Chemistry Reference
In-Depth Information
Knowing that the desired accuracy is not present in some models should spur
more innovation and demand for more enhanced approaches and fundamental
science. This could include data from experimentalist, human's expert eye and
tailored scoring functions.
Regarding evaluation and standards, we should hope for some sort of 'good
computing' certification for software that meets certain community expectations.
For the future, a lot will depend less on the hardware and more on the software
(providing error estimates and some indication of confidence).
It would be commendable for software to be able to estimate additional
information such as diverse physical properties; provide available experimental
results; experimental and predicted values of related compounds; required
precision for applications; clustering of like compounds by docking programs; list
of related compounds and sample availability; chemical, physical, biological
properties and synthetic procedures.
We need to also develop software that are goal-oriented,
i.e.
that could help us
make decisions. 1) Tell us if a model is able to yield accurate protein/drug
interaction energies. 2) Anticipate whether the correlation will be poor. 3) Driven
by few point of high leverage. 4) Inform whether number of descriptors are too
large. 5) Predict whether we can obtain the required accuracy and precision. 6)
Use conglomerate of approaches/expert systems/decision trees assembled towards
our goals. 7) Include table of look-up pre-calculated values. 8) Request for expert
intervention. 9) Inform when key data is not available. 10) Recommendations for
further experiments. 11) Inform if needed models are already available 12) Yield
precise calculations of physical properties from first principles. 13) Inform us of
potentially unobtainable goals. 14) Show the route for more tractable goals.
We should not have over-expectations of what our methods can do. We should not
leave to experiment and interpretation the evaluation of our success. As computer
speed increases, we will be able to run larger, more complex quantum mechanical
systems and perform simulations with longer/more realistic comparisons with
experiment/obtain more data and better statistics in order to improve our scientific
quests.
