PN: Homology Modeling

• experimental protein structure determination is expensive and time-consuming and sometimes fails

• for many applications, we do not need to know the exact positions of all atoms, just of residues, and theoretical calculations can provide that

• (not sufficient for drug design, though)

• Darwin: Evolutional Biology

• Boltzmann: Statistical Thermodynamics

• Existent proteins are derived from a common ancestor by mutation (duplication, insertions, deletions, …)

• Evolutionary related (homologous) proteins are similar in structure.

• There is a model to describe the evolution of proteins based on sequence alignment.

  • sequence alignment: quantify how similar two sequences are, tells you distance in evolution

• When a structure of a protein that is homologous to our target protein exists, we can model the evolutionary changes that happened since they diverged from the common ancestor.

  • evolutionary related / homologous proteins keep a similar structure even they have differences in their sequences.

• sometimes it works well

• requirements

  • good homologous protein template available

  • we need to find said protein

• find the weight of all transformations (insertions, deletions, mutations) that transforms sequence a in sequence b

  • weights are determined by scoring matrix, e.g. BLOSUM

• align the amino acids so that the weight of transformations is maximized

• “Descent from a common ancestor”

• we assume homology when biological objects have similar properties

• Homology is not the same as sequence similarity! But it can be assumed when sequences/structures are similar.

  • “45 % homology”: 45 % of the two proteins have derived from a common ancestor, 55 % from different ancestors. It does NOT mean that they necessarily share 45 % of their sequence.

  • homology: qualitative

  • sequence similarity: quantitative

• Assumption: The native protein structure/conformation is stable and located in the global free energy minimum.

• To find a native fold, the peptide chain is sampled into the native structure as well as non-native decoys.

  • anyway, not all possible conformations are tried (this would take way too long)

  • some sequences just “fall” into certain folds (e.g. α-helices) – there are libraries for that

• A potential-energy function is used to identify the native structure (the one with the lowest energy)

  • find the funnel

• We do not have the computational power to calculate all possible structures.

• We do not have sufficiently accurate potential-energy functions to identify the correct native fold.

ROSETTA combines the Boltzmann and the Darwin approach.

• You have the sequence of your protein of interest.

• You search the PDB for a homologous sequence.

• You align the two sequences.

  • important! If there are mistakes here, you get the wrong structure! Always check if the alignment makes sense.

• You create a model using the template.

• Check the initial alignment of sequences! If this is wrong, you cannot get a correct model.

• You need to validate your structure! Computational methods usually give you several alternatives, some of which will be wrong

  • globally (wrong fold/wrong orientation of domains)

  • locally (bad loops, alignment shifts)

Therefore, you should filter out wrong models and check for wrong regions in partially correct models.

• Your structures can be checked by comparison with the experimental data you have

  • residue-residue distance (disulfide bridges, chemical cross-linking, etc.)

  • solvent accessibility

  • secondary structures (NMR, CD)

  • shape and size (cryo-EM, SAXS,…)

• Also check for bond lengths, angles, stereochemistry, chemical environment of residues, etc.

• protein engineering

  • antibody design

  • implementation of new functions into known enzymes

  • de-novo enzyme design

  • stabilization of protein structures (e.g. thermostability for industry)

• reduce entropy difference between folded and unfolded proteins

• stabilize α-helix dipoles             

  • introduce negatively charged amino acid at the N-terminus (which has positive charge)

• increase hydrophobic interactions in the core

• reduce hydrophobic surface accessible to water

• introduction of hydrogen bonds, disulfide bridges, salt bridges, etc.

• Choose a catalytic mechanism.

• Create active site (quantum mechanical calculations to find optimal transition state stabilization).

• RosettaMatch: search PDB for backbone structures that could support the active site.

• Geometry optimization

• Redesign of residues surrounding the transition state.