Theory of modelling and related techniques

WHAT IF: A molecular modeling and drug design program.
G.Vriend,
J. Mol. Graph. (1990) 8, 52-56.

This is the original WHAT IF article. Don't read it, it does not contain much science, but explains what the WHAT IF program can do. This is however the article that you should refer to if you use the WHAT IF program for general purposes. If you use it for structure validation please refer to the note in NATURE on structure validation.

Predicting local structural changes that result from point mutations.
V. de Filippis, C. Sander, G. Vriend,
Prot. Engin.. (1994) 7, 1203-1208.

This is one of the key articles for the data based modelling as implemented in WHAT IF. It describes how one extracts rotamer libraries from PDB files. This article also describes that rotamer distributions are position specific, and that this knowledge can help improve models built by homology. Rotamer libraries are implemented in WHAT IF in the DGLOOP menu.

The use of position specific rotamers in model building by homology.
G. Chinea, G. Padron, R.W.W.Hooft, C.Sander, G.Vriend,
PROTEINS (1995) 23, 415-421.

After determining that position specific rotamer distributions are a powerful tool for the prediction of side chain conformations (see the article above), this method got implemented in WHAT IF to improve the modelling module. WHAT IF does not only use position specific rotamers, but combines this information with packing quality control, hydrogen bond energy, bumps and the fit quality of the database fragment on the actual modelling template.
The use of rotamer libraries and the other information is implemented in WHAT IF in the BLDPIR option in the PIRPSQ menu.

Detection of common three-dimensional substructures in proteins.
G. Vriend, C. Sander,
PROTEINS (1991) 11, 52-58.

This article describes a method that allows you to find the best superpositioning of two protein structures. Many programs could do this already in 1991, but this WHAT IF option was the first algorithm that did not require any manual pre-alignment. The main idea is that a distance geometry algorithm is used to very quickly detect all fragments in the two proteins similar that have similar backbone conformations. Clustering these fragments gives the largest possible sub-structure between the two proteins.
Many people have written programs based on this principle. Later, Liisa Holm has written the DALI method which, I think, is the only method that works better than our method. The WHAT IF superposition method is described in chapter 15 of the writeup.

A database of protein structure families with common folding motifs.
L. Holm, C. Ouzounis, C. Sander, G. Tuparev, G. Vriend,
Protein Science (1992) 1, 1691-1698.

This article compares three methods to determine optimal protein superpositions, it describes the FSSP database and the format of the files. Nowadays only the results of DALI are stored in the FSSP database because they are normally 'the best'.

Parameter relation rows: a query system for protein structure function relationships.
G.Vriend,
Prot. Engin. (1990) 4, 221-223.

This article explains the parameter correlation method that is implemented in WHAT IF. This is an extremely simple, but fast and elegant method to get questions answered such as "Where are all the buried unsatisfied hydrogen bond donors and acceptors"?

A very fast program to visualize protein surfaces, channels and cavities.
R.Voorintholt, M.T.Kosters, G.Vegter, G.Vriend, W.G.J.Hol,
J. Mol. Graph. (1989) 7, 243-245.

This article describes the surface and cavity options in WHAT IF. The idea is that an electron desity like grid is put over the molecule. Every grid point gets a value that is a function of its distance to the nearest atom. A normal three dimensional contouring algorithm can than be used to rapidly display the cavities and surfaces. By using different contour levels one can display the results of surface calculations for probes with different radii. This article describes several options in the MAP menu in WHAT IF.

A novel search method for protein sequence-structure relations, using property profiles.
G. Vriend, C. Sander, P.F.W. Stouten,
Prot. Engin. (1994) 7/1 23-29.

This article describes a novel type of database (actually not a real database, but a fast retrieval system) specialized in protein structure queries, the SCAN3D query system. The main idea is that proteins are inherently sorted, and thus, one of the main points of relational databases (data is assumed unsorted) can be skipped. This leads to a 10 - 1000 fold query speed up in most cases. It also allows us to quickly search for a series of residues in a row, something that is very relevant in the real world. This system is implemented in WHAT IF.

How to extract non-trivial information from protein structure and sequence data.
P. Stouten, G. Vriend.
CDA News (1992) 7/5, 18-23.

This is a low level explanation of the SCAN3D system. The CDA journal aims at managers rather than practicing scientists.

Molecular docking using surface complementarity.
V.Sobolev, R.C.Wade, G.Vriend, M.Edelman,
PROTEINS (1996) 25, 120-129.

This is one of the two docking methods we are involved in. The other one is the FLEX program written in the group of T. Lengauer at the GMD in Bonn. Both with LIGIN and with FLEX, WHAT IF is used as an interface. We are only involved in discussing the scientific aspects, but not in the actual implementation. You find the interfaces in the DRUG chapter of the WHAT IF writeup. See the Additional information available on this subject.

