Antibody Modelling Flashcards
What’s the situation with sequence and structural data?
We have loads of sequence data but not that much structural data. The theory is that sequence determines structure, so we should be able to use sequences to predict structures.
Why are predictions useful?
They are useful because:
- Help us understand the sequence-structure gap
- The structural information can tell us about the function
- It can guide rational drug design
- It can guide mutagenesis studies
- It helps to solve experimental structures
- It focuses on fundamental chemistry of protein structures
How can protein similarity be measured?
- Sequence identity - compares the sequences
- Root mean squared (RMS) - compares structures as they are superimposed and the average distance between 2 equivalent points is calculated. 2.5Angstrom is good, lower is better.
What is CASP?
Critical Assessment of Protein Structure
This allows the blind trials of protein prediction software so they can be truly assessed with no bias and compared for other users
Describe the 2 ab-initio energy calculation methods for structure prediction.
Ab-initio methods use calculations and simulations to determine structure. They were not the most successful methods at protein prediction.
Energy Minimisation - The atoms are described in terms of bond length, bond angle, bond dihedral rotations and interactions. The conformation with the lowest energy is then searched for. Hydrophobicity terms need to be added and structures can get stuck in false energy minimums.
Molecular Dynamics - The atoms are described in terms of bond length, bond angle, bond dihedral rotations and interactions. Newtons laws of motion are solved over time which allows jumps over energy barriers to find the lowest energy conformation.
How does secondary structure prediction work?
Based on the idea that local sequences determine local structure. Programs aim to predict secondary structure elements: alpha-helices, beta-strands and coils.
How can you measure the accuracy of structure prediction?
Accuracy (Q3 - because its a 3-state model) = no. of residues correctly predicted/total no. of residues considered.
For a typical protein with an average mix of alpha-helices, beta-strands and coils Q3 = 40%
How does the Chou-Fasman principle work?
This is an early method that gives residues scores and then adds additional rules to determine secondary structure elements. It’s based on the idea that amino acids have a preference for certain secondary structure elements.
Propensity of (e.g.alanine in a helix) = (no.of alanines in helices/no.of alanines in the database)/(no. of amino acids in helices/ no. of amino acids in database)
Propensity = 1 - average amino acid
Propensity > 1 - indicates a preference
Propensity < 1 - indicates a dislike
Describe stereochemical methods for secondary structure prediction.
This recognises the hydrophobicity of residues and the way they favour particular secondary structures
e.g. hydrophobic residues are in the core, polar or charged residues are exposed.
This method produces 60% accuracy.
Describe the use of artificial neural networks for structure prediction.
Artificial neural networks and machine learning is used as you can give input sequences and output structures, the machine will learn the weights for signals and will try different architectures so that it can predict outputs from inputs.
Why has secondary structure improved?
Better algorithms
more structural data
more sequence data
What is template-based modelling and what are some of the other terms used to describe the process?
Template-based modelling uses the basis that homologous proteins have similar structures so using these as templates is the best way to predict 3D structures. This process is also called: -homology modelling -threading -fold recognition -comparative modelling
Describe the 6 steps of template-based modelling.
- Find the template sequence - compare query sequences and predicted secondary structures to the database looking for at least 15% identity. MSAs are used if possible.
- Align the sequences - Multiple structures allow gaps and loops to be predicted.
- Substitute the sequence for structure- if residues match replace them, if there is a clear substitution use the backbone as a template, if there is an indel put it in a suitable space, depending on the size of it.
- Identify and model loop regions - If there’s a conserved backbone this is fairly easily however if it is not conserved search the database for similar sequences and orientations then use these as a template. There are various algorithms for this, short loops are predicted better.
- Add side chains - Use rotamer libraries and probabilities to ensure allowed chi angles and no steric clashes. Consder the packing and formation of H-bonds, disulphides etc
- Model refinement - Use energy calculations to minimise the structure, but don’t minimise too much.
Describe Phyre2.
A webserver that carries out template-based homology. -input the query sequence -create an MSA -predict the secondary structure -make a HMM using secondary structure and MSA -produce an alignment using the HMM -build the backbone based on this information -build loops -add side chains -give an accuracy/confidence score PhyreAlarm --> checks for updates BackPhyre --> structure to sequence
How do monoclonal antibodies work and what are some of their modes of action?
They recognise and alter the activity of specific proteins. They can be used as a therapeutic method e.g. cancer treatment.
- Direct cytotoxicity - the binding of the antibody leads to the destruction of the cell
- Immune Modulation - It targets cytokines to cells which leads to cell death
- Pretargetting - It directs drugs or radioactive molecules to cells via binding
