Lecture 4 Construct

Question 1

How are protein constructs constructed and why

Answer 1

AI:
- Can make a predicated model using Alphafold 3, roseTTAfold, Boltz-1, chai-1

You do this to make a predicated modfied version of the protien that is able to crystallize better than the regular unmodified protien

If you used affinity rage to purify the protein you crystallized, do they need to be removed

Also if you want to design the nucleic acids in a protien-nucleic acid complex (length of the nucleic acids are important for crystallization

Question 2

Describe alphafold

How is it trained

Answer 2

Predicts structure of protien from their sequences

Works better with 30-50 related seqeunces because residues that covary are close in the 3D structure:
- covary meaning have asp and arg next to each other and interacting, if asp mutated then arg changes to keep the interaction with asp in 3D
- having many sequences that are related can show patterns of covariation

Trained:
- it uses the many protein structures in databank and the protein sequences
- does multiple sequence alignment

Question 3

Explain alphafold 3 process

Answer 3

Give it seqeunces ligands or covalent bonds (not just sequence)

It does:
- template search: looks at models
- genetic search: does MSA
- conformer generation: forms bits of the protien structure

Question 4

What is alphafold pLDDT

Answer 4

Chatgpt doesn’t give this

The predicted local difference test :
- Gives a per-residue confidence score of 0-100 in the local structure

Catergories:

Darker blue:
- expected to be accurate (>90 backbone and side chain atoms are typically predicted well)

Light blue:
- expected to be to have good backbone predication (>70 backbone atoms are well predicted but some side chains are misplaced).

Yellow:
- low confidence

Orange:
- ribbon like appearance, 3D coordinated should not be interpreted, correlated with disorder

Question 5

What is alpha fold predicated aligned error(PAE)

Answer 5

Measures the confidence of the relative position of two residues in the predicted model

Question 6

Explain the PAE plot

Answer 6

ASSESING RELATIVE POSTION TO EACH OTHER

Y axis aligned residue: # of the AA residue in the predicted model that is also aligned in the true structure
X axis scored residue: the residue in the model that

Colour at position x,y shows the expected position error of the predicted residue x when compared to the predicted and aligned to the true residue y

Postion error: darker means less error

dark diagonal line:
- if x postion is correct in relation to y, the expected postion error is low and dark green along the diagonal line

Blocks:
- blocks of dark green represent domains of the structure
- block of light green are less well predicted like linker regions or loops

Question 7

What are the limitations to the alphafold models

Answer 7

Only the protien is modelled, there is no information on the environment (like a membrane):
- unless you specifically include these things (like covalent modifications of protiens, add in DNA, RNA, ligands like ADP,NAD,ions)
- DNA it can do good, not RNA though

The information it’s give is only based off the data in the PDB and PDB has:
- only statics structures
- no disordered proteins
- includes both good and poor geometry
- AF3 is biased toward conformations that it has already seen (won’t tell us something new)

Question 8

How can AF model be used as hyposthsis for crystallization

Answer 8

Want a homogenous protien, don’t want disorder that changes conformation in the protein, they also might not be packed into the crystal:

• It can cut off disordered regions (like n-term, c-term, loops)
• more disorder, lower confidence of the prediction
• Easy to cut off at c term and n term

It can also crystallize individual domains:
- if don’t know the orientation of one domain in the protien in relation to the other
- use the PAE to select domain boundaries and make more than one truncation at each end of the domain
- but don’t want to cut where interactions are occuring (like h bond with backbone at a certain residue)

Predict the structure with PTMs and ligands:
- could help decide if you should add ligands or PTM for the crystallization

Question 9

What are affinity tags in crystallization

Answer 9

Tag added to protien for purification usually at n or c term connected by a linker

Ex.
His6

Shorter tags are left on the protiens, longer tags are cleaved off via the cleavage site in the protien (ex. TEV)

Question 10

example of nucleic acids helping crystallization

Answer 10

they bound ssDNA to the protien whcih was self complimentary and the 5’ end which allow the crystal to form easier and make a packing region (allows the protiens to co crystallize in one unit cell)