欢迎使用CSDN-markdown编辑器

Note for Contact-assisted protein structure modeling by global optimization in CASP11

---

Three Model

- CSA global optimization method: the optimization of the energy function
- common energy functions:
    1. E_{stereo-chemistry} 
        - the stereo-chemical terms related to bond lengths, bond angles, torsion angles, and improper torsion angles according to CHARMM22
    2.  E^{vdw}_{repul}
        - the repulsive part of the van der Waals potential

Tp protocol

• predicted residue–residue contact information was provided.

• earched for protein 3D models which were diverse and had low energie
  

  $$E_{Tp}=E_{stereo-chemistry}+E_{repul}^{vdw} + E_{contact}+E_{phys}$$
  $$E_{contact}=w_{contact}\sum_i\left\{ \begin{aligned} \frac{(r_i-r_{upper})^2}{(r_i-r_{upper})^2+\delta^2},r_i>r_{upper} \\ 0 , otherwise \\ \end{aligned} \right.$$
  r_i is the distance between two Cb (Ca for glycine) atoms for each contact i;

• provided residue–residue contacts were likely to be incorrect

• not want to penalize contact violation too excessively

• ->functional form of the Lorentzian function was used for E_contact.
  

  $$E_{phys}=E_{dfa}+E_{dfire} + E_{GOAP}+E_{hbond}$$
  E_phys includes all the necessary modeling energy terms for free modeling (ab initio modeling) :

• dfa: dynamic fragment assembly terms

• dfire: dfire statistical potential term
• hbond: the hydrogen bonding term
• GOAP: the GOAP term

Total procedure

1. For initial structures -> used fold recognition models generated by nns protocol and all the server models provided from the CASP11 homepage
2. Using CSA, typically generated up to 200 structures which were clustered by the structural similarity
3. Each cluster was ranked based on the average value of Econtact and the average QA3 score
4. The best cluster was identified and the final models were selected based on E_contact and the QA3 score in the cluster.

Ts protocol

The modeling of Ts targets is closely related to the NMR structure determination using NOE distance restraints
the contact restraints provided for Ts targets were highly ambiguous —> many atom–atom pairs were assigned to a peak
CSA optimization is applied to the following energy function:

$$E_{Ts}=E_{stereo-chemistry}+E_{repul}^{vdw} + E_{chiral}+E_{CMAP}+E_{NOE}$$

• Estereo-chemistry: constrains bond lengths, bond angles, torsion angles and improper torsion angles, and taken from the CHARMM22 force field
• E_vdw^repul: the repulsive part of the van der Waals potential
• E_chiral: keep the chirality of the amino acid residue in the L form
  
  • Since the driving force for the structure modeling of Ts targets is hydrogen–hydrogen distance restraints, without Echiral, the mirror image of a given structure (either in L or D form) will be as stable as the given structure
• ECMAP: used to provide a cross-term correction for two adjacent torsion angles
  
  • which is a standard stereo-chemical term in the CHARMM force field
• NOE: for a given set of ambiguous restraints
  $$E_{NOE}=\sum_i\left\{ \begin{aligned} k_i(R_i-d_i^{upper})^2,R_i>d_i^{upper} \\ 0 , d_i^{lower}<R_i<d_i^{upper} \\ k_i(R_i-d_i^{lower})^2,R_i<d_i^{lower}\\ \end{aligned} \right.$$
  $$R_i=(\sum_jd_i^{-6})^{\frac{-1}{6}}$$
  
  • the sum over j is taken for all possible distance pairs provided for the ith restraint.
    $$k_i=10^3/d_i^{upper}kcal mol^{-1}A^{-1}$$
    
    • the conventional usage of the effective distance Ri suits our purpose of solving the two-tier optimization problem, where the variation of an assigned peak to an atom–atom pair is possible
• Depending on the initial protein structure, the minimization of E_Ts will provide various peak assignment possibilities (i.e., the identity of the atom–atom pair contributing the most in 1/di^6 varies)

Total Procedure

1. For initial structures, we used the same initial models as used in the Tp protocol
   
   • generated up to 700 structures which were subsequently
     reminimized using the LBFGS local minimization method on the new energy function:
     $$E_{Ts}'=E_{Ts}+E_{electro-static} + E_{gbsa}$$
     
     • the second and the third terms respectively correspond to the CHARMM22 electro-static potential and gbsa solvation energy
2. Final models were selected based on the values of E_ts and the Ramachandran-favored score calculated by the MolProbity program

Tc protocol

total 24 Tc target:17 (Ts,Tc) + 7 (Tc)
- For 17 (Ts,Tc):

$$E_{Tc}^1=E_{Ts}+E_{contact}$$
- For 7 (Tc):
$$E_{Tc}^2=E_{Tp}$$
- Note: arguments is different from E_ts and E_tp.