© Wiley Publishing All Rights Reserved. Protein 3D Structures

Learning Objectives Review the basics of protein structures Know how to predict secondary structures Work with the PDB database Manipulate 3D structures Appreciate both the potentials and limitations of 3D analysis

Outline Predicting secondary structures Retrieving a PDB structure from the PDB database Guessing the 3D structure of your sequence Further applications of 3D analysis

Primary, Secondary and Tertiary Structures Proteins are made of 20 amino acids Proteins are on average 400 amino acids long Protein structure has 3 levels: The primary structure is the sequence of a protein The secondary structure is the local structure The tertiary structure is the exact position of each atom on a 3D model

Secondary Structures Helix Amino acid that twists like a spring Beta strand or extended Amino acid forms a line without twisting Random coils Amino acid with a structure neither helical nor extended Amino-acid loops are usually coils

Guessing the Secondary Structure of Your Protein Secondary structure predictions are good If your protein has enough homologues, expect 80% accuracy The most accurate secondary structure prediction server is PSIPRED

PSIPRED Output Conf = Confidence 9 is the best, 0 the worst Pred = Every amino acid is assigned a letter: C for coils E for extended or beta-strand H for helix

Predicting Other Secondary Features It is also possible to predict these accurately: Transmembrane segments Solvent accessibility Globularity Coiled/coil regions All these predictions have an expected accuracy higher than 70%

Predicting 3D Structures Predicting 3D structures from sequences only is almost impossible The only reliable way to establish the 3D structure of a protein is to make a real-world experiment in X-ray crystallography Nuclear magnetic resonance (NMR) Structures established this way are conserved in the PDB database “The PDB of my protein” is synonymous with “The structure of my protein”

Retrieving Protein Structures from PDB All PDB entries are 4-letter words! 1CRZ, 2BHL... Sometimes the chain number is added: 1CRZA, 1CRZB... To access all PDB entries, go to PDB contains 42,000 entries PDB contains the structure of 16,000 unique proteins or RNAs You can download the coordinates and display the structure

Displaying a PDB Structure You can use any of the online viewers to display the structure They will let you rotate the structure, zoom in and out, or color it PDB files themselves are not human- readable

Predicting the Structure of Your Protein The bad news: It is very hard to predict protein 3D structures The good news: Similar proteins have similar structures If your favorite protein has a homologue with a known structure... You can do homology modeling How? Start with a BLAST (more about that in the next slide)

BLASTing PDB for Structures BLAST your protein against PDB If you get a very good hit, it means PDB contains a protein similar to yours Your protein and this hit probably have the same structure

Be Careful! Sometimes only one of the domains contained in your protein has been characterized If that’s the case, the PDB will only contain this domain Always check the alignments Red line = full protein in PDB Blue line = one domain only in this entry

Structures and Sequences Highly conserved sequences are often important in the structure Make a multiple-sequence alignment to identify these important positions Highly conserved positions are either in the core or important for protein/protein interactions

3D Predictions If you want to predict the structure of your protein automatically, try the Swiss Model Swiss Model makes the BLAST for you The program does a bit of homology modeling The process delivers a new PDB entry You can access it at swissmodel.expasy.org Swiss Model gives good results for proteins having homologues in PDB

3D-BLAST Use this technique if you have a structure and you want to find other similar structures Use VAST or DALI to look for proteins having the same 3D shape as yours

3D Movements Most proteins need to move to do their job Predicting protein movement is possible using molecular dynamics Check out this site: molmolvdb.mbb.yale.edu Good molecular dynamics requires extremely powerful computers Don’t expect miracles from standard online resources

Protein Interaction Knowing “who interacts with whom” is one of the ultimate goals of predicting protein movement These predictions are called docking analyses Docking analyses are very difficult Try