Books

• Structural Bioinformatics (2003) edited by Philip E. Bourne and Helge Weissig, Wiley-Liss, 649 pages. Complete contents at www.structuralbioinformaticsbook.com. Reviewed by Eric Martz in Biochem. Mol. Biol. Education 31:370-1, 2003.
  A goldmine of fundamental and practical information:
  • Section I introduces bioinformatics, fundamentals of protein structure, crystallography, NMR, electron microscopy, and visualization in a bit over a hundred pages.
  • Section II: the PDB format, mmCIF, the Protein Data Bank, the Nucleic Acid Database, and other databases in a bit over a hundred pages.
  • Section III: Structural Comparison of Proteins (SCOP), CATH (Class, Architecture, Topology, Homology), structural quality assurance, comparision and alignment in about a hundred pages.
  • Section IV: assignment of secondary structure, identifying domains, and inferring function from structure in about fifty pages.
  • Section V: prediction of protein-protein interactions from evolution, and electrostatic interactions in about twenty pages.
  • Section VI: proteins as drug targets in about thirty pages.
  • Section VII: protein structure prediction in about a hundred pages.
  • Section VIII: twenty pages on structural genomics.

• Introduction to Bioinformatics - A Theoretical and Practical Approach (2003) edited by Stephen A. Krawetz and David D. Womble, Humana Press. 746 pages plus CD.
  • Part I, Biochemistry, Cell and Molecular Biology provides an introduction to the underlying science in about a hundred pages.
  • Part II, Molecular Genetics introduces genomics and clinical human genetics in about a hundred pages.
  • Part III, The UNIX operating system introduces unix and Solaris, general management of bioinformatics servers, and GCG in about a hundred pages.
  • Part IV, Computer Applications introduces software resources for DNA sequences, genome databases, multiple sequence alignment, 3D visualization, and gene expression microarray tools in about three hundred pages.

• Introduction to Protein Structure, Carl Branden and John Tooze, second edition (1999), Taylor and Francis. 410 pages.
  Lavishly illustrated and eminently readable. A wonderful overview of protein structure.

• Introduction to Protein Architecture - The structural biology of proteins, Arthur M. Lesk (2000), Oxford University Press. 360 pages, numerous illustrations.

• Nature's Robots: A History of Proteins, Charles Tanford & Jacqueline Reynolds (2001), Oxford University Press. 304 pages.
  A fascinating, meticulous and scholarly account of the history of protein science embroidered with details of the personalities and events in the lives of those involved. Begins in the 18th century, before the word "protein" was established, and traces protein science through the discovery of the genetic code in the 1960's.

• An introduction to hydrogen bonding, George A. Jeffrey, Oxford University Press, 1997. 314 pages.

Journals

• Bioinformatics
• Nature
• Nature Structural Biology
• Proteins: Structure, Function, Genetics (or use their Journal Finder).
• Science
• Trends in Biochemical Sciences via BioMedNet or Trends.Com.

Journal Articles (most recent first)

• Conformational diversity and protein evolution -- a 60-year-old hypothesis revisited. Leo C. James and Dan S. Tawfik. (2003). Trends Biochem. Sci. 28:361. PubMed.
  A "new view" of proteins is presented based on the notion that conformational diversity for a single sequence may support multifunctionality, and that this in turn fosters relatively rapid evolution of function from existing folds. A spontaneous pre-equilibrium (pre-ligand equilibrium) may exist between conformers supporting different functions (Monod-Wyman-Changeaux model). Typically, ligand for a given function may stabilize the conformer that binds it at higher affinity, thus pulling the equilibrium towards the high-affinity conformer. This contrasts with induced fit in which ligand binding induces the high-affinity conformer de novo from a stable low-affinity form (Koshland-Nemethy-Filmer model). Experimental evidence for pre-equilibrium conformational diversity, and its links to functional diversity, are reviewed. A pre-equilibrium, multifunction hypotheses proposed for antibodies in the 1930's by Landsteiner and Pauling now has some experimental support. A glossary is included with terms such as promiscuity (a different function employing the original active site), and moonlighting (a different function employing a site distinct from the original active site). Intrinsicially disordered proteins show function without a stable fold, perhaps providing a clue to the evolutionary precursors of primordial folds. The "new view" is that function co-evolves with fold, rather than fold implausibly having to evolve before function. However, "a direct proof for the role of functional diversity in protein evolution is yet to be obtained".