CASP2 Molecular docking predictions with the Ligin software.
V.Sobolev, T.M.Moallem, R.C.Wade, G.Vriend, M.Edelman,
PROTEINS (1998) Suppl 1:210-214.

Vladimir Sobolev used the LIGIN docking method in the CASP2 competition, and despite the relative crudeness of the program performed reasonably well. The numbers given by Dixon in this same special volume of PROTEINS are not realistic because Vladimir gave for every experiment the best 10 solutions without puting weights on them. The judges have therefore weighed the worst solution in all cases equally heavy as the best. Also, LIGIN was used 100% automatically without helping it by telling where the active site was. Considering this, LIGIN did well. Not excellent, but really well.

Positioning hydrogen atoms by optimizing hydrogen-bond networks in protein structures.
R.W.W.Hooft, C.Sander, G.Vriend,
PROTEINS (1996) 26, 363-376.

This article describes the HB2 options in WHAT IF. In contrast to most other hydrogen bond calculation programs, this method tries to optimize the whole hydrogen bonding network of the full protein, rather than just looking around for the nearest suitable partner. Additionally, the method looks if flipping the sidechains of Asn, Gln and His residues might improve the hydrogen bonding network. The idea behind this is that in most Xray projects the difference between C, N and O atoms cannot been seen. If in such cases the local H-bond network is too complicated for the human brain to solve, the chance that the sidechain is put into the density 'the wrong way around' is considerable.All PDB files were looked at with this tecnique and the outcome is available on the WWW.

Model building of a thermolysin-like protease by mutagenesis.
F.Frigerio, I.Margarit, R.Nogarotto, G.Grandi, G.Vriend, F.Hardy, O.R.Veltman, G.Venema, V.G.H.Eijsink,
Prot. Engin. (1997) 10, 223-230.

Normally we build models to help the protein engineer. But things can also work the other way around.... This article describes how mutations can be designed to shed light on troublesome areas in a model that was built using (not enough) homology.

The PDBFINDER database: A summary of PDB, DSSP and HSSP information with added value.
R.W.W. Hooft, C. Sander and G. Vriend.,
CABIOS (1996) 12, 525-529.

This paper describes the construction of the PDBFINDER database, which contains summary information from a number of structural databases in an easily searchable form. The database itself and more information are available.

PRODRG: a program for generating molecular topologies and unique molecular descriptors from coordinates of small molecules.
D.M.F.v.Aalten, R.Bywater, J.B.C.Findlay, M.Hendlich, R.W.W.Hooft, G.Vriend
JCAMD (1996) 10, 255-262.

This paper describes a program that can be used to get WHAT IF topology entries for small molecules. The program can be used by everybody via a server. Additionally, a MOL2 file and a GROMOS topology are generated.

GPCRDB: an information system for G protein-coupled receptors.
F.Horn, J.Weare, M.W.Beukers, S.Hoersch, A.Bairoch, W.Chen, O.Edvardsen. F. Campagne, G.Vriend, NAR (1998) 26, 275-279.

Often modelbuilding by homology is not possible because of the absence of a good template structure. GPCRs are such a case. Bacteriorhodopsin is not a good template, although it was used as template many, many times. The GPCRDB project is a pilot project for class specific databases, aimed at harvesting all the heterogenous data that is available for this giant family of mediaclly important molecules.

Homollogy modelling, model and software evaluation: three related resources.
R.Rodriguez, G.Chinea, N.Lopez, T,Pons, G.Vriend
CABIOS (1998) 14: 523-528.

This article describes the WHAT IF based homology modelling server, the model validation server, and the server that compares a model with the real structure to see how well we did with the modelling.

Heterozygous Germline mutations in the p53 homolog p63 are the cause of EEC syndrome.
J.Celli, P.Duijf, B.C.J.Hamel, M.Bamshad, B.Kramer, A.P.T.Smits, R.Newbury-Ecob, R.C.M.Hennekam, G.v.Buggenhout, A.v.Haeringen, C.G.Woods, A.J.v.Essen, R.d.Waal, G.Vriend, D.A.Haber, A.Yang, F.McKeon, H.G.Brunner, H.v.Bokhoven
Cell (1999) 99, 1-20.

The EEC syndrome possibly gives people a series of extremely unpleasant problems suc as: Ectrodactyly, Ectodermal dysplasia (whatever those are), cleft lip, etc. It influences skin, teeth, hair and nails. It can also give lacrimal duct abnormalities, urogenital problems, conductive hearing loss, facial dysmorphism, etc. Brunner and Bokhoven found the gen that holds the mutations that lead to EEC. It was our contribution to model p63 based on p53. this was easy. High homology, and no indels anywhere near the residues of interest. The nice thing was that all five EEC causing mutations could easily be explained from the model. Four would hinder DNA binding, and one would prevent proper folding of the DNA binding domain. So, presently a screaning method is tested that is based on binding a piece of DNA.