Application and implementation of probabilistic

profile-profile comparison methods

for protein fold recognition

mgr inż. Jakub Paś

Dissertation supervisor: dr hab. Marcin Hoffmann

Auxiliary supervisor: dr Krystian Eitner

Introduction

● The purpose of Phd thesis is to show that profile - profile methods may

outperform other fold recognition approaches in comparison and analysis

of distantly related proteins.

● The work presents advantages of usage probabilistic profile-profile

methods over comparable fold recognition techniques in research

performed by author.

● The thesis is based on several author’s publications in area of gene

identification, detection of distant homologous, domain boundaries

detection, protein modeling, evolutionary analysis and protein - ligand

interaction

● The work shows both applications and implementations of such methods in

molecular biology software.

What is fold Recognition

● Fold recognition are the methods of fold detecting and protein tertiary

structure prediction applied for proteins lacking homologues sequences of

known fold and structure deposited in the Protein Data Bank.

● They are based on assumption that there is strictly limited number of

different protein folds in nature, mostly as a result of evolution and due to

basic physical and chemical constraints of polypeptide chains.

● Fold recognition methods are useful for protein structure prediction,

evolutionary analysis, metabolic pathways analysis, enzymatic efficiency

prediction, molecular docking and drug design.

Sequence comparison methods

used for fold recognition

Sequence based

comparison methods:

● Smith Waterman

● BLAST

Sequence - Profile methods:

● PSI - BLAST

● RPS-BLAST

Profile - profile methods.

FFAS:

● Meta-BASIC

● ORFEUS

Other (non sequence based) fold

recognition methods:

● Threading (HHpred, Raptor)

● Ab initio (Rosetta)

What are probabilistic profile -

profile comparison methods?

● A profile or PSSM simply is a set of vectors, where each vector contains

the frequency of each type of amino acid in a particular position of the

multiple sequence alignment.

● In profile - profile alignments, we have to compare two frequency vectors

and this can be done in several different ways, including calculating the

sum of pairs, the dot-product, or a correlation coefficient between the two

vectors

● The performance of a profile - profile methods depends on: calculation of

the score between two profile positions, alignment methodology and

evolutionary distance between two sequences in study.

● Because profiles contain more information they are more sensitive and

provide better alignments than sequence - sequence methods.

Improvement and assessment

of fold recognition methods

Benchmarking of fold recognition methods:

● CASP(Critical Assessment of protein Structure Prediction)

● CAFASP(Critical Assessment of Fully Automated Structure Prediction)

● LIVEBENCH (Live Benchmark)

● EVA

Application vs implementation

Implementation - realization of an application, or execution of

a plan, idea, model, design, specification, standard, algorithm,

or policy.

Eg. Implementation of profile - profile comparison methods to

create “tools” (molecular biology software)

Application - to put something into effect or into use after it

was implemented.

Eg. Usage of such software in for science discovery

(operating “tools”)

Implementations of profile-profile

comparison methods

GRDB

J Pas, P Stepniak, L Rychlewski, GRDB – Gene

Relational DataBase. Bioinfobank Library Acta, 2011.

2659.

ELM

CM Gould, F Diella, A Via, P Puntervoll, C Gemünd, S

... J Pas, S … ELM: the status of the 2010 eukaryotic

linear motif resource

Nucleic acids research 38 D167-D180 137 2010

Implementations of profile-profile

comparison methods

ORFeus

K Ginalski, J Pas, LS Wyrwicz, M Von Grotthuss, JM

Bujnicki, L Rychlewski ORFeus: Detection of distant

homology using sequence profiles and predicted

secondary structure.

Nucleic Acids Res, 2003. 31(13): p. 3804-3807

PDB-Preview

Fischer, D., J. Pas, and L. Rychlewski, The PDB-

Preview database: a repository of in-silico models

of 'on-hold' PDB entries.

Bioinformatics, 2004. 20(15): p. 2482-4.004. 20(15)

Applications of profile-profile

comparison methods

ELM: the status of the 2010 eukaryotic linear motif resource

CM Gould, F Diella, A Via, P Puntervoll, C Gemünd, S ... J Pas at al.