• All-atom contacts: a new approach to structure validation. Jane S. Richardson. (2003). Chapter 15 in Structural Bioinformatics (2003).
  Protein crystallographers have, for the most part, traditionally omitted hydrogen atoms in their models. Richardson and her team have developed software to insert hydrogens, checking for steric fit and optimizing hydrogen bonding. The steric clashes they detect reveal occasional correctable errors in crystallographic models, such as side chains of Asn, Gln, and His flipped 180^o (where the best rotamer typically cannot be discerned from electron density alone), or other rotamer errors such as for Thr or Val (where the best rotamer can be discerned from the electron density map only if it has adequate local resolution). van der Waals clashes are clearly and easily visualized as Kinemages. The resulting corrections, gathered from high-resolution structures, enabled development of improved rotamer libraries and better-defined favorable regions in Ramachandran plots, which are available for use by crystallographers during model development. "All-atom contact analysis" is a powerful new method for structure validation and improvement that complements traditional methods. The resulting improvements in the model often reduce the R value and free R. The software is free, can be downloaded for local (secure) operation, or you can use it via the "MolProbity" server at kinemage.biochem.duke.edu.

• Taxonomic Group	Domain families/genome	Average domains/family
  Vertebrates	~750	~50
  Invertebrates, Plants	~670	~20
  Yeast, Bacteria with large genomes	~550	~8
  Small parasitic bacteria	~220	~2

  Sequence identity	Function
  >= 40%	Identical
  25-40%	Broadly Similar
  <25%	Can be very different

  Evolution of the protein repertoire. Cyrus Chothia, Julian Gough, Christine Vogel, Sarah A. Teichmann (2003). Science 300:1701-3. PubMed
  "A [protein] domain ... is an evolutionary unit whose coding sequence can be duplicated and/or undergo recombination. ... Domains typically have 100-250 residues. ... we only clearly know the family relationships and domain structures of those proteins that either have a known structure, or are homologous to proteins of known structure. At present, close to 50% of the sequences in the currently known genomes are homologous to proteins of known structure." (This is based on hidden Markov model methods that find about twice as many relationships as do pairwise sequence comparisons, Gough et al., 2001.) "Rough estimates indicate that two-thirds of prokaryote proteins have two or more domains. In eukaryotes ... about four-fifths of proteins are multidomain. ... The evolutionary relationships of domains in proteins of known structure are described in the Structural Classification of Proteins (SCOP) database. ... [evidence] suggests that [most] of the protein repertoire is formed by members of families that go back to the origin of eukaryotes or the origin of the different kingdoms. ... [and] that it is much easier to evolve new binding sites than new catalytic mechanisms."
  
   
  
• Computational design of receptor and sensor proteins with novel functions, Loren L. Looger, Mary A. Dwyer, James J. Smith, and Homme W. Hellinga (2003). Nature 423:185-90. PubMed - Comment
  In this breakthrough in computational design of protein binding sites for small ligands, existing binding sites were successfully redesigned to accomodate three unrelated and quite different ligands. When empirically tested, half of the redesigned proteins bound their targeted ligands with K_d's of <10 micromolar (some close to the affinities for the original ligands). All displayed impressive specificity when challenged with analog decoys.

  Five members of the E. coli periplasmic binding protein superfamily that bound monosaccharides or amino acids were redesigned to bind TNT, L-lactate, or serotonin. These target ligands are neutral, anionic, and cationic respectively. Only L-lactate is chiral. Successes with all three and multiple original specifities exclude the results being merely a lucky fluke. Ligand-contacting residues were mutated computationally, while the ligand underwent translation and rotation restricted to a volume roughly occupied by the natural ligand. The combinatorial problem (10⁵³ - 10⁷⁶ cases) was solved with a novel algorithm based upon dead-end elimination theorems. Each redesign took about three days on a 20-processor computer cluster.

