To be folded or to be unfolded journal club 120604

1
To be folded or to be unfolded(journal club
12/06/04)
S. O. Garbuzynskiy, M. Y. Lobanov and O. V.
Galzitskaya in Protein Science (2004), 13,
2871-2877
Apostol Gramada San Diego Supercomputer Center,
2School of Pharmacology, University of California
at San Diego, La Jolla, California, USA
2
Unfolded proteins (natively unfolded,
intrinsically unstructured)

• Natively unfolded proteins lack ordered structure
  under conditions of neutral pH in vitro.
• In vivo, they are likely to be stabilized by
  binding of specific targets, ligands (small
  molecules, substrates, cofactors, other proteins,
  nucleic acids, membranes, etc.).
• It has been suggested that the lack of globular
  structure is a functional advantage, allowing
  them to interact more efficiently with several
  targets. Moreover, a disorder-order transition
  induced during binding represent a simple
  mechanism for regulation of numerous cellular
  processes.

3

• There is a considerable number of unfolded or
  partially unfolded proteins under normal
  physiological conditions (more than 100).
• It has been predicted that more than 15000
  proteins contain long (40 residues or more)
  disordered segments.
• Are implicated in a wide range of functional
  roles. Ex entropic machines (bristles, spacers,
  linkers, clocks) but also implicated in protein
  transport (metal ion binding), signal
  transduction and molecular recognition.
• Some of these proteins are associated with
  neurodegenerative disorders (BSE, CJD,
  Alzheimers and Parkinsons diseases).

4
Unfolded proteins

• A paradigm shift from the common dogma A 3D
  structure is imperative for function.
• The Structure-function paradigm is being
  reconsider
• Protein Trinity Paradigm three thermodynamic
  states, ordered, molten globule and random coil
  with function arising from any of the three
  conformations and transitions between them
  (Dunker et al, (2002), Biochemistry 41, 6573).
• Protein Quartet Model premolten globule state
  added to the picture (Uversky (2002), Protein
  Science, 11, 739).
• AA sequence determines 3D structure, then it
  should also determine lack of 3D structure (not
  an obvious assumption).
• It appears that natively unfolded proteins have
  low hydrophobicity and high net charge (previous
  work by Uversky).

5
To be folded or to be unfolded the paper

• Are there intrinsic properties of amino acid
  residues that are responsible for the absence of
  a fixed structure at physiological conditions?
  Two sets of indicators are proposed and studied
  in comparison with others by Uversky
• Expected average number of contacts per residue
• Artificial parameters obtained by Monte-Carlo
  optimization
• A weaker version of the assumption that sequence
  determines lack of structure.
• The goal is to separate as well as possible two
  classes of proteins natively unfolded and
  globular
• Do the results support the claim that these
  indicators are effective?

6
Average number of contacts - motivation

• The globular structure is the result of the
  competition between entropic and energetic free
  energy.
• A contact reduces the energy of interaction.
• The more compact the structure, the less
  conformational entropy.
• The larger of potential number of contacts in a
  globular state, the smaller the energy of
  interaction and therefore the larger the
  conformational entropy they can compensate for.

7
Databases

• Globular Proteins (extracted from SCOP 1.63)
• one chain only
• nonhomologous single-domain without modified
  residues
• no serious errors in connectivity
• no disulfide bonds and ligands
• all heavy atoms resolved
• from the four general classes of SCOP
• The length of proteins 54-500 residues.
• 80 proteins overall.

8

• Natively unfolded (from Uversky et al. (2000) and
  SWISS-PROT
• Nuclear magnetic resonance chemical shifts of a
  random-coil
• Lack of significant secondary structure
• Show hydrodynamic dimensions close to those
  typical of an unfolded polypeptide chain
• Length 50-1827
• 90 proteins overall.

9

• Data base for number of contacts (extracted from
  SCOP 1.61)
• 6626 domains from seven general classes (a-g)
  with less than 80 sequence identity 1122 (?
  from a), 1644 (? from b), 1617 (?/? from c), 1435
  (?? from d), 142 (multidomain from e), 127
  (membrane from f) and 528 (small from g).
• Calculated the average number of contacts for
  each of the 20 residues.

10
Properties of amino acid residues

• Average number of contacts is the sum over all
  residues in the sequence of the residue specific
  number of contacts divided by the length of the
  chain.

11
Results for the expected number of contacts
12

• With a border at 20.73 contacts per residue, the
  prediction accuracy is 89.

13
Other indicators considered

• Only the artificial parameters and the expected
  of contacts (even though, this one marginally)
  are statistically different for the two classes.
  The next closest seem to be the hydrophobicity.

14
Hydrophobicity
15

• The prediction accuracy is 83.

16
Charge (per residue)
17

• The prediction accuracy is 76.

18
Two parameters simultaneously

• More than two parameters doesnt help to
  separate. However, number of contacts and
  hydrophobicity lead to an 8 accuracy.

19
Artificial parameters

• The parameters are optimized using a Monte-Carlo
  algorithm.

20
Artificial parameters results
21
Artificial parameters - results

• The accuracy reaches 5, still not fully
  separated.

22
Parameter correlation

• The best correlation is between the artificial
  parameters and expected number of contacts.
• The next best does not ( of C and S atoms have a
  poor separation).

23
Conclusions

• None of the parameters is able to separate the
  classes.
• The idea of uniquely separation of naturally
  unfolded proteins in the charge-hydrophobicity
  space is somewhat challenged here.
• The set of artificial parameters are better
  correlated with the number of contacts than the
  other parameters with good separation of the two
  classes.
• The use of several structural properties may
  possibly separate the databases.
• The scores are only sensitive to composition but
  not to the order in the sequence that might be
  one of the reasons for the above conclusion
  according to authors. Contact profiles may be
  also able to predict unfolded regions for
  partially structured proteins.