antibody H3 modelling protocol for nanobody gives “[ ERROR ] Inaccurate G!” messages

[Anonymous]

Hello all,

I am trying to model the H3 loop of a nanobody (VHH). The protocol runs fine and completes successfully but it gives a lot of the following error messages:

……

core.optimization.LineMinimizer: [ ERROR ] Inaccurate G! step= 3.8147e-06 Deriv= -0.121688 Finite Diff= 0.149698

core.optimization.LineMinimizer: [ ERROR ] Inaccurate G! step= 1.90735e-06 Deriv= -0.0181975 Finite Diff= 0.0799868

core.optimization.LineMinimizer: [ ERROR ] Inaccurate G! step= 3.8147e-06 Deriv= -0.0763095 Finite Diff= -0.0049378

core.optimization.LineMinimizer: [ ERROR ] Inaccurate G! step= 3.8147e-06 Deriv= -0.0891257 Finite Diff= 0.0424726

core.optimization.LineMinimizer: [ ERROR ] Inaccurate G! step= 3.8147e-06 Deriv= -0.0358391 Finite Diff= 0.0313933

core.optimization.LineMinimizer: [ ERROR ] Inaccurate G! step= 3.8147e-06 Deriv= -0.117232 Finite Diff= -0.0115376

……

I could not find much on line. What are they? How shall I interprete them? Any help would be greatly apreaciated

Cheers,

Alfredo

 

[Anonymous]

This error sounds way scarier than it is.

What it means is that the minimizer has run into an algorithmic problem. It moves down the conformational energy landscape by computing derivatives and then taking steps in a direction “down” the gradient.  The error means that it took such a step and the energy unexpectedly went up.

Generally speaking this means there is some combination of strange local microstructures in your protein that are triggering edge case bugs in some term of the energy function.

The main effect of the bug is simply that a lot of compute time gets wasted in the minimizer.  The scores for structures from such runs are accurate; whether the Monte Carlo trajectory was efficient is a different question (probably not).

[Anonymous]

Thank you very much! I had the feeling it was more like a warning than a proper error, but def worth checking. Thanks for clarifying

[Author]

Posts