Nucleic acids research 38 (suppl 1), D167-D180 137 2010

Predicting protein structures accurately

M von Grotthuss, LS Wyrwicz, J Pas, L Rychlewski

Science 304 (5677), 1597-1599 5 2004

How unique is the rice transcriptome?

LS Wyrwicz, M von Grotthuss, J Pas, L Rychlewski

Science (New York, NY) 303 (5655), 168; author reply 168 6 2004

Structure prediction, evolution and ligand interaction of CHASE

domain.

J Pas., LS Wyrwicz, L Rychlewski, J Barciszewski

FEBS Lett, 2004. 576(3): p. 287-90.

Linear motif identification in

distant proteins

Protein structure prediction

Gene identification

Protein topology Detection

Applications of profile-profile

comparison methods

Two sequences encoding chalcone synthase in yellow lupin

(Lupinus luteus l.) may have evolved by gene duplication

D Narożna, J Pas, J Schneider, CJ Mądrzak

Cellular & molecular biology letters 9 (1), 95-105 5 2004

Molecular phylogenetics of the RrmJ/fibrillarin superfamily of

ribose 2'-O-methyltransferases

M Feder, J Pas, LS Wyrwicz, JM Bujnicki

Gene 302 (1-2), 129-138 66 2003

Application of 3D-Jury, GRDB, and Verify3D in fold recognition.

Proteins

M Grotthuss, J Pas, L Wyrwicz, K Ginalski, L Rychlewski

Proteins, 2003. 53 Suppl 6: p. 418-23.

Predicting protein structures accurately

LS Wyrwicz, M Von Grotthuss, J Pas, L Rychlewski

Science 304 (5677), 1597

Gene duplication detection

Molecular phylogeny

Fold Recognition

Distant homologues detection

Gene identification and detection of

distinct homologuesApplications of profile-profile comparison methods

D Narożna, J Pas, J Schneider, CJ Mądrzak, Two sequences encoding

chalcone synthase in yellow lupin (Lupinus luteus l.) may have evolved by gene

duplication. Cell Mol Biol Lett, 2004. 9(1): p. 95-105.

● In the publication about detection of chalcone synthase (CHS)

sequences encoded in yellow lupin profile-profile fold recognition

methods were used to detect two full copies of CHS.

● Using the molecular clock calibration the duplication of genes was

estimated to happened about 16 millions years ago.

● Initial multiple alignment of distant homologues from 52 plant

species was created using multiple profile-profile comparison

methods.

Gene identification and detection of

distinct homologuesApplications of profile-profile comparison methods

D Narożna, J Pas, J Schneider, CJ Mądrzak, Two sequences encoding

chalcone synthase in yellow lupin (Lupinus luteus l.) may have evolved by gene

duplication. Cell Mol Biol Lett, 2004. 9(1): p. 95-105.

● Amino-acids involved in ligand binding has been

detected.

● Catalytic, evolutionary conserved amino acids

helped to produce full multiple sequence alignment

of known sequences of whole CHS family.

Detection of domain boundaries and

modeling of complex proteinsApplications of profile-profile comparison methods

● Domain boundaries and the homologs of the Tn-C domains were

identified using Gene Relate Sequence Database (GRDB)

● The characteristic profiles were computed for every domain using

protein families collected from Pfam, COG and from other genomic

sources.

● The comparison of the target families with about 100,000 other

families was performed using Meta-BASIC program

● The models of Tenascin-C domains were performed

J Pas., et al., Analysis of structure and function of tenascin-C. Int J Biochem

Cell Biol, 2006. 38(9): p. 1594-602.

Detection of domain boundaries and

modeling of complex proteinsApplications of profile-profile comparison methods

● Usage of sensitive profile - profile sequence comparison analysis allowed to detect the order of

functional elements in large multidomain tenascin-C protein, all variable part of a molecule as

well as all isoforms.

● The number of putative fibronectin repeats was corrected. Also previously not identified HSP33

domain with was described.

J Pas., et al., Analysis of structure and function of tenascin-C. Int J Biochem

Cell Biol, 2006. 38(9): p. 1594-602.

Profile-profile comparison

allowed shows

conservation of sequence

patterns and secondary

structure despite the low

amino acid identity which

helped to study evolution

of PYP-like family.

