I wrote pharma software for a few years focusing on small molecule drug discovery, and did academic research on structural based drug discovery of allosterics.
Different companies do it differently, but the customers I had went through a series of steps. I'll try to summarize them briefly here.
1. Target choice: Typically research focuses on a single "target". In small molecules, this is typically a protein that you want to activate somehow to change a biological pathway. For other sub-fields, such as biologics, "target" can mean other things.
2. Iteration / core choice: You start with a small organic molecule "core" that can be the backbone of your drug. Molecules with similar cores can "fit" into similar places, whereas active groups that branch off of the core affect their efficacy.
3. Screening / lead optimization: Pharma companies have libraries of chemical structures, their properties, and how they fit into certain proteins. The idea is to find a molecule with a core that will fit, and active groups that will make the drug 'plug' the active site of the protein. When doing drug discovery, there are all sorts properties you care about, ranging from pka to toxicity.
- For many drugs, there is experimental data that is available. For other drugs, experiments must be performed to obtain data. At this stage it basically becomes a data problem for the chemist. Lots of time looking at pivot tables of raw experiment data, 2d molecule structures, 3d structures, computational predictive results, etc. The idea is to to find holes in the data where potential molecules with the right core and active groups could have matching properties, etc.
- A lot of it is data based, but a lot of it is also intuitive.
- There are other sources of data other than assays (experiments). One of the most useful is X-Ray crystallography images, which are 3D images of molecules in active sites.
4. Discovery Candidate: At some point (hopefully) you reach the stage where you're pretty sure a drug has good properties. This is when you lock down the IP, not just by patenting that molecule, but all other molecules that are similar. Basically "putting it in a box" to start the more expensive animal / human testing.
5. POST-IND FDA approval process. You can read about this here [1], but basically this is where the expensive / time consuming fun begins. Animal tests, efficacy tests on humans, etc. This is pretty expensive but also well defined by the FDA
Note that I am far from an expert, some of the people I worked with had done this for decades and would probably correct me and / or point out steps I'm missing.
