Drug Design with Computers

How Do We Turn a Protein into a Medicine?

Lecture 1 — From Protein to Drug Target

"It costs on average £1.5 billion and 10–15 years to bring one drug to market. Computational biology is changing that."

You already know more than you think

You already know	We will build on this
Proteins are chains of amino acids	The sequence folds into a precise 3D shape
Enzymes have active sites	The shape of that site determines which molecules can bind
Inhibitors block enzyme activity	We can design molecules to be better inhibitors
Mutations change protein sequence	A single amino acid change can make a drug stop working

One question drives following steps

                      "A disease is caused by a protein working too well.
How do we design a molecule to switch it off?"
                

Which protein do we target?
Where exactly on that protein does the molecule need to bind?
What shape and chemistry does the molecule need to have?

COX reaction pathway

The sequence is the blueprint
the shape is the machine

"The same protein chain always folds into the same 3D shape. That shape determines everything the protein can do — including whether a drug can bind to it."

The 20 amino acids have very different personalities

Amino acid	Property	Drug design relevance
Alanine (Ala, A)	Tiny, non-polar	Used as a probe in alanine scanning — its side chain does almost nothing
Phenylalanine (Phe, F)	Large, hydrophobic aromatic ring	Likes to sit in hydrophobic pockets — often lines binding sites
Arginine (Arg, R)	Positively charged	Forms strong interactions with negatively charged drug molecules
Serine (Ser, S)	Polar, has -OH group	Forms hydrogen bonds — important for aspirin's covalent binding to COX-2

Side chains are what make each amino acid different.
It's the side chains that determine how a drug binds.

Disease often begins with a protein doing its job too well

Normal State:

Stimulus → Enzyme active → Normal amount of product → Healthy outcome

Disease State:

Overactivation → Enzyme too active → Too much product → Inflammation / infection

Example 1: COX-2 & Inflammation
COX-2 converts arachidonic acid into prostaglandins (causes pain). Overproduction = chronic inflammation. This is the target of ibuprofen (Projects 3 & 4).

Example 2: β-Lactamase & Resistance
Bacteria produce an enzyme that destroys penicillin. Natural selection favours bacteria with this enzyme (Project 6).

Not every protein is a good drug target

1. Essential to the disease process.
If you switch it off, the disease gets better. (e.g., Blocking COX-2 relieves pain).
2. Has a binding pocket the drug can physically fit into.
Many proteins are flat. Good targets have a deep, well-defined cavity (like the COX-2 hydrophobic channel).
3. Different enough from healthy human proteins.
Prevents serious side effects. Bacteria have β-Lactamase; humans don't.

Drugs work by getting in the way

Competitive Inhibition

Non-competitive Inhibition

Both types work by changing the geometry. 3D shape matters — a drug that's the wrong shape simply doesn't fit!

The Protein Data Bank (PDB)

The world's protein library

A global, open-access database of experimentally determined protein structures (>220,000 structures as of 2025). Every structure contains the 3D coordinates of every atom.

                    Grow crystals ➔ 
                    Shoot X-rays ➔ 
                    Calculate 3D map ➔ 
                    Deposit in PDB
                

What a PDB entry contains:

Unique 4-character ID (e.g., 1LYZ for Lysozyme, 4PH9 for COX-2)
3D coordinates (x, y, z) of every atom
The amino acid sequence & Resolution (Å)
Bound ligands (drugs, cofactors)

Resolution: How sharp is the picture?

1.0 Å — Every atom precise

2.0 Å — Good for side chains

3.5 Å — Backbone visible only

Resolution	Quality	What you can see
< 1.5 Å	Exceptional	Individual hydrogen atoms sometimes visible
1.5–2.5 Å	Good	All heavy atoms reliable — good for drug design
2.5–3.5 Å	Moderate	Backbone reliable, some side chain uncertainty
> 3.5 Å	Low	Use with caution — atomic positions approximate

Ångströms: atomic distance unit

1 Ångström (Å) = 0.1 nanometres = 10^-10 metres

1 mm
(Flea) 1 µm
(Bacterium) 1 nm
(Protein) 1 Å
(Bond length) 0.1 Å
(Nucleus)

A carbon–carbon bond is 1.54 Å
A hydrogen bond is 1.8–3.5 Å
The COX-2 binding channel is roughly 6–8 Å wide
The ibuprofen molecule is about 8 Å across

The Drug Discovery Pipeline

Computation compresses the 15-year timeline

Target ID
(2 yrs)

COMPUTATIONAL SCREENING
(weeks–months)

Hit ID
(1–2 yrs)

Lead Opt.
(3–4 yrs)

Clinical Trials
(6–7 yrs)

Without computational docking	With computational docking
Synthesise drug (weeks, £10,000+)	Screen digitally (hours, nearly free)
Test on cells (months)	Test only the top candidates in the lab
>90% of candidates fail at this stage	Dramatically reduces the failure rate

Your computational pipeline (Roadmap)

Project	What we do	The question it answers
1	Measure Lysozyme stability; mutate	How stable is it, which amino acids matter?
2	Compare 5 crystal structures	Which structure is highest quality?
3	Dock NSAIDs into COX-2	Which drug binds most tightly?
4	Map COX-2 binding site	Which residues touch the drug?
5	Alanine scan of binding site	Which residues are essential for binding?
6	Model resistance mutations	How do bacteria evolve to resist antibiotics?

"By the end of Project 6, you will have run the same computational pipeline that professional drug discovery teams use."

Summary & Preparation

A drug works by binding to a specific protein and disrupting its function.
The protein's 3D shape determines what can bind to it.
Good drug targets are essential, have binding pockets, and differ from healthy human proteins.
The Protein Data Bank holds >220k real 3D structures.
Resolution (Å) tells you how precisely a structure was determined.
Computational docking screens thousands of drugs in hours, saving years of work.

                    Preparation for Lecture 2:

                    "Think about this: if you unfold a protein — stretch it out into a straight line — does that require energy, or does energy get released? Why?"