Postgenomic Era

1
Postgenomic Era

• Having learned how to sequence entire genomes, we
  want to move beyond just the sequence
• First major area Proteomics
• Study of all proteins within a cell/organism
• One Clear Difficulty
• The proteome is tissue dependent
• Every one of your cells has the identical genome,
  but the proteome of your cells varies by both
  time and space and can be environmentally
  dependent

2
Protein Function

• Protein functions are often divided into three
  subparts
• Biological Process
• Why is the protein doing what it is doingwhat is
  the goal (e.g., lactose metabolism)
• Molecular Function
• What does the protein do? (e.g., its an enzyme
  that helps convert sugar A to sugar B)
• Cellular Component
• Where is the protein expressed and active (e.g.,
  pancreatic cells)
• Can any of these be answered computationally?

3
Molecular Function

• In order to understand what a protein actually
  does, we need to know what it looks like
• About 20,000 proteins have experimentally
  determined structures
• As of June 2003, more than 1 million protein
  sequences had been determined

4
Protein Structure

• Protein structure can be thought of in multiple
  dimensions
• Primary Structure Protein sequence
• Human C2 protein
• MGPLMVLFCLLFLYPGLADSAPSCPQNVNISGGTFTLSHGWAPGSLLTYS
  CPQGLYPSPASRLCKSSGQWQTPGATRSLSKAVCKPVRCPAPVSFENGIY
  TPRLGSYPVGGNVSFECEDGFILRGSPVRQCRPNGMWDGETAVCDNGAGH
  CPNPGISLGAVRTGFRFGHGDKVRYRCSSNLVLTGSSERECQGNGVWSGT
  EPICRQPYSYDFPEDVAPALGTSFSHMLGATNPTQKTKESLGRKIQIQRS
  GHLNLYLLLDCSQSVSENDFLIFKESASLMVDRVRNQESACSRGLPVLTI
  SLCLLPLLRTPLTAHLLQEVFSDYTHAM

5
Protein Structure

• Can be sequenced directly or derived from DNA
  sequence
• Molecular structure of proteins amino acid
  chain
• Amino acid

R the side chain this is what makes one amino
acid different from another
Glycine R a single hydrogen atom (H) Alanine
R CH3 Tryptophan R a double ring
6
Protein Structure
Amino acid chain
Peptide Bond

• The side chains are what give specific properties
  to amino acids
• Size
• Electric Charge
• Polarity
• Shape and rigidity
• These properties describe how amino acids can
  interact
• Example positive and negative charged amino
  acids can interact as salt bridges

7
Protein Structure

• Secondary Structure
• Basic folding of primary sequence into common
  subparts
• Backbone of a protein is made up of the non-side
  chain atoms
• Peptide bond (C O NH) is rigid and planar
• All flexibility in the backbone is due to the
  bonds from the alpha carbon to the neighboring C
  and N, since these bonds can rotate

8
Protein Structure

• Major Secondary Structures
• Alpha helices where both bonds have -60
  degree angles
• Beta strands where the bonds alternate -130 and
  135 degree angles
• Beta strands combine to form Beta sheets
• Beta turns The U turn between beta strands in a
  beta sheet

9
Protein Structure
Alpha Helix
10
Protein Structure
Beta Sheet
Beta Strand
11
Protein Structure

• Tertiary Structure
• Folding of subparts in the full three-dimensional
  protein
• Quarternary structure
• How multiple proteins interact to form larger
  molecules the interaction may cause changes in
  tertiary structure

12
Graphical Display
Graphical methods for displaying 3d structure can
be very complicated and difficult to understand
without training
13
Structural Proteomics

• Goal To discover the exact three-dimensional
  location of every atom within a protein
• Experimental approach
• Purify large amounts of the protein
• Crystalize the protein
• Use X-ray crystallography of nuclear magnetic
  resonance imaging to find locations of the atoms
• Requires computational analysis of the X-ray/NMR
  results
• This is difficult and time consuming

14
X-Ray Crystallography
15
Structural Proteomics

• Computational Prediction
• Theoretically it should be possible to predict
  full protein structure from a sequence
• Every protein sequence naturally folds into shape
  in a fraction of a second therefore, simply
  knowing the protein sequence show provide all
  necessary information to predict the structure
• In practice this is extraordinarily difficult
• One approach is to take the atomic structure of
  the protein (the atom by atom diagram of the
  entire amino acid chain) and calculate the
  optimal energy binding of Van der Waals forces

