A poster presentation on my current work in biomedicine was made for the MCBIOS IV conference at New Orleans on February 1-3. The aim was to introduce new approaches for locating DNA and peptide sequences in databases of medical journals by using Support Vector Machines (SVMs). While the project is still under development, significant improvements were made in term of speed of computation. Newly developed SVMs can be trained on sets containing up to one million instances in less than 90 seconds by using an average computer. The accuracy is nearly perfect for locating DNA substrings and still acceptable for peptide sequences. Nevertheless further accuracy improvements are under way for peptide substrings.
The reference for the poster is:
- R. C. Gilbert, T. B. Trafalis, and J. D. Wren, “Scalable methods of locating peptide sequence data within full-text articles,” in 4th Annual Conference of the MidSouth Computational Biology and Bioinformatics Society, MCBIOS IV, New Orleans, Louisiana, USA, 2007.
You can also visit the websites of the co-authors: