This site is maintained by John H. Van Drie, as an open-source resource for all things
related to pharmacophores.
Open-source pharmacophore-based 3D Database searching
Rajarshi Guha (Indiana University) and I have recently made available as open-source some pharmacophore search software. Look at
Rajarshi's page to download that Java software, and get some
cryptic documentation. Contact me for more info. This is brand-new, and is evolving, but it seems to work. It's all command-line driven
at the moment, and functions a lot like the old ALADDIN code (Van Drie, Weininger & Martin, JCAMD, 1989, 3:255).
This new open-source stuff is based on the open-source
Chemistry Development Kit (CDK).
Building 3D databases with public or open-source software
A number of options are now available beyond the offerings from the commercial vendors:
- Rajarshi Guha's SMI23D - generates a single reasonable conformer per molecule
- Mikko Vainio's Balloon - generates a set of conformers per molecule
- RDKIT.ORG from Greg Landrum et al - generates a set of conformers per molecule
Note too that PubChem is now available in 3D format, PubChem3D
and available for download.
Pharmacophore discovery
These methodologies are also called 'pharmacophore identification', 'pharmacophore perception', or 'receptor mapping'. The first commercial software
for pharmacophore discovery was the Catalyst software, of which
I was one of the co-authors. (DISCO is often cited as the first, but that appeared from Abbott only after they'd seen initial versions
of Catalyst. DISCO did indeed appear in the literature before Catalyst). Catalyst's approach to pharmacophore discovery has
numerous weaknesses, which I've addressed in my latest work on pharmacophore discovery, DANTE, which is described in the following papers.
- J.H.VanDrie, "An inequality for 3D database searching and its use in evaluating the treatment of conformational flexibility", J. Comp.-Aided Mol. Design, 1996, 10, 623.
- J.H.VanDrie, "Strategies for the determination of pharmacophoric 3D database queries", J. Comp.-Aided Mol. Design, 1997, 11, 39.
- J.H.VanDrie "'Shrink-wrap' surfaces: A new method for incorporating shape into pharmacophoric 3D database searching", J. Chem. Inf. and Comp. Sci., 1997, 37, 38.
- J.H.Van Drie and R. A. Nugent, "Addressing the challenges of combinatorial chemistry: 3D databases, pharmacophore recognition and beyond ", SAR and QSAR in Env. Res, 1998, 9, 1-21.
I recently published a review in Curr Pharm Design, 2003, 9(20):1649, and also have a book chapter reviewing pharmacophore discovery:
Computational Medicinal Chemistry and Drug Discovery, Tollenaere Bultinck de Winter and Langenaeker (eds.), NY: Dekker.
Here is a script and a small C program which one can use with the commercial Catalyst software
to implement the Mayer et al method, the algorithm at the heart of DANTE:
Frequently-asked questions (FAQ's) about the practical issues of constructing pharmacophores
There is a tremendous amount of confusion surrounding many issues surrounding pharmacophores and their use. I would like to clarify some
of these here. Many thanks to Helene Decornez of AMRI for constructing an initial list of questions, which formed the basis for these FAQ's.
During my time at Abbott (where I developed ALADDIN, the progenitor to all pharmacophore search systems),
at BioCAD (where we developed Catalyst, now marketed by Accelrys), and at Upjohn, I probably built over one
hundred pharmacophores, and have learned a number of practical tips on how to build and use pharmacophores.
Twenty years ago, one built pharmacophores manually, and many of these were good pharmacophores (see the first 5 volumes of JCAMD for such
examples). Now, people mainly rely on the low-quality automated pharmacophore generation tools commercially available; as a result,
almost an entire generation of CADD scientists have never even seen a good pharmacophore.
- What is the definition of a pharmacophore? (Many people view this as an ill-defined and fuzzy concept)
- There is an official IUPAC definition of a pharmacophore
(IUPAC = Int'l Union of Pure and Applied Chemistry)
- I tend to put it more simply: a pharmacophore is a spatial arrangement of functional groups essential for biological activity.
It is a pattern that emerges from a set of molecules with a common biological activity.
