XLs restraint: stat file reports only the likelihood as score

[iecheverria]

When adding all the restraint scores reported in the stat file (i.e. excluded volume, connectivity, xls) they do not add to the total score reported in the stat file. For example:

Total_Score != ExcludedVolumeSphere_Score + CrossLinkingMassSpectrometryRestraint_Data_Score+ Sum(Connectity_Score)

To recover the total score is also necessary to add the CrossLinkingMassSpectrometryRestraint_PriorPsi_Score term.

Total_Score = ExcludedVolumeSphere_Score + CrossLinkingMassSpectrometryRestraint_Data_Score+ Sum(Connectity_Score) + CrossLinkingMassSpectrometryRestraint_PriorPsi_Score term.

This can be misleading when evaluating the contribution of individual scores since the PriorPsi_Score term can be fairly large (in absolute value) when Psi is close to zero (very low number of false positives).

Is there a reason not to report the log likelihood and prior/priors as individual terms, but making the score the -log(likelihood*prior) in the stat file? In such way, it would be consistent with the score used for Monte Carlo steps.

[Pellarin]

This follows the structure of the posterior score, where likelihoods and priors are separated terms. In the example that you give:

posterior_score = XL_likelihood_score + Excluded_Volume_Prior + Connectivity_Prior + XL_prior

In such way, it would be consistent with the score used for Monte Carlo steps.

I don't get this point. Can you explain?

Riccardo Pellarin, PhD

Institut Pasteur CNRS UMR 3528 25, rue du Docteur Roux 75724 Paris Cedex 15, France [email protected] +33 (0)1 44 38 93 63

On Fri, Feb 9, 2018 at 2:37 AM, Ignacia Echeverria <[email protected]

wrote:

When adding all the restraint scores reported in the stat file (i.e. excluded volume, connectivity, xls) they do not add to the total score reported in the stat file. For example:

Total_Score != ExcludedVolumeSphere_Score + CrossLinkingMassSpectrometryRestraint_Data_Score+ Sum(Connectity_Score)

To recover the total score is also necessary to add the CrossLinkingMassSpectrometryRestraint_PriorPsi_Score term.

Total_Score = ExcludedVolumeSphere_Score + CrossLinkingMassSpectrometryRestraint_Data_Score+ Sum(Connectity_Score) + CrossLinkingMassSpectrometryRestraint_PriorPsi_Score term.

This can be misleading when evaluating the contribution of individual scores since the PriorPsi_Score term can be fairly large (in absolute value) when Psi is close to zero (very low number of false positives).

Is there a reason not to report the log likelihood and prior/priors as individual terms, but making the score the -log(likelihood*prior) in the stat file? In such way, it would be consistent with the score used for Monte Carlo steps.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/salilab/imp/issues/1000, or mute the thread https://github.com/notifications/unsubscribe-auth/ACXOkPEYwZO8pS6zedVrB_pNjdXrGVkhks5tS6FugaJpZM4R_Suj .