Protein World

1
Protein World

• SARA
• 12-12-2002 Amsterdam
• Tim Hulsen

2
Genome sequencing

• Since 1995 sequencing of complete genomes
  (DNA) A/C/G/T order
• ACGTCATCGTAGCTAGCTAGTCGTACGTATG
• TGCAGTAGCATCGATCGATCAGCATGCATAC
• At this moment more than 80 genomes have been
  sequenced and published, of all kinds of
  organisms
• Animals
• Plants
• Fungi
• Bacteria

3
Genomes ? Proteins

• Transcription and translation of specific
  regions of the genome leads to proteins,
  consisting of twenty types of amino acids
• ATG ACG CTG AGC TGC GGA CGT TGA -gt TLSCGR
• Proteins are responsible for all kinds of life
  processes
• All the proteins that can be produced in an
  organism
• together are called the proteome
• Sequence comparisons make
• possible the classification of
• proteins

4
Protein families

• e.g. The GPCR family
• Sequence comparison helps in predicting the
  function of new proteins

5
Determining protein functions

• Function of 40-50 of the new proteins is unknown
• Understanding of protein functions and
  relationships is important for
• Study of fundamental biological processes
• Drug design
• Genetic engineering

6
Sequence comparison

• Smith-Waterman dynamic programming algorithm
  (1981) calculates similarity/distance between
  two sequences
• Query ---PLIT-LETRESV-
• Subject NEQPKVTMLETRQTAD
• (boldsimilar)
• Results in a SW-score that is a measure for how
  similar the two sequences are to each other
• Disadvantage score is dependent of length
• After the alignments, the proteins are
  clustered (divided into families) according to
  their similarity

7
Existent databases

• Domain-based clusterings Prosite, Pfam, ProDom,
  Prints, Domo, Blocks
• Protein-based clusterings ProtoMap, COGs,
  Systers, PIR, ClusTr
• Structural classifications SCOP, CATH, FSSP
• Why should there be another database?

8
Another method

• Enhanced Smith-Waterman algorithm Monte-Carlo
  evaluation (Lipman et al., 1984)
• How big is the chance that two sequences are
  similar but not related?
• One of the two sequences is randomized and
  recalculated (200 times). Randomization leads to
  sequences with the same length and the same
  composition, but different order
• Method leads to calculation of the Z-value
• S(A,B) - µ
• Z(A,B) -------------------
• s

9
Advantages

• The obtained Z-value is a very reliable measure
  for sequence, compared to SW-score
• SW-score is dependent of length, Z-value is not
• Amino acid bias does not affect the Z-value
• Independent of the database size
• Easier updating of the database, without a total
  recalculation

10
Disadvantage

• LOTS of calculation time needed, especially when
  all proteins in all proteomes are compared to
  each other (all-against-all)!
• ? SARA

11
SARA calculation

• Proteomes of 82 organisms compared
  all-against-all with the use of the Monte Carlo
  algorithm more than 400,000 proteins!
• 21,600 CPU days (520,000 CPU hours)
• 21,600 PCs running parallel over 24 hours / 1
  PC running for 60 years
• Using supercomputer TERAS (1024-CPU SGI Origin
  3800) at SARA less than two months!

12
Parties involved

• Gene-IT (Paris, France)
• SARA (Amsterdam, the Netherlands)
• CMBI (Nijmegen, the Netherlands)
• Organon (Oss, the Netherlands)
• EBI (Hinxton, UK)

13
Supporting parties

• Financed by NCF, foundation in support of
  supercomputing
• Under the auspices of BioASP, the new Dutch
  knowledge and service center for Bioinformatics

14
Results available through BioASP

• http//www.bioasp.nl
• Log in and click on links Research and Protein
  World

1
2
15
Results available through BioASP

• Organism selection screen

16
Results available through BioASP

• Results screen

17
Results available through BioASP

• Alignment screen

18
Conclusions

• Currently the most comprehensive and most
  accurate data-set of protein comparisons
• A start for a maintainable and unique database of
  all proteins currently known
• A rich data-source for clustering, data-mining
  and orthology determination

19
Orthology determination

• Orthologs genes/proteins in different species
  that derive from a common ancestor
• Orthologs often have the same function
• Interesting! Information from other species could
  help in annotating a protein

20
Thank you for your attention

• Any questions?