Drug research at home – (how) is that possible? – Part 1

In the current day and age of open access information, combined with cheap computing power, it is rather simple to do (some) drug research from the comfort of your home, be it as private person for fun or out of interest, or as a small (start-up) company. Actually, big pharma companies use some of the same resources combined with their own in-house data and programs as well – so why shouldn’t you?  

Where is this data? What kind of data?

There are a number of public- so called open access – databases available these days, curated over many years by high profile institutes, as e.g. the National Institute of Health, NIH for Pubchem.  Many more institutions and specific initiatives have evolved over many years, some appearing literally right now, depending on the field and data. Databases on chemical compounds, small molecules, have been around the longest, afik, with structure, properties, literature references and biological data associated.

Listing all of them would require an entire Wikipedia page (or more), and that work has already been done – you can find a substantial list here for example http://oad.simmons.edu/oadwiki/Data_repositories, though in terms of life science, on this NIH site, you can really knock yourself out: https://www.ncbi.nlm.nih.gov/guide/all/#databases_. The scientific literature has regularly some article on databases and software, as well as many blogs do, but that is outside of this scope.

More focused for our purpose of drug research, you have sites such as PubChem, BindingDBZinc, or e.g. GuideToPharmacology. I’d say with these you can get pretty far.  Curated from literature and also patents, these databases connect structures to biology, i.e. mechanism of action, structure of the target, how much is know about it (or not).  All sites and db-s are arranged differently, some you can search on the web, via an API, some by browsing, or a combination thereof. Then, there are also the semi-public databases, such as CDD-Vault – you can register and search within the public databases (all via the web, independent of your machine power), though you cannot download or batch process on the free account. It might still be worth a look at times considering you find data which is not in literature/patent based curated databases.

What will you need?

A certain understanding of the drug discovery process, chemistry and some degree of biology. If not yourself, then a good friend who might have that knowledge and can support you (though this seems like a unlikely scenario?). Some IT-skills certainly don’t hurt. Below I will focus on data-mining as the core task of the home research, methods such as docking or quantum mechanic calculations I will leave out for now.

Hardware
  • A(ny) computer – Windows, Linux, Mac – doesn’t matter.
    In my experience though when it comes to chemistry, the Windows platform still offers a broader range of both commercial and freeware programs .
  • How powerful?
    Simply put, also doesn’t matter. Sure, the more power, the smoother your experience, though for mining purpose I would go for more memory before processing power. An Intel i3 with (minimum) 16GB of RAM can get you pretty far with little money. Only for large data sets and more complicated calculations I feel this being a bit of a bottleneck. If you have an i7 or Xeon available, good for you!
    What about graphic cards? That actually doesn’t matter for data-mining and simple visualizations. Once you want to do some visual 3d-docking though, that’s another story.
  • An alternative, or even complimentary solution is a (powerful) workstation, placed “anywhere”, which could e.g. be shared with someone else sharing investment costs and then access it via any (simple) PC/Laptop via remote access, e.g. TeamViewer. Cloud computing@home so to say.
  • Reasonably fast internet connection  – for mining those web-services.
Software
  • Knime (available on all platforms) allowing for flexible, visual and fast development of search and analysis workflows.  Combined with some know-how on Java or XML and you have quite a powerful package. To start your journey, you can use some of the readily available (example) workflows before getting into details.
  • A chemical drawing program – there are a rather larger number out there, it is difficult to really make a good suggestion. Knime itself comes with a “myriad” of plugins for structural input and output, thus you actually don’t really need a separate program. Myself, I do have the free Marvin package by Chemaxon installed.
  • DataWarrior – a great package for visually guided “manual” mining, sort of “Spotfire light”, if you will.
  • Excel – or similar, can be used as light weight DataWarrior alternative, but also useful for sharing or storage (as would be Word or Powerpoint (and alternatives).
  • Scripting languages – R or Python – are not necessary, though they can make a good complement, depending on your requirements.
  • Java – also not necessary, but since Knime is built on Java, it sometimes can help for certain work-arounds.
  • XML, HTML, REST – some basics might be helpful when accessing certain services via network API.

What if you don’t know Java and such? Don’t fret, initially, I for example didn’t either. If you are though a person who is more of a “learning by doing”, then the knowledge will come automatically. Obviously, you can learn these in courses as well.

Continued in part 2.