Fundamentals of Bioinformatics - Ogurtsov A.N. 2013

Information Principles in Biotechnology
Protein Analysis and Prediction
Challenges in Protein Function Prediction

Ideally, we can reconstruct a protein's Structure from its Amino Acid Sequence, and subsequently deduce its function from that structure.

In reality, despite the confidence that similar Amino acid sequences will yield similar protein structures, the relationship between Structure and function is more complex.

Proteins with identical structures, and even identical sequences, can be employed by organisms to carry out entirely different Functions.

Conversely, vastly different proteins can perform the same function.

Furthermore, just as unrelated sequences can share a similar structure, unrelated proteins with distinct structures can execute the exact same function.

During evolution, proteins may:

✵ preserve function and Specificity;

✵ preserve function while altering specificity;

✵ shift to a similar or identical function within a different metabolic "context";

✵ switch to an entirely different function.

Often, a shift in function is not tied to any change in protein sequence or structure at all. Some proteins are multifunctional, requiring no structural alterations to change their biological function.

For instance, in ducks, active Lactate dehydrogenase and enolase function as crystallins within the eye lens, even though they never encounter their substrates in situ. In other cases, crystallins are closely related to Enzymes, although some divergence resulting in the loss of catalytic activity has already occurred (proving that enzymatic activity is not required within the eye lens).

Another example is a protein from E. coli called Do, DegP, or HtrA, which functions as a chaperone (catalyzing protein folding) at low temperatures, but turns into a proteinase when the Temperature rises above 42° С. This is apparently explained by the fact that under normal conditions or moderate heat stress, the task of Do is to protect proteins that are having difficulty folding; whereas under harsher heat stress, when saving proteins is no longer possible, they must be degraded.

As another example, the E. coli enzyme lipoate dehydrogenase also serves as an essential subunit in the Pyruvate dehydrogenase, 2-oxoglutarate dehydrogenase, and Glycine Cleavage complexes (see Section 12.3).

These Examples of the relationship between structure and function lie at the boundaries of a wide range of possible protein behaviors.

It is often difficult to quantify differences in function—that is, to formulate a criterion for distinguishing when two different functions are more similar to each other than two other different functions. In some cases, an altered function may share an underlying mechanism. For instance, the enolase family contains several homologous enzymes that catalyze different reactions while retaining common Features of the reaction mechanism. This group includes enolase itself, mandelate racemase, muconate-lactonizing enzyme I, and D-glucarate dehydratase. All of them Abstract a proton from the a-carbon atom of a carboxylic acid, forming an enolate intermediate. However, the subsequent transformations and The Nature of the product vary from enzyme to enzyme. These proteins generally share very similar Triosephosphate isomerase (TIM-barrel) structures. Different sets of Amino Acids in the active sites of these enzymes are responsible for the fact that they catalyze different reactions.

Functional Divergence: Orthologs and Paralogs. The family of Chymotrypsin-like Serine proteases includes closely related enzymes that have retained their function, as well as highly diverged homologs that have acquired new functions. Trypsin is a mammalian digestive enzyme that catalyzes the Hydrolysis of peptide bonds adjacent to a positively charged Arg or Lys residue (see [9], Section 13.4).

Trypsins with similar sequences, structures, functions, and specificities exist in many species, including humans, cows, Atlantic salmon, and even the bacterium Streptomyces griseus. The similarity of the S. griseus enzyme to vertebrate trypsins suggests Horizontal Gene Transfer.

For the three vertebrate enzymes (human, bovine, and salmon), each pair of sequences shares more than 64% identical residues in the alignment, whereas the bacterial homolog shares only about 30% identity with them. Nevertheless, all of them are very similar in structure. These enzymes are orthologs—homologous proteins found in different species (see Section 10.1).