16
Structural Proteomics

• No computers that exist can currently do this
• One of the fastest computers in the world is
  being designed by IBM to solve this very problem
  (started in 1999)
• Called Blue Gene
• 100 million development costs
• We may still be decades or more away from this
  approach succeeding

17
Secondary Structure Prediction

• What can be done today?
• Secondary structure prediction
• Chau-Fasman algorithm
• There are many others, but this is one of the
  most common ones
• Each amino acid has been assigned as set of
  parameters
• P(a), P(b), P(turn), f(i), f(i1), f(i2), f(i3)
• Parameter values were estimated from a large set
  of proteins with known structure
• The Ps represent the propensity of a specific
  amino acid to be found in alpha helices, beta
  strands, and beta turns
• The fs represent the frequency with which each
  amino acid is found in a hairpin turn

18
Secondary Structure Prediction
Example of the table
19
Secondary Structure Prediction

• Algorithm
• Alpha helices
• Find all regions where 4 out of 6 consecutive
  amino acids have P(a) gt 100
• Extend each region until 4 consecutive amino
  acids with P(a) lt 100
• Add P(a) and P(b) for each region
• If region is longer than 5 amino acids and SP(a)
  gt SP(b), it is assumed to represent an alpha
  helix
• Beta strand
• Same exact approach, but SP(b) gt SP(a) and
  average P(b) gt 100

20
Secondary Structure Prediction

• If helix and sheet prediction overlap, the one
  with the greater sum P(a) vs P(b) is assumed to
  be correct
• Beta turns
• For each amino acid, calculate turn probability
  by multiplying f(i) f(i1) for the next amino
  acid f(i2) for the next f(i3) for the next
• Assumed to be a hairpin turn if
• Product is gt 0.00075
• The mean P(turn) for all four amino acids is gt
  100
• S P(turn) gt SP(a) and SP(b) for all 4 amino acids

21
Secondary Structure Prediction

• Many other possible algorithms
• Hidden Markov Models, Neural networks, etc.
• Best methods tend to identify about 75 of
  secondary structures

22
Tertiary Structure

• Much more difficult than secondary structure
• Important features
• Hydrophobicity
• Disulfide bonds
• Cysteine residues cross-link to form a very
  strong, stabilizing bond
• Cysteine molecules are often more strongly
  conserved than other amino acids and can be
  extremely important for alignment
• Interacting amino acids can be fairly close or
  very far apart in the protein chain!

23
Tertiary Structure

• Three basic approaches
• Homology modeling
• BLAST protein for a similar sequence with known
  structure
• If protein sequences are more than 30 identical,
  align (or multiple align) and use known structure
  as a template for estimating the structure of
  unknown protein
• Order of construction
• Model core backbone
• Model loops
• Loops will likely vary more between homologous
  proteins than the core backbone
• Model side chains

24
Tertiary Structure

• How good is homology modeling?
• When gt 50 of the sequence is identical, results
  are usually excellent
• Alignment error is one of the main causes or
  prediction mistakes

25
Tertiary Structure

• Second approach Fold Recognition
• Similar to BLAST search but where youre
  searching for secondary structure patterns
  (rather than the primary sequence) in a database
• Essentially trying to align folds rather than
  amino acids
• If you find proteins with similar secondary
  structure patterns, you assume the tertiary
  structures are similar
• There are known cases or proteins with almost no
  similarity in primary sequence having extremely
  similar three-dimensional shapes
• One form of advanced fold recognition is known as
  Threading
• Does a more thorough search than standard fold
  recognition sort of an optimization algorithm
  for protein folding

26
Tertiary Structure

• Third approach ab initio prediction
• Predict structure from physical properties
• Goal of IBMs Blue Gene project
• Derive secondary structure from amino acid
  sequence (alpha helix, beta sheets, coils, etc.)
• Fold these structures into tertiary structures
  based on physical properties
• Currently these methods are restricted to the
  carbon atoms of the peptide backboneside chains
  are still too difficult
• Accuracy is relatively poor compared to other
  methods
• Accuracy is measured as the difference in
  angstroms between predicted and experimentally
  determined locations of atoms

27
Protein Prediction Competition

• CASP Critical Assessment of Structure
  Prediction
• Held every two years
• Contest for protein structure prediction
• A group experimentally determines the structure
  of novel proteins
• Sequences are published, but the structures are
  not
• Outside groups try to predict the structures and
  submit their results
• A big meeting/party is held to announce how
  everyone did

28
RNA structure

• Similar idea for the most part
• Secondary structure consists of complementary
  base pairing between a single stranded sequence
  and itself
• Tertiary structure has similar problems as those
  found with proteins