- Do pharmacophores work? In other words, can one really find novel biologically active molecules searching a
3D database using a pharmacophore?
- Yes, emphatically. A number of important drug candidates originated from leads discovered via pharmacophoric 3D database
searching. Our initial experiences at Abbott in the discovery of a novel D1 agonist lead were described by
Y. C. Martin in J Med Chem in 1992. The first marketed drug which
originated in a lead found by pharmacophoric 3D database
searching is Merck's Aggrastat, approved by the FDA in 1999; that lead discovery was
described coyly here
A number of reviews are available describing the successes of pharmacophoric 3D database searching, including my chapter in the Alvarez and Shoichet
book Virtual Screening.
- Dataset selection
- Number of structures
- One can use as few as 2-3 compounds, or have as many as thousands.
- Activity range
- The most important thing is to have a few extremely potent molecules (< 100 nM). A few single-digit nM compounds convey an
enormous amount of information.
- Congeneric series vs. diverse structures
- It's best to have many diverse, distinct structures. But, with luck, one can find a pharmacophore even given a congeneric series.
- Diversity of features (number of key points, acc, donor, arom, etc.)
- Obviously, it's better to have feature-rich mol's, as opposed to anthracene analogs.
- Are there good or bad mixing of features?
- There is no such thing. If the molecules' activity is real, then whatever features are necessary to elicit
biological activity is what should appear.
- Pharmacophore generation
- optimum number of features
- Most good pharmacophores have 2-4 features. I've never seen a good example requiring > 4, and often find good 'dyad pharmacophores'
(2-feature pharmacophores)
- combination of features
- Criteria for a good pharmacophore
- selectivity vs. known inactives
- This can vary greatly. The important thing is the selectivity relative to a database of drug-like mol's (see my 1997 JCAMD paper)
- Selectivity vs. random library of cmpds
- Again, see my 1997 JCAMD paper for info on the 'principle of selectivity'
- Impact of a min # of confs
- Conformational analysis is key to getting good pharmacophores. Most people do not appreciate how sensitive pharmacophore
discovery is to the quality of the conformational analysis. The total number of conf's is less critical, and depends highly on the
flexibility of the molecule. I usually prefer Macromodel's MCMM, with a target number in the range of 200 confs.
- Tweaking pharmacophores
- What steps should be taken (tighten radius of features? Chg input confs?
- Don't tweak - use the Mayer et al algorithm, cited in my 1997 JCAMD paper.
- Application of excl vol's (guessed from known SAR? from Xray or homol model?)
- If you have an Xray struc or homol model, use it (see, eg, paper of P. Greenidge et al, J Med Chem, 1998).
- To determine it from the known SAR, use the 'shrink-wrap' algorithm (JHVD, JCICS, 1997).
- How to make sure we are not making a bad model say what we want?
- See Guha and Van Drie, JCIM, 2008, 48:1716.
- automated pharmacophore discovery. These questions are answered in my 2003 Curr Pharm Design paper.
- key factors to monitor?
- what features are best hand-generated?
Who was the originator of the concept of a pharmacophore, and when did this occur?
- Monty Kier, in a series of papers 1968-1971, introduced the concept of a pharmacophore. See
my article in the IEJMD, an issue dedicated to Monty Kier. Garland Marshall
did more than anyone to propagate the concept of a pharmacophore, via his Active Analog Approach, embodied in the early Tripos software. The concept of
a pharmacophore is widely but
erroneously believed to originate from Ehrlich in the early 20th century, a mistake due to Ariens, propagated by P. Gund, and further
propagated by O. Guener, sloppy scholarship that now has reached Wikipedia.
3D database searching
I developed the ALADDIN software, collaborating with Yvonne Martin at Abbott Labs. That pioneered a new approach to 3D database searching,
and was used in the first success for virtual screening, the discovery of a novel template for D1 agonists.
Link to PubMed citations
Contact me
Those interested in anything related to matters above:
send me an e-mail and let's initiate a dialogue.
Definition of the term pharmacophore
The material that was here for many years has now been moved to a refereed journal,
in a 2007 article by me
in the IEJMD, an issue dedicated to Monty Kier.