  In a further tour de force, redesigned sugar chemotaxis receptors were shown to enable E. coli to respond to TNT and L-lactate (using a re-engineered signal transduction pathway that increased b-galactosidase gene expression). Potential biosensor applications include locating the sources of underwater TNT plumes leaching from unexploded military ordnance, or locating landmines. Biosensors for L-lactate or serotonin have potential clinical applications. Chirally selective receptors could aid the "preparation of optically pure pharmaceuticals from racemic mixtures".

  Finally, the methods may be extended to design of enzymes, with the advantage that the virtual transition-state intermediate may more accurately reflect the true transition state than the chemically stable transition state analogs heretofore used to select catalytic antibodies.

• Membrane proteins: the 'Wild West' of structural biology. Jaume Torres, Tim J. Stevens, and Montserrat Samsó (2003). Trends Biochem. Sci. 28:137.
  Membrane proteins are greatly underrepresented in the Protein Data Bank because they are less amenable to the classical methods of X-ray crystallography and solution NMR. This article reviews recently developed experimental and predictive methods, their pros and cons, and successes for membrane proteins. It touches upon electron crystallography, single particle methods, atomic force microscopy, solid state NMR (SSNMR), oriented samples NMR (OS NMR), magic-angel spinning NMR, site-directed spin labelling electron paramagnetic resonance, site-specific infrared dichroism, and global search molecular dynamics simulations coupled with evolutionary conservation data. With the latter, "a model for the transmembrane domain of glycophorin, <1 Å root-mean-square deviation away from the structure painstakingly determined using NMR, was obtained in just a few days of simulations, without any experimental data". See Integral Membrane Proteins in the Atlas of Macromolecules.
   
  
• Extended disordered proteins: targeting function with less scaffold. Kannan Gunasekaran, Chung-Jung Tsai, Sandeep Kumar, David Zanuy and Ruth Nussinov (2003). Trends Biochem. Sci. 28:81.
  Argues that proteins involved in extensive protein-protein interactions can function effectively despite having their structure depend upon such interactions, so that as monomers they are natively disordered. Dispensing with the structural framework (scaffold) needed to maintain a stable fold in the monomer increases efficiency by reducing size. This may account for the large percentage (roughly half) of all proteins that are predicted to be natively disordered.

• Intrinsically unstructured proteins. Tompa, Peter. (2002) Trends Biochem. Sci. 27:527. PubMed
  The disorder of intrinsically unstructured proteins (IUP's) is crucial to their functions. They may adopt defined but extended structures when bound to cognate ligands. Their amino acid compositions are less hydrophobic than those of soluble proteins. They lack hydrophobic cores, and hence do not become insoluble when heated. About 40% of eukaryotic proteins have at least one long (>50 residues) disordered region. Roughly 10% of proteins in various genomes have been predicted to be fully disordered. Presently over 100 IUP's have been identified; none are enzymes. Obviously, IUP's are greatly underrepresented in the Protein Data Bank, although there are a few cases of an IUP bound to a folded (intrinsically structured) protein. Here, Tompa suggests five functional categories for intrinsically unstructured proteins and domains: entropic chains (bristles to ensure spacing, springs, flexible spacers/linkers), effectors (inhibitors and disassemblers), scavengers, assemblers, and display sites.

• Intramolecular interactions at protein surfaces and their impact on protein function. Robertson, Andrew D. (2002) Trends Biochem. Sci. 27:521. PubMed
  "... in [soluble] globular proteins, 33% of surface residues are apolar ..." and these are often in small clusters.