J Pas., et al., Structure

prediction, evolution and

ligand interaction of

CHASE domain. FEBS

Lett, 2004. 576(3): p.

287-90.

Evolutionary analysis and

protein - ligand interactionApplications of profile-profile comparison methods

Evolutionary analysis and

protein - ligand interactionApplications of profile-profile comparison methods

● Molecular model of CRE1

receptor from A. thaliana

docked with (a)trans-zeatin, (b)

kinetin confirms that ligands are

entirely buried

● The visible side chain belongs

to threonine 278 whose

mutation is responsible for loss

of function.

● Molecular modeling and

docking confirmed that ligand

was entirely buried.

● Importance of threonine 278 for

the catalytics activity of the

enzyme was confirmed.

J Pas., et al., Structure prediction, evolution and ligand interaction of CHASE

domain. FEBS Lett, 2004. 576(3): p. 287-90.

Implementation: PDB PreviewImplementations of profile-profile comparison methods

● Not all the entries in the Protein Data Bank (PDB) are publicly available.

● A new structure can be deposited as an “on-hold” entry, non-accessible for

public before final release.

● To access 3D structure before release it is possible in most cases to

generate relatively accurate automatically created computational models.

● The PDB-Preview provide biologists with relatively accurate 3D models for

not yet released PDB shortly after they are deposited in the PDB, and well

before the experimental structure is released.

● Additionally the resulting PDB-CAFASP analysis provides computational

biologists with a continuous blind evaluation of their methods.

D Fisher., J Pas, and L Rychlewski, The PDB-Preview database: a

repository of in-silico models of 'on-hold' PDB entries. Bioinformatics,

2004. 20(15): p. 2482-4.

Implementation: Gene Relational

DataBase (GRDB)Implementations of profile-profile comparison methods

● GRDB is the web service dedicated for searching for distant homologues

of protein sequences which may not be detected using different

approaches such as direct sequence search.

● It performs the comparison of the target family with 100,000 other families,

using Meta BASIC profile-profile comparison methods. (SCOP, CATH,

COG)

● In contrast other methods it allows to use manually build profile as input

and perform comparison between whole protein families.

● GRDB was successfully used for comprehensive classification of proteins

folds and identification of novel families and their representatives in human

(Kuchta, et al., 2009).

J Pas., et al., GRDB – Gene Relational DataBase. Bioinfobank Library Acta,

2011. 2659.

Conclusions

● Profile-profile based sequence comparison methods are usually superior to

sequence methods and may have more possible applications in molecular

biology.

● PDB statistics shows that in recent years, only a limited number of

completely new protein folds appear although several thousand new

structures are deposited to the PDB. Most of the single-domain proteins

can be aligned to a protein already deposited in the PDB.

● The output of experimentally determined protein structures from X-ray

crystallography and NMR spectroscopy are still expensive and time

consuming despite the efforts in structural genomics.

● As more and more novel sequences are produce from the genome projects,

the profile-based methods can be expected to become even more sensitive

(new alignment, scoring methods etc.)

Acknowledgements

Dissertation supervisor: dr hab. Marcin Hoffmann

Auxiliary supervisor: dr Krystian Eitner

My wife: Agnieszka Paś

Recenzja: Prof. Tadeusz Kuliński

● Współczynnik impact factor oraz cytowalność. Autor niestety w żaden

sposób nie wyróżnia tych prac, które stanowią spójny tematyczne zbiór

artykułów i faktycznie wchodzą w skład dysertacji.

● Literówki i błedy językowe

● Błąd merytoryczny. W publikacji:Two sequences encoding chalcone synthase in yellow lupin (Lupinus luteus l.) may have

evolved by gene duplication

Narożna, J Pas, J Schneider, CJ Mądrzak

Cellular & molecular biology letters 9 (1), 95-105 5 2004

istotnie chodziło o 52 sekwencje roślinnej suntazy chalkonowej

w tym 2 z L. Luteus

Recenzja: Prof. Andrzej Koliński

● Uwagi formalne - brak oświadczeń - zostały dosłane

● Uwagi dotyczace polskich odpowiedników wyrazów angielskich w polskim

streszczeniu.

● ab inition - de novo