Makes sense now, thank you!

Mariana


From: Min-yi Shen <minyishen@gmail.com>
Sent: Friday, June 19, 2020 5:27:08 PM
To: Modeller Caretaker
Cc: Mariana GONZALEZ MEDINA; modeller_usage@salilab.org
Subject: Re: [modeller_usage] z-DOPE score
 
Original DOPE author here. The Z-dope is effectively an estimated z-score of your protein model based on the protein composition and size, as the raw DOPE scores are size and protein dependent. Z-dope used a linear model to estimate the mean and sd of a protein’s DOPE score based on their sizes and compositions. It was trained by millions of models from ModBase. -1.5 means your model’s DOPE score is -1.5 standard deviation lower than the average... usually a good indicator of good native overlap.

I can’t believe I can still recall all these... 

On Thu, Jun 18, 2020 at 11:59 Modeller Caretaker <modeller-care@salilab.org> wrote:
On 6/18/20 9:48 AM, Mariana GONZALEZ MEDINA wrote:
> I have a question regarding the zDOPE score. I have read some paper that
> state this: "The Z-DOPE scores ranged from −1.63 to −1.85, where a score
> of less than −1 indicates a “reliable” model (i.e., 80% of its Cα atoms
> are within 3.5 Å of their correct positions)"
>
> I am wondering how is it that a score of less than -1 indicates that 80%
> of its Cα atoms are within 3.5 Å of their correct positions?

I would say "implies" rather than "indicates". DOPE is a model
evaluation function. Of course we don't know the true structure so we
can only guess with some degree of confidence (the more negative the
score, the more confident I would be) whether a good DOPE score really
means a correct model. Benchmarking does show that this is typically the
case though; see the DOPE paper.

        Ben Webb, Modeller Caretaker
--
modeller-care@salilab.org             https://salilab.org/modeller/
Modeller mail list: https://salilab.org/mailman/listinfo/modeller_usage
_______________________________________________
modeller_usage mailing list
modeller_usage@salilab.org
https://salilab.org/mailman/listinfo/modeller_usage