• Data mining the Protein Data Bank: residue interactions. Oldfield, Thomas J. (2002) Proteins 49:510-528. PubMed
  Oldfield has created software to identify recurrent 3D structural motifs of interactions of 3 or more amino acids, using a completely objective method that is minimally biased by expectation. This paper does not report any surprising new 3D interactions, but rather concentrates on showing that the expected and well-known interactions are properly detected, in order to validate the methods. Future work may reveal novel interactions. He started by reducing the PDB to about 1,700 nonredundant, high quality domain structures (PDB file fragments, mean 130 residues). This method, dictated by the necessity for computational feasibility, precludes identification of interdomain 3D motifs. He detected 237 3D interactions of 3 to 7 residues (ignoring Gly, Ala, Leu, Ile, Val) that occurred more than five times. These were collated into least-squares aligned results and templates of 1,972 common geometrical configurations. The template coordinates can be used to search the entire PDB, or a single PDB file, for occurrences. The results included the expected metal binding sites, ligand binding sites, catalytic triads, 3-residue salt bridges, polar uncharged interactions, and aromatic interactions (most edge to ring plane).

• Discriminating between homodimeric and monomeric proteins in the crystalline state. Ponstingl, H., K. Henrick, and J. M. Thornton (2000). Proteins. 41:47-57. PubMed.
  See also Crystal Contacts and Probable Quaternary Structures.

• Extent and nature of contacts between protein molecules in crystal lattices and between subunits of protein oligomers. Dasgupta, S., G. H. Iyer, S. H. Bryant, C. E. Lawrence, and J. A. Bell. (1997). Proteins. 28:494-514. PubMed.
  Dasgupta et al. studied 58 oligomers and 223 nonvirus crystals with good resolution and completeness. They found that "crystal contact patches are frequently smaller than patches involved in oligomer interfaces". Contact patches of 10 to 100 atoms were common at crystal contacts. Patches involving 100-1,000 atoms were common in oligomer interfaces but rarely seen in crystal contacts. Nevertheless, the total number of atoms involved in crystal contacts vs. oligomer interfaces were about the same; that is, the larger number of small crystal contact patches involved about the same number of atoms as the smaller number of large oligomer interfaces. They also observed that crystal contacts tend to involve more polar interactions, while nonpolar interactions tend to predominate in oligomer interfaces. "Hydrophobic interactions lead to disordered precipitates, and not to crystals." See also Crystal Contacts and Probable Quaternary Structures.

Crystallography

Websites

• Nature of 3-D Structural Data is a very short overview available at the Protein Data Bank.

• Earliest Macromolecular Crystal Structures, part of the History of Visualization of Biological Macromolecules.

• Hendrickson, Wayne A. Synchrotron crystallography. 2000. Trends Biochem. Sci. 25:637-43. PubMed.
  Overview of the previous quarter century of crystallography by a leader in the field, emphasizing breakthroughs that matured in the 1990's. "Six prime developments have evolved into maturity, and they completely change the scene for rapid structure determination. These are undulators, charge-coupled device (CCD) detectors, cryopreservation, MAD phasing, selenomethionyl proteins and structure-solving automation." Each is explained in some detail.

• Kleywegt, GJ, AT Brünger. 1996. Checking your imagination: applications of the free R value. Structure 4:897-904.
• Kleywegt, G. J., and T. A. Jones. 1996. Phi/psi-chology: Ramachandran revisited. Structure 4:1395-1400.
• Kleywegt, G. J., and R. J. Read. 1997. Not your average density. Structure 5:1557-1569.
• Kleywegt, GJ. 2000. Validation of protein crystal structures. Acta. Crystallogr. D. Biol. Crystallogr. 56:249-265.

• Laskowski, Roman A. 2003. Structural quality assurance. Chapter 14 in Structural Bioinformatics.

• Rhodes, Gale. Crystallography made crystal clear. Academic Press, second edition, 1999.
  Highly readable introduction recommended for an overview of the method and interpretation of the results, regardless of whether you plan to do crystallography yourself.

• Richardson, Jane S. All-atom contacts: a new approach to structure validation. See reference and precis above.