Protein Secondary Structure Prediction Dong Xu Computer Science Department 271C Life Sciences Center 1201 East Rollins Road University of Missouri-Columbia Columbia, MO (O)

Outline  What is Secondary Structure  Introduction to Secondary Structure Prediction  Chou-Fasman Method  Nearest Neighbor Method  Neural Network Method

Structures in Protein Language: Letters  Words  Sentences Protein: Residues  Secondary Structure  Tertiary Structure

 helix  Single protein chain (local)  Shape maintained by intramolecular H bonding between -C=O and H-N-

 sheet  Several protein chains  Shape maintained by intramolecular H bonding between chains  Non-local on protein sequence

 -sheet (parallel, anti-parallel)

Classification of secondary structure l Defining features å Dihedral angles å Hydrogen bonds å Geometry l Assigned manually by experimentalists l Automatic å DSSP (Kabsch & Sander,1983) å STRIDE (Frishman & Argos, 1995) å Continuum (Andersen et al.)

Classification l Eight states from DSSP  H:  helix  G: 3 10 helix  I:  -helix  E:  strand  B: bridge  T:  turn  S: bend  C: coil l CASP Standard  H = (H, G, I), E = (E, B), C = (C, T, S) E H < S R H < S N < K ! C I E -cd 58 89B L E -cd 59 90B V E -cd 60 91B G E -cd 61 92B 0

Dihedral angles

Ramachandran plot (alpha)

Ramachandran plot (beta)

Outline  What is Secondary Structure  Introduction to Secondary Structure Prediction  Chou-Fasman Method  Nearest Neighbor Method  Neural Network Method

What is secondary structure prediction? l Given a protein sequence (primary structure) HWIATGQLIREAYEDYSS GHWIATRGQLIREAYEDYRHFSSECPFIP l Predict its secondary structure content (C=Coils H=Alpha Helix E=Beta Strands) EEEEEHHHHHHHHHHHHH CEEEEECHHHHHHHHHHHCCCHHCCCCCC

Why secondary structure prediction? o An easier problem than 3D structure prediction (more than 40 years of history). o Accurate secondary structure prediction can be an important information for the tertiary structure prediction o Protein function prediction o Protein classification o Predicting structural change

Prediction methods o Statistical method oChou-Fasman method, GOR I-IV oNearest neighbors oNNSSP, SSPAL oNeural network oPHD, Psi-Pred, J-Pred oSupport vector machine (SVM) oHMM

Accuracy measure l Three-state prediction accuracy: Q 3 correctly predicted residues number of residues l A prediction of all loop: Q 3 ~ 40% l Correlation coefficients

Improvement of accuracy 1974 Chou & Fasman~50-53% 1978 Garnier63% 1987 Zvelebil66% 1988 Qian & Sejnowski64.3% 1993 Rost & Sander % 1997 Frishman & Argos<75% 1999 Cuff & Barton72.9% 1999 Jones76.5% 2000 Petersen et al.77.9%

Prediction accuracy (EVA)

How far can we go?  Currently ~76%  1/5 of proteins with more than 100 homologs  >80%  Assignment is ambiguous (5-15%).  non-unique protein structures, H-bond cutoff, etc.  Some segments can have multiple structure types.  Different secondary structures between homologues (~12%). Prediction limit  88%.  Non-locality.

Assumptions o The entire information for forming secondary structure is contained in the primary sequence. o Side groups of residues will determine structure. o Examining windows of residues is sufficient to predict structure. o Basis for window size selection:   -helices 5 – 40 residues long   -strands 5 – 10 residues long

Outline  What is Secondary Structure  Introduction to Secondary Structure Prediction  Chou-Fasman Method  Nearest Neighbor Method  Neural Network Method

Secondary structure propensity l From PDB database, calculate the propensity for a given amino acid to adopt a certain ss-type l Example: #Ala=2,000, #residues=20,000, #helix=4,000, #Ala in helix=500 P( ,aa i ) = 500/20,000, p(  p(aa i ) = 2,000/20,000 P = 500 / (4,000/10) = 1.25

Chou-Fasman algorithm l Helix, Strand 1. Scan for window of 6 residues where average score > 1 (4 residues for helix and 3 residues for strand) 2. Propagate in both directions until 4 (or 3) residue window with mean propensity < 1 3. Move forward and repeat l Conflict solution Any region containing overlapping alpha-helical and beta-strand assignments are taken to be helical if the average P(helix) > P(strand). It is a beta strand if the average P(strand) > P(helix). l Accuracy: ~50%  ~60% GHWIATRGQLIREAYEDYRHFSSECPFIP

Initiation Identify regions where 4/6 have a P(H) >1.00 “alpha-helix nucleus”

Propagation Extend helix in both directions until a set of four residues have an average P(H) <1.00.

Outline  What is Secondary Structure  Introduction to Secondary Structure Prediction  Chou-Fasman Method  Nearest Neighbor Method  Neural Network Method

Nearest neighbor method o Predict secondary structure of the central residue of a given segment from homologous segments (neighbors)  (i) From database, find some number of the closest sequences to a subsequence defined by a window around the central residue, or  (ii) Compute K best non-intersecting local alignments of a query sequence with each sequence. o Use max (n , n , n c ) for neighbor consensus or max(s , s , s c ) for consensus sequence hits

Environment preference score l Each amino acid has a preference to a specific structural environments. l Structural variables: å secondary structure, solvent accessibility l Non-redundant protein structure database: FSSP

“Singleton” score matrix Helix SheetLoop Buried Inter Exposed Buried Inter Exposed Buried Inter Exposed ALA ARG ASN ASP CYS GLN GLU GLY HIS ILE LEU LYS MET PHE PRO SER THR TRP TYR VAL

Total score l Alignment score is the sum of score in a window of length l: T R G Q L I R i-4 i-3 i-2 i-1 i i+1 i+2 i+3 i+4 E A Y E D Y R H F S S E C P F I P...E C Y E Y B R H R.. j-4 j-3 j-2 j-1 j j+1 j+2 j+3 j+4 | | | | | L H H H H H H L L

Neighbors 1 - L H H H H H H L L - S L L H H H H H L L - S L E E E E E E L L - S L E E E E E E L L - S 4 n - L L L L E E E E E - S n n+1 - H H H L L L E E E - S n+1 :  max (n , n , n L ) or max (  s ,  s ,  s L )

Evolutionary information l “All naturally evolved proteins with more than 35% pairwise identical residues over more than 100 aligned residues have similar structures.” l Stability of structure w.r.t. sequence divergence (<12% difference in secondary structure). l Position-specific sequence profile, containing crucial information on evolution of protein family, can help secondary structure prediction (increase information content). l Gaps rarely occur in helix and strand. l ~1.4%/year increase in Q3 due to database growth during past ~10 years.

How to use it l Sequence-profile alignment. l Compare a sequence against protein family. l More specific. l BLAST vs. PSI-BLAST. l Look up PSSM instead of PAM or BLOSUM. position amino acid type

Outline  What is Secondary Structure  Introduction to Secondary Structure Prediction  Chou-Fasman Method  Nearest Neighbor Method  Neural Network Method

Neurons normal state addictive state

Neural network Input signals are summed and turned into zero or one 3. J1J1 J2J2 J3J3 J4J4 Feed-forward multilayer network Input layerHidden layerOutput layer neurons

Enter sequences Compare Prediction to Reality Adjust Weights Neural network training

Simple neural network Error = | out_net – out_desired |

Training a neural network

Simple neural network with hidden layer

Neural network for secondary structure

PsiPred l D. Jones, J. Mol. Boil. 292, 195 (1999). l Method : Neural network l Input data : PSSM generated by PSI-BLAST l Bigger and better sequence database å Combining several database and data filtering l Training and test sets preparation å Secondary structure prediction only makes sense for proteins with no homologous structure. å No sequence & structural homologues between training and test sets by PSI-BLAST (mimicking realistic situation).

Psi-Pred (details) l Window size = 15 l Two networks l First network (sequence-to-structure): å 315 = (20 + 1)  15 inputs å extra unit to indicate where the windows spans either N or C terminus å Data are scaled to [0-1] range by using 1/[1+exp(-x)] å 75 hidden units å 3 outputs (H, E, L) l Second network (structure-to-structure): å Structural correlation between adjacent sequences å 60 = (3 + 1)  15 inputs å 60 hidden units å 3 outputs l Accuracy ~76%

Reading Assignments l Suggested reading: å Chapter 15 in “Current Topics in Computational Molecular Biology, edited by Tao Jiang, Ying Xu, and Michael Zhang. MIT Press ” l Optional reading: å Review by Burkhard Rost: ev_dekker/paper.html ev_dekker/paper.html

Develop a program that implements Chou- Fasman Algorithm 1. TA will give you a matrix table of Chou-Fasman indices 2. Using the FASTA as input format for sequence 3. Output format: Project Assignment KVFGRCELAA AMKRHGLDNY RGYSLGNWVC AAKFESNFNT QATNRNTDGS HHHHHH HHHH HHHHHH HHHHHH EEE TDYGILQINS RWWCNDGRTP GSRNLCNIPC EEE EE