Hey guys,
I've been trying to develop a way to calculate an accurate metastability for a kinetic intermediate state in my protein system, and I came up with I think is an interesting method to do so, but I'm not sure about the soundness of it, so I'm putting it out there. Feedback would be greatly appreciated.
Basically the idea is I have a protein with multiple states (native, unfolded, etc.) and I want to isolate all the microstates in the MSM that are consistent with that one state. For the state I want, I would take a representative structure (i.e. the crystal structure if I wanted the native state) and order all the microstates by some order parameter relative to that representative structure (e.g. native contacts). I would then repeatedly build macrostate models by successively lumping those states into their own macrostate, and plotting the self-transition probability against the number of states I lumped to make the macrostate. The local maximum of the plot is what I would pick to be the most representative metastability.
My reason for this is because as soon as I lump an "illegal" microstate into my ensemble, the fast intrastate transitions involving that illegal microstate will be treated as a fast interstate transition in my desired ensemble, which will artificially decrease the metastability. I did a benchmark study on the native state of CheY using native contacts as the metric (attached image), and found that the local maximum at n = 32 recapitulated the lumping results of PCCA.
An extension of this idea is that one may be able to judge the quality of order parameters by qualitatively looking at the features of the graph. For a two state system U <-> N for example, a good order parameter would give a single local maximum, i.e. where the U to N boundary is, and for a 3-state system U <-> I <-> N, you would see two local maxima, etc., whereas a bad order parameter will give you many more local maxima that don't make any sense.
Hey guys, I've been trying to develop a way to calculate an accurate metastability for a kinetic intermediate state in my protein system, and I came up with I think is an interesting method to do so, but I'm not sure about the soundness of it, so I'm putting it out there. Feedback would be greatly appreciated.
Basically the idea is I have a protein with multiple states (native, unfolded, etc.) and I want to isolate all the microstates in the MSM that are consistent with that one state. For the state I want, I would take a representative structure (i.e. the crystal structure if I wanted the native state) and order all the microstates by some order parameter relative to that representative structure (e.g. native contacts). I would then repeatedly build macrostate models by successively lumping those states into their own macrostate, and plotting the self-transition probability against the number of states I lumped to make the macrostate. The local maximum of the plot is what I would pick to be the most representative metastability.
My reason for this is because as soon as I lump an "illegal" microstate into my ensemble, the fast intrastate transitions involving that illegal microstate will be treated as a fast interstate transition in my desired ensemble, which will artificially decrease the metastability. I did a benchmark study on the native state of CheY using native contacts as the metric (attached image), and found that the local maximum at n = 32 recapitulated the lumping results of PCCA.
An extension of this idea is that one may be able to judge the quality of order parameters by qualitatively looking at the features of the graph. For a two state system U <-> N for example, a good order parameter would give a single local maximum, i.e. where the U to N boundary is, and for a 3-state system U <-> I <-> N, you would see two local maxima, etc., whereas a bad order parameter will give you many more local maxima that don't make any sense.
Curious to see what you guys think. Thanks! Jade