Homology Modeling in Biology and Medicine - LSIR - EPFL

The goal of Homology ModelingUse homologous sequences to construct amodel of 3D structureAnalyze relationships between DNAsequence and 3D structure of proteinsKnow protein’s 3D structure to understandinteractions with other moleculesCreate computer-aided drug design,mutagenesis and protein engineering01/12/2004 3

Homologous sequencesGene A in DNA sequenceDuplication ofthe gene AGene AParalogyGene AOrthologyMutationinvolvingSpeciationXenologyGene AduplicatedCopytransferred tothe specie froman otherorganismGene A indifferentorganismGene A Gene A’Gene A inspecies 1Gene A inspecies 201/12/2004 4

How determine the protein structure?By experimentationX-RayNMR (nuclear magnetic resonance spectroscopy)Today, Sequence Analysis have explodedWe have the dataWe need to construct 3D modelsThe idea!Use similar structure to identify constraints and build foldcorrespondingHomology Modeling01/12/2004 7

Where find the data?Protein Data Bank (PDB)http://www.rcsb.org/pdb/> 10,000 structures of proteinsText file contain: coordinates for each heavy (nonhydrogen)atom from the first residue to the lastATOM 1 N SER A 2 29.089 9.397 51.904 1.00 81.75ATOM 2 CA SER A 2 27.883 10.162 52.185 1.00 79.71ATOM 3 C SER A 2 26.659 9.634 51.463 1.00 82.64ATOM 4 O SER A 2 26.718 8.686 50.686 1.00 81.02ATOM 5 CB SER A 2 28.039 11.660 51.932 1.00 75.59ATOM 6 OG SER A 2 27.582 12.038 50.639 1.00 43.28-------ATOM 1737 CD1 ILE A 229 39.535 21.584 52.346 1.00 41.62TER 1738 ILE A 22901/12/2004 8

The way to visualize the protein It is impossible to read this text file without the help ofgraphic viewers such as RASMOL http://www.bernstein-plus-sons.com/software/rasmol Different way to visualize:All-atom model, in ball-andstickrepresentationCα TraceColoring: by structureSpace-filling model01/12/2004 9

Structure and homologous sequencesWith at least 30% identity between two sequences, adefinite correlation exists between sequence andstructureIn particular, homologous sequences show verysimilar structures, with strong conservation insecondary structural elementsSome folds are preferred by vastly different sequencesto conserve the structure of the active siteOn the other hand, some proteins adoptvery similar structures, with no obvioussequence similarity01/12/2004 10

Why homology modeling?Other way to construct 3D modelPrediction method Ab initio ThreadingBut :Expansive in time and in calculationThe solution of Homology ModelingFrom 3D structure for each protein familyConstruct model from this known structure template structure01/12/2004 11

Before building a model…Elements of sequence analysis, essential forbuilding a molecular model, will be consideredMultiple sequence alignmentAlignment checksProtein domain,…01/12/2004 12

Some problems have to be solved Homologous sequences are identified by using databasesearch methods (BLAST) To build a model, we require the alignment of completeprotein sequences, collected from database searches Identical residues must be lined up The rest should be arranged, based on observed substitution in protein families chemical similarity charge similarity Where none of the 19 residues is suitable, the alignmentsimply skips that position a ‘gap’ (insertion/deletionregions) CLUSTALW/CLUSTALX, MAXHOM, MALIGN (MSA method)01/12/2004 13

After alignment, check the resultThe function of a protein depends on thelocalization in space of a few key residuesSome residues are critical for the stability of theprotein fold or for the formation of functionalquaternary structuresConserve all residues usually indicate someconserved structural or functional role, especiallyburied charges01/12/2004 14

The most important step: find ahomologous structureThe criteria:Alignment Score and E value (discarded: low scoresand high values (> 0.005) )Domain coverage (at least 60% of the domain)Gaps (the fewer the gaps, the better the structuralmodel)For small proteins, specific search (disulfide bond)No structure found: prediction method used(second and tertiary structure prediction method)01/12/2004 17

Selection of template sequenceSingle structural homologue one uniquechoice for template selectionSeveral equally structural homologues areidentified how many and which one(s) shouldwe choose?Improve one template in viewingsimple phylogenetic tree (show the most similarstructure)Completeness of structural information (viewingPDB information by RASMOL and verify thecompleteness of the structure)X-ray and NMR entries01/12/2004 18

One or many templates?When we have selected many templateswith same quality and similarityCompare 3D structure to check the uniqueinformation each templates providesStructure alignment of Cα atomsIf 2 templates are very close, keep only oneKeep templates that provide new information01/12/2004 19

Align the template and the targetsequencesIn case of homology (>40%), the alignment isconstant and every method is availableIn the other cases, the use of multiple sequenceimprove the qualitySome checks are needed to increase the satisfactionof the modelResidue conservation checks (pattern and function)Visual inspection of indel regions (RASMOL)01/12/2004 20

And finally… Build the modelIt is the moment to use a programIn input: target, template sequence and theiralignmentIn output: the 3D structure responding ofthe constraints01/12/2004 22

Which program to choose?WHATIF (1990)SWISSMODEL (1993)MODELLER (1994)ICM (1994)CPH Models (1997)SDSC1 (2000) 3D-JIGSAW (2001)01/12/2004 23

Advantages of Modeller Implements comparative protein structuremodelling by satisfaction of spatial restraints (2,3)Can perform many additional tasks including de novo modelling of loops in protein structuresOptimize various models of protein structure with respectto a flexibly defined objective functionPerform multiple alignment of protein sequences and/orstructuresSearch sequence in databasesCompare protein structures01/12/2004 25

Optimization with iteration01/12/2004 26

To concludeProtein structure determine functionsImportance to know protein structure forapplication in biology and medecineHomology modeling :From a known structure in protein familyBuild a model of homologous sequence01/12/2004 27