Transcript Faik Bioinformatics PowerPoint 2-2006

Tools to analyze protein characteristics
3-D fold model
Identification of
conserved regions
-Family member
-Multiple alignments
Evolutionary
relationship (Phylogeny)
Protein
sequence
Protein sorting and
sub-cellular localization
Some
Signal sequence
(tags)
Anchoring into
the membrane
nascent proteins contain a specific signal, or targeting sequence
that directs them to the correct organelle. (ER, mitochondrial, chloroplast,
lysosome, vacuoles, Golgi, or cytosol)
Questions
Can
we train the computers:
To detect signal sequences and predict protein destination?
To identify conserved domains (or a pattern) in proteins?
To predict the membrane-anchoring type of a protein?

(Transmembrane domain, GPI anchor…)
To predict the 3D structure of a protein?
Learning
algorithms are good for solving problems in pattern
recognition because they can be trained on a sample data set.
Classes
of learning algorithms:
-Artificial neural networks (ANNs)
-Hidden Markov Models (HMM)
Artificial neural networks (ANN)
learning algorithms that mimic the brain. Real brains,
however, are orders of magnitude more complex than any
ANN so far considered.
Machine
ANN
is composed of a large number of highly interconnected
processing elements (neurons) working simultaneously to solve
specific problems.
like people, learn by example. ANNs cannot be programmed
to perform a specific task.
ANNs,
The
first artificial neuron was developed in 1943 by the
neurophysiologist Warren McCulloch and the logician Walter Pits.
Hidden Markov Models (HMM)
Used
to answer questions like:
What is the probability of obtaining a particular outcome?
What is the best model from many combinations?

HMM
is a probabilistic process over a set of states, in which the
states are “hidden”. It is only the outcome that visible to the
observer. Hence, the name Hidden Markov Model.
HMM
has many uses in genomics:
Gene prediction (GENSCAN)
SignalP
Finding periodic patterns

The ExPASy (Expert Protein Analysis System)
Expasy
server (http://au.expasy.org)
is dedicated to the analysis of
protein sequences and structures.
Sequence
analysis tools include:
DNA -> Protein [Translate]
Pattern and profile searches
Post-translational modification and
topology prediction
Primary structure analysis
Structure prediction (2D and 3D)
Alignment


PredictProtein: A service for sequence analysis, and structure prediction
http://www.predictprotein.org/newwebsite/submit.html

TMpred:

TMHMM: Predicts transmembrane helices in proteins (CBS; Denmark)
http://www.ch.embnet.org/software/TMPRED_form.html
http://www.cbs.dtu.dk/services/TMHMM-2.0/

big-PI : Predicts GPI-anchor site:http://mendel.imp.univie.ac.at/sat/gpi/gpi_server.html

DGPI: Predicts GPI-anchor site: http://129.194.185.165/dgpi/index_en.html

SignalP: Predicts signal peptide: http://www.cbs.dtu.dk/services/SignalP/

PSORT: Predicts sub-cellular localization:

TargetP: Predicts sub-cellular localization: http://www.cbs.dtu.dk/services/TargetP/

NetNGlyc: Predicts N-glycosylation sites:http://www.cbs.dtu.dk/services/NetNGlyc/

PTS1: Predicts peroxisomal targeting sequences
http://www.psort.org/
http://mendel.imp.univie.ac.at/mendeljsp/sat/pts1/PTS1predictor.jsp

MITOPROT: Predicts of mitochondrial targeting sequences
http://ihg.gsf.de/ihg/mitoprot.html
Hydrophobicity: http://www.vivo.colostate.edu/molkit/hydropathy/index.html
Multiple alignment
Used
to do phylogenetic analysis:
Same protein from different species
Evolutionary relationship: history

Used
to find conserved regions
Local multiple alignment reveals conserved regions
Conserved regions usually are key functional regions
These regions are prime targets for drug developments
Protein domains are often conserved across many species

Algorithm

for search of conserved regions:
Block maker: http://blocks.fhcrc.org/blocks/make_blocks.html
Multiple alignment tools
Free
programs:
Phylip and PAUP: http://evolution.genetics.washington.edu/phylip.html
Phyml: http://atgc.lirmm.fr/phyml/

The
most used websites :
http://align.genome.jp/
http://prodes.toulouse.inra.fr/multalin/multalin.html
http://www.ch.embnet.org/index.html (T-COFFEE and ClustalW)

ClustalW:

Standard popular software

It aligns 2 and keep on adding a new sequence to the alignment

Problem: It is simply a heuristics.
Motif

discovery: use your own motif to search databases:
PatternFind: http://myhits.isb-sib.ch/cgi-bin/pattern_search
Phylogenetic analysis
Phylogenetic
Describe
Major
trees
evolutionary relationships between sequences
modes that drive the evolution:
Point
mutations modify existing sequences
Duplications (re-use existing sequence)
Rearrangement
Two
most common methods
Maximum
parsimony
Maximum likelihood
Parsimony vs Maximum likelihood
Parsimony
is the most popular method in which the simplest
answer is always the preferred one.
It involves statistical evaluation of the number of mutations need
to explain the observed data.
The best tree is the one that requires the fewest number of
evolutionary changes.

In
contrast, maximum likelihood does not necessarily satisfy
any optimality criterion. It attempts to answer the question:
What parameters of evolutionary events was likely to produce the
current data set?
This is computationally difficult to do. This is the slowest of all
methods.

Likelihood
generally performs better than parsimony
Definitions
Homologous:Have
Orthologous:
Paralogous:
a common ancestor. Homology cannot be measured.
The same gene in different species . It is the result of
speciation (common ancestral)
Related genes (already diverged) in the same species. It is
the result of genomic rearrangements or duplication
Determining protein structure

Direct measurement of structure
X-ray crystallography
NMR spectroscopy



Site-directed mutagenesis
Computer modeling
Prediction of structure
Comparative protein-structure modeling

Comparative protein-structure modeling

Goal:Construct 3-D model of a protein of unknown
structure (target), based on similarity of sequence to
proteins of known structure (templates)

Procedure:
Template selection
Template–target alignment
Model building
Model evaluation

Blue: predicted model by PROSPECT
Red: NMR structure
The Protein 3-D Database
The
Protein DataBase (PDB) contains 3-D structural data
for proteins
Founded
in 1971 with a dozen structures
As
of June 2004, there were 25,760 structures in the database.
All structures are reviewed for accuracy and data uniformity.
80% come from X-ray crystallography
16% come from NMR
2% come from theoretical modeling

Structural
data from the PDB can be freely accessed at
http://www.rcsb.org/pdb/
High-throughput methods
Most used websites for 3-D structure prediction
Protein
Homology/analogY Recognition Engine (Phyre) at
http://www.sbg.bio.ic.ac.uk/phyre/html/index.html
PredictProtein
at
http://www.predictprotein.org/newwebsite/submit.html
UCLA
Fold Recognition at
http://www.doe-mbi.ucla.edu/Services/FOLD/