syswerda@bbn.com (Gilbert Syswerda) (12/05/90)
I am looking for the results of work people have done in the area of predicting the chemical and physical properties of molecules. What would be ideal is a program that could compute properties of a molecule based on its formulation. The kind of information that would be useful is hydrophobicity/philicity, electron withdrawing/donation, aliphatic/aromatic, flexibility/rigidity, planar/non-planar (and other topological features), and heterocyclic/non-heterocyclic. Properties of molecules in solution would also be very useful. The idea here is that there are large databases of molecules with known interactions (e.g. binding rate at a receptor site in the nervous system). Various techniques (e.g. ID3) could be applied to these databases to construct classification schemes, and these could then be used to predict the properties of new molecules. For this to work, a meaningful set of parameters (like those listed above) must exist for each molecule. These parameters are difficult and costly to obtain experimentally; it is usually easier to simply try the molecule to see if it has the desired effect. The parameters are also probably difficult to compute as well, but some are easy (e.g. molecular weight). At any rate, I would like to find out what can be computed today and what the general state-of-the-art is in this area.
djh@xipe.osc.edu (David Heisterberg) (12/05/90)
In article <61332@bbn.BBN.COM> syswerda@bbn.com (Gilbert Syswerda) writes: >I am looking for the results of work people have done in the area of >predicting the chemical and physical properties of molecules. You might contact some of the large molecular modeling software vendors. In their glossy propaganda they usually include references discussing the force-field model they use. From that you get names of the principals in the area, and there you go. Try SYBYL Charmm/Quanta Tripos Associates, Inc. Polygen Corp. 1699 S. Hanley Rd. Suite 303 200 Fifth Ave. St. Louis, MO 63144 Waltham, MA 02254 (314)647-1099 (617)890-2888 Insight/Discover Biosym Technologies, Inc. 10065 Barnes Canyon Rd. San Diego, CA 92121 (619)458-9990 -- David J. Heisterberg djh@osc.edu And you all know The Ohio Supercomputer Center djh@ohstpy.bitnet security Is mortals' Columbus, Ohio 43212 ohstpy::djh chiefest enemy.
fax0236@uoft02.utoledo.edu (12/06/90)
In article <61332@bbn.BBN.COM>, syswerda@bbn.com (Gilbert Syswerda) writes: > I am looking for the results of work people have done in the area of > predicting the chemical and physical properties of molecules. What would be > ideal is a program that could compute properties of a molecule based on its > formulation. The kind of information that would be useful is > hydrophobicity/philicity, electron withdrawing/donation, > aliphatic/aromatic, flexibility/rigidity, planar/non-planar (and other > topological features), and heterocyclic/non-heterocyclic. Properties of > molecules in solution would also be very useful. You don't ask much. Some of these are trivial to calculate, others don't even need to be calculated (i.e. benzene is planar, you don't have to calculate that). But almost nothing except molecular weight can be calculated from molecular formula (not formulation). The number of possible isomeric structures of any system tends to be huge. > > The idea here is that there are large databases of molecules with known > interactions (e.g. binding rate at a receptor site in the nervous system). > Various techniques (e.g. ID3) could be applied to these databases to > construct classification schemes, and these could then be used to predict > the properties of new molecules. For this to work, a meaningful set of > parameters (like those listed above) must exist for each molecule. These > parameters are difficult and costly to obtain experimentally; it is usually > easier to simply try the molecule to see if it has the desired effect. > The kinds of data bases you are thinking of are readily available in the literature (they may not be compiled lists, but certainly much of the data is known and could be found with some library work) or in corporate data bases (probably proprietary). There would be no need to get much of this data by repeating work, either experimentally or computationally. > The parameters are also probably difficult to compute as well, but some are > easy (e.g. molecular weight). At any rate, I would like to find out what > can be computed today and what the general state-of-the-art is in this area. State of the art is complex, since some things can be calculated with a high degree of precision and accuracy, some can only be estimated, and many are not unique solutions. Your best bet is to take two or three years to study computational chemistry as a graduate student or post-doctoral fellow. Even then, you will only learn a small fraction of the available methods. Sorry if this sounds harsh, but you seem to have a vague idea of what might be an interesting approach (although one that is certainly being pursued by others right now) but you don't know enough yet to formulate the question properly. One person I know of who is doing a lot of this type of work is Dr. George Famini at Aberdeen (the Army, for God's sakes!). His work involves using known physico-chemical data to develop models that allow predictions of activity for many types of compounds. I can provide his e-mail address if you write to me. Doug Smith Assistant Professor of Chemistry The University of Toledo Toledo, OH 43606 FAX0236@UOFT02.UTOLEDO.EDU