An Introduction to Bioinformatics

An Introduction to Bioinformatics

Protein Structure Prediction

Aims

• Understand the use of algorithms

• Recognize different approaches

• Understand the limitations

Objectives

• Predict occurrence of aspects of structure

• To select appropriate tools

Introduction

• Structure has several levels– 1 primary– 2 secondary– 3 tertiary– 4 quaternary

1 primary

• Amino acid sequence

NH2-MRLSWYDPDFQARLTRSNSKCQGQLEV YLKDGWHMVC SQSWGRSSKQWEDPSQASKVCQRLNCGVPLSLGPFLVTYTPQSSIICYGQLGSFSNCSHSRNDMCHSLGLTCLE-COOH

2 secondary

• Localized organisation -helices and -sheets

3 tertiary

Three-dimensional organisation

4 quaternary

Multi protein assembly

The problem…..

• The best way is by X-ray crystallography or NMR etc…

• Structure databases only hold about 10,000 + structures

• Therefore devise programs to deduce structural solutions

• Complex!

Secondary Structure prediction

• Signal peptides

• Intracellular targeting

•Trans-membrane -helices

• -helices and -sheets

•Super-secondary structure (motifs)

Signal peptides

• Short N-terminal amino acid sequences

• Direct to membrane

• Cleaved after translocation

• SignalP – Nobel Prize 1999 Günter Blobel

SignalP predicts signal peptide cleavage sites

Only first 50-70

Using neural networks

Is the sequence a signal peptide?

# Measure Position Value Cutoff Conclusion max. C 25 0.910 0.37 YES max. Y 25 0.861 0.34 YES max. S 12 0.960 0.88 YES mean S 1-24 0.892 0.48 YES# Most likely cleavage site between pos. 24 and 25: SRA-LE

Intracellular targeting

• TargetP

• Predict subcellular location of eukaryotic protein

• Presequences – Chloroplasts– Mitochondria– signal peptide

Transmembrane Domains

• Lots of programs

• TMHMM -helices– hydrophobic – helix topology– R or K +ve charge cytoplasmic

side– Hidden Markov Modelling

Paste as FASTA file

e.g Serotonin Receptor

Predicts the transmembrane domains and orientation

-helices and -sheets

• GOR algorithim• Assigns each residue to one conformational state of -helix, extended chain, reverse turn or coil• 64.4% accurate• Many other sites

• most use multiple alignments

-helices and -sheets

10 20 30 40 50 60 70 | | | | | | |MKFSWRTALLWSLPLLVVGFFFWQGSFGGADANLGSNTANTRMTYGRFLEYVDAGRITSVDLYENGRTAIcccceeeeeecccceeeeeeeeccccccccccccccccccchhhhcceeeeccccceeeeeeccccceeeVQVSDPEVDRTLRSRVDLPTNAPELIARLRDSNIRLDSHPVRNNGMVWGFVGNLIFPVLLIASLFFLFRReeccccccchhhhccccccccchhhhhhhhhccccccccceecccceeeeecccccchhhhhhhhheeecSSNMPGGPGQAMNFGKSKARFQMDAKTGVMFDDVAGIDEAKEELQEVVTFLKQPERFTAVGAKIPKGVLLcccccccccchhhhcchhhhhhhhccceeeecchhhhhhhhhhhhhhhhhhcccchhhhhcccccceeeeVGPPGTGKTLLAKAIAGEAGVPFFSISGSEFVEMFVGVGASRVRDLFKKAKENAPCLIFIDEIDAVGRQRecccccchhhhhhhhhcccccceeecccccceeeeeecccchhhhhhhhhcccccceeeecchhhhccccGAGIGGGNDEREQTLNQLLTEMDGFEGNTGIIIIAATNRPDVLDSALMRPGRFDRQVMVDAPDYSGRKEIccccccccchhhhhhhhhhhhhcccccccceeeeeeccccchhhhhhccccccceeeeecccccccchhhLEVHARNKKLAPEVSIDSIARRTPGFSGADLANLLNEAAILTARRRKSAITLLEIDDAVDRVVAGMEGTPhhhhhhhhccccccchhhhccccccccchhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhheeeccccccLVDSKSKRLIAYHEVGHAIVGTLLKDHDPVQKVTLIPRGQAQGLTWFTPNEEQGLTTKAQLMARIAGAMGcccccccchhhhhcccceeeeeecccccccceeeecccccccceeccccccccchhhhhhhhhhhhhhhhGRAAEEEVFGDDEVTTGAGGDLQQVTEMARQMVTRFGMSNLGPISLESSGGEVFLGGGLMNRSEYSEEVAhhhhhhhcccccceeeccccchhhhhhhhhhhhhhhccccccccccccccceeeecccccccccchhhhhTRIDAQVRQLAEQGHQMARKIVQEQREVVDRLVDLLIEKETIDGEEFRQIVAEYAEVPVKEQLIPQLhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhcccccchhhhhhhhhhcccccccccccc

Super-secondary Structure• Secondary structure elements

combined into specific geometric arrangements known as motifs

Beta corner

Super-secondary Structure

Several programs/websites for specific domains e.g.

• PAIRCOIL and MULTICOIL - detect coiled-coiled regions– regions separating domains

• TRESPASSER - detects Leucine Zippers– Leu-X6-Leu-X6-Leu-X6-Leu protein interaction domain

• NPS@nalysis Helix-Turn-Helix– Protein interaction/DNA binding

Integrated stucture prediction

• One stop shop!• Predict Protein at EBI

– secondary structure

– solvent accessibility globular regions

– transmembrane helices coiled-coil regions

– a multiple sequence alignment ProSite sequence motifs

– low-complexity retions

– ProDom domain assignments

Tertiary Structure Prediction

• Homology modelling

• Fold recognition

• Threading

• Model building

Protein sequence(primary structure)

Database searchingfor homologues

Homologue ofknown structure

No homologue ofknown structure

Comparativemodelling

3D-structure

Fold prediction,ab initio methods etc.

Homology Modelling

• Method of choice following BLAST search

• SWISSModel is agood WWWInterface

URL: http://www.expasy.ch/swissmod/SWISS-MODEL.html

• Requires at least one sequence of known 3D-structure with significant similarity to the target sequence.

• Compare the target sequence with database - FastA and BLAST.

• Sequences with a FastA score 10.0 standard deviations above the mean of the random scores or a P(N) lower than 10-5 (BLAST) considered for the model building

• Restrict to those which share at least 30% residue identity

Homology Modelling

Homology Modelling

• Framework construction– compare atom positions - Cs

• Build non-conserved loops

• Complete backbone - add other atoms

• Add side chains

• Refine

Insulin like gene from C.elegansRed = InsulinBlue = ILGF1

What if I have no homologue?

Ab initio methods - Threading

• Sequence of unknown structure

• Thread through a through a sequence of known structure

• Move query sequence through residue by resudue and compare computationally

– include thermodynamic criteria, solvent accessibility, secondary structure information

• Computing intensive

http://www.cs.bgu.ac.il/~bioinbgu/form.html