necessary.
If you consider how today it is difficult for WCG scientists to support GPU, you can completely forget an ASIC-based approach.
By the way, a RPi is a "classical" computer and not an ASIC.
Cheers,
Yves
---
PS: WCG sciences are a little bit more complex than coin mining.

This would include at least a suspicion of a "for-profit" factor in this project. A very important reason I and probably a lot of others contribute here is exactly because of the not-for-profit idea.

Basically just donating the use of existing hardware is what WCG asks from the members, and that is just fine for me.

In the specific case of WCG project, the integrity of the results is very important for comparing results and enabling research decisions.
As soon as the science software would be modified even for optimisation purpose, the integrity of the results could not be ensured any more, making the complete work useless.
Cheers,
Yves

It's a great idea, maybe not as specialised as an ASIC but some sort of hardware that is optimised for the typical usage. Like if autodock spends most of its time doing one type of calculations then have hardware that is better at that specific part to speed it up.