HOME- R&D
- Bryn Mawr Conference
- Hyderabad Conference
- Workshops & Training
- Program
- Drug Discovery Innovation
- Proteins
- Quantum Biochemistry
- Screening
- Web Services
- Abstracts
- Speakers
- Bender, A
- Bryant, S
- Coles, S
- Covell, D
- De Sousa, J
- Dimitropoulos, D
- Gasteiger, J
- Golovin, A
- Henrick, K
- Irwin, J
- Kel, A
- Krissinel, E
- Nicklaus, M
- Poroikov, V
- Rassokhin, D
- Schwede, T
- Pharmacophores
- Graph Mining
- Nanotech
- Membranes & Ion Channels
- Exhibition
- Registration
- Jobs
- Contact
- Schedule
|
|
|
|
|
|
|
| Joao Aires de Sousa, Universidade Nova de Lisboa |
|
| João Aires-de-Sousa was born in Lisbon, Portugal, in 1970. He studied chemistry at the New University of Lisbon, where he graduated in 1993, and obtained his Ph.D. (1997) in Organic Chemistry with A. Lobo and S. Prabhakar. His Ph.D. thesis was on 'Chiral Synthesis of N-Arylaziridines'. In 1997 he was the founder of ORGLIST - the Organic Chemistry Mailing List on the Internet, which he coordinates since then. In 1998 he joined the group of J. Gasteiger in Erlangen, Germany, as a post-doctoral fellow, and in 2002 he was appointed as an Assistant Professor at the New University of Lisbon, where he currently leads a research group in Chemoinformatics. His scientific interests include prediction of NMR spectra, representation of molecular chirality for computer prediction of observable properties, and data mining of chemical reactivity databases. His web site is at http://www.dq.fct.unl.pt/staff/jas
|
| |
|
|
A Web Interface to Neural Network Prediction of 1H NMR Chemical Shifts
João Aires-de-Sousa, Departamento de Química, CQFB and REQUIMTE, Faculdade de Ciências e Tecnologia, Universidade Nova de Lisboa, 2829-516 Monte de Caparica, Portugal
Fast and accurate predictions of 1H NMR spectra of organic compounds are highly desired for automatic structure validation, automatic structure elucidation, spectra-activity relationships, or for computer-assisted interpretation of spectra by chemists and spectroscopists.
We developed a neural networks-based system to predict 1H NMR chemical shifts of organic compounds from their molecular structures. Feed-forward neural networks were trained with 744 protons and their experimental chemical shifts. The protons were represented by a set of descriptors, selected from a pool of 120 topological, physico-chemical and geometric descriptors [1].
Additional experimental data could be integrated by means of associative neural networks [2,3] improving the accuracy of the predictions. For an independent test set of 952 protons an average error of 0.20 ppm was achieved. Such a procedure avoids retraining the networks with the new data. It can be particularly useful when predictions are desired for a specific class of compounds, and experimental data are available for related structures.
A web interface to the neural network system was built, and is currently accessible at www.dq.fct.unl.pt/spinus or www2.chemie.uni-erlangen.de/services/spinus. It incorporates a Java molecular editor from ChemAxon [4]. The output is wrapped in an HTML document, which includes automatically generated JavaScript and MDL Chime programming [5] to allow interactive visualization of the predictions.
References
1. Binev, Y.; Aires-de-Sousa, J. J. Chem. Inf. Comput. Sci. 2004, 44, 940-945.
2. Binev, Y.; Corvo, M.; Aires-de-Sousa, J. J. Chem. Inf. Comput. Sci. 2004, 44, 946-949.
3. Tetko, I. V. J. Chem. Inf. Comput. Sci., 2002, 42, 717-728.
4. http://www.chemaxon.com
5. http://www.mdl.com
|
|
|
|
|
|
|