Evolution has also produced related enzymes with different specificities within organisms of the same species. Chymotrypsin and pancreatic Elastase are also digestive enzymes that, like trypsin, cleave peptide bonds; however, chymotrypsin cleaves bonds adjacent to large, flat hydrophobic residues (Phe, Trp), whereas elastase cleaves bonds adjacent to small residues (Ala) (see [9], Section 13.4). Another homolog, leukocyte elastase, is essential for phagocytosis and defense against infection. Under certain conditions, leukocyte elastase is responsible for lung damage leading to emphysema.

Some trypsin homologs have acquired fundamentally new functions:

✵ Haptoglobin is a chymotrypsin homolog that has lost its proteolytic activity. It acts as a chaperone preventing unwanted protein aggregation. Haptoglobin forms a compact complex with Hemoglobin fragments released from red Blood Cells, thereby producing several beneficial effects, such as preventing iron loss.

✵ The rhinovirus serine protease has evolved a special, independent function by forming an initiation complex during RNA Synthesis, utilizing amino acids on the side of the molecule opposite to the proteolytic Active Site. This is an example not of active site modification, but of the creation of a novel function.

✵ Subunits homologous to serine proteases appear in plasminogen-like growth factors. The Role of these subunits in growth factor activity is still unknown, but they cannot perform a proteolytic function because the essential catalytic Amino acids have been lost.

✵ The insect "immune" protein scolexin is a distant homolog of serine proteases that induces hemolymph coagulation in response to infection.

In the chymotrypsin family, we observe Structural and functional conservation in closely related proteins, alongside progressive functional divergence in some (though not all) distant homologs. The key takeaway is that the overall protein folding pattern is an unreliable guide for predicting function, especially for highly divergent homologs. To accurately predict the function of distant proteins, analyzing the architecture of the active site is a far more reliable approach. For example:

✵ viral 3C proteases have been identified as distant chymotrypsin homologs, even though the serine in their catalytic triad is replaced by Cysteine;

✵ distant Homology between retroviral and aspartic proteases was uncovered through conserved Asp, Thr, and Gly residues.

Analyzing characteristic patterns of amino acid packing within the active site makes it possible to predict protein function even in the absence of experimental data on overall 3D structure.

This active-site-based approach to functional prediction is similar to the Methods used in drug design to predict potential substrates capable of binding a given protein. It is also essential to incorporate other experimental data, such as tissue-specific protein expression patterns and interacting protein networks.

Attempts to determine function directly, for example via gene knockout, can occasionally succeed, but this approach proves unproductive in the case of a lethal phenotype or when function is redundantly shared by a group of proteins.

THE CONTRIBUTION OF bioinformatics to predicting protein function from amino acid sequence and structure will likely not be reduced to a single algorithm establishing a straightforward mapping between function and structure. Although there is hope that Cell/13.html">Protein Structure might one day be predicted directly from sequence, utilizing bioinformatics data to design effective experiments and interpret their results is far more critical for successful functional prediction.

METABOLISM/35.html">Selection/41.html">Review Questions and Exercises

1. What two reasons prevent predicting protein structure based solely on fundamental physical principles?

2. The difference between which two opposing contributions determines the stabilization energy of a protein?

3. Why do cumulative errors from summing a massive number of individual atomic contributions to the total stabilization energy prevent predicting protein structure using only basic physical principles?

4. What alternative approach is used for Protein Structure Prediction when no suitable homologs are available for a given protein?

5. What underlying assumption forms The basis of the Chou-Fasman method?

6. What underlying assumption forms the basis of the GOR method?

7. What are neural networks?

8. How is a simple neural network organized?

9. How is Secondary structure prediction performed using the nearest-neighbor method?

10. What three algorithms are used for protein structure Prediction Based on the evaluation of homolog structures?

11. List the steps of the homology modeling algorithm for 3D protein structures.

12. What is protein threading?

13. What two methodologies must be used for successful protein structure recognition via threading?

14. Why is the analytical Determination of the protein state with maximal stability such a challenging problem?

15. Which energy contributions must be accounted for when estimating Conformational Energy?

16. To overcome which two problems in protein structure prediction was The Molecular Dynamics method developed?

17. List the main steps of the computational Procedure in the Monte Carlo Method.