Team:UANL Mty-Mexico/Modeling/3Dmodel/structure prediction

Structure Prediction

Theoretical background

Basically, our modelling approach to predict the tertiary structures of the proteins related to our system relies on the Anfinsen's thermodynamic hypothesis (Anfinsen 1973); our goal is to obtain models which satisfy the condition of the minimum free-energy states. Although every modelled structure will be validated by means of its energetic state, the sampling in the conformational space will differ according to the type of protein modelled and the type of input information available.

Methodology

Our modelling approach is divided into the two sub-modules which are listed below:

• rhMT

• Assemblies

• rhMT-L2

• OmpA-L2

• Mocr-[rhMT]x

Where rhMT stands for the recombinant human metallothionein, whereas the group “assemblies” includes six fusion proteins: 1.rhMT fused to the E. coli ribosomal protein L2, 2.Outer membrane protein A fused to L2, and 3-6.Mocr tag protein fused to rhMT (subscript “x” refers to the number of copies as integers 1-4).

We used algorithms of the Rosetta3.4 package (Srivastan et al. 2009) such as AbinitioRelax (ab initio modelling), Comparative modelling, and Clustering application (RMSD calculations and sorting by free-energy). The ab initio approach was used to predict the tertiary structure of rhMT by using a template native structure of an homologous metallothionein (PDB: 4mt2) found by Jpred3 (Cole et al. 2008), in order to obtain an accurate model, we used 3mer and 9mer rotamer library fragments generated by Robetta server. Radius gyration, contact-order and sheet filters as well as secondary structure prediction files were used in order to avoid some noise in a fail-model context. We used default loop and helix reweight factors as well as ab initio cycles. The key flags used in our model with the AbinitioRelax algorithm are showed below.

AbinitioRelax.linuxgccrelease \

-database /rosetta_database/ \

-abinitio:relax \

-use_filters true \

-abinitio::increase_cycles 10 \

-abinitio::rg_reweight 0.5 \

-abinitio::rsd_wt_loop 0.5 \

-abinitio::rsd_wt_helix 0.5 \

-relax::fast \

-in:file:fasta rhMT.fas

-in:file:frag3 rhMT3_05.200_v1_3

-in:file:frag9 rhMT9_05.200_v1_3

-in:file:native 4mt2.pdb

-psipred_ss2 rhMT.psipred_ss2

-out:file:silent rhMT_silent.out

-nstruct 20000

rhMT.fas> [MKFGSMGKAAAACSCATGGSCTCTGSCKCKECKCNSCKKAAAACCSCCPMSCAKCAQGCVCKGASEKCSCCKKAAAAGS]

Clustering application was used to obtain clusters by calculating RMSD between all the structures and joining them to their corresponding neighbours, the resulting clustered models were sorted by their free-energy in order to obtain the model with the lowest value —which is intended to satisfy the Anfinsen's thermodynamic hypothesis— thus, the most accurate approximation to the native state (it's sensitivity depends strongly on the size of the explored portion of the conformational space).

Results

As molecular simulations are time-consuming (computationally), we only have partial results for rhMT ab initio modelling (partial model generated by using 5,000 simulated structures as input for the Cluster application). Unfortunately, this model is not accurate enough describe the native structure of rhMT.

Partial 3D model for rhMT structure obtained by clustering the 5,000 generated models. This structure represents the lowest free-energy state obtained in the sampled conformational space..