Abstracts Keynote & Plenary
Abstracts Keynote & Plenary
Abstracts Keynote & Plenary
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Profile-profile alignment may be the most sensitive and useful computational resource for identifying<br />
remote homologies and recognizing protein folds. However, profile-profile alignment is usually much<br />
more sophisticated on algorithm and slower on time than sequence-sequence or profile-sequence<br />
alignment. The profile or PSSM (position-specific scoring matrix) can be used to represent the<br />
mutational variability at each sequence position of a protein by using a vector of amino acid<br />
substitution frequencies and it is a much richer encoding of protein sequences. Consensus sequence,<br />
can be considered as the simplified profile, was used early to improve sequence alignment. Recently,<br />
several studies were carried on to improve Psiblast remotely related protein identification performance<br />
by using the alignment between Psiblast profile and consensus sequences (profile-consensus alignment).<br />
There are several ways can be used to compute consensus residues at each position of a sequence<br />
which capture different information of a profile. Based on this observation, we propose a method that<br />
combined the information of different type of consensus sequences and profiles<br />
with the assistance of support vector machine learning, and results suggest that our<br />
method can further<br />
improve Psiblast fold recognition performance. In addition, we also compared the fold recognition<br />
ability of our method with COMPASS.<br />
Keywords: Psiblast, consensus sequence, profile, SVM<br />
OR-014<br />
Human Oral<br />
Bioavailability Prediction of Four Kinds of Drugs<br />
Aixia Yan Zhi Wang, Meng Meng<br />
Corresponding author phone: +86-10-64421335;<br />
fax: +86-10-64416428;<br />
E-mail: aixia_yan@yahoo.com or yanax@mail.buct.edu.cn<br />
State Key Laboratory of Chemical Resource Engineering, Department of Pharmaceutical Engineering,<br />
P.O. Box 53, Beijing University of Chemical Technology, 15 BeiSanHuan East Road, Beijing 100029,<br />
P. R. China.<br />
In this work, four<br />
quantitative bioavailability prediction models were built for four kinds of drugs using<br />
MLR (Multiple Linear Regression). For the model of Angiotensin Converting Enzyme Inhibitors or<br />
Angiotensin Ⅱ Receptor Antagonists, correlation coefficient r=0.91, MAE (Mean absolute error)<br />
=5.97; for the model of Calcium Channel Blockers, r=0.98, MAE =2.93; for the model of Sodium and<br />
Potasium Channels Blockers, r=0.97, MAE =6.29; and for the model of Quinolone Antimicrobial<br />
Agents, r=0.91, MAE =6.72. Explorations into subsets of compounds were performed and good<br />
quantitative relationship can be built for these four kinds of drugs which were considered have same<br />
pharmacological activity.<br />
Keywords: Human Oral Bioavailability,<br />
Multiple Linear Regression (MLR), Absorption, Distribution,<br />
Metabolism, and Excretion (ADME)<br />
OR-015<br />
Prediction of directly regulated genes of transcription factors important for ascidian early<br />
development<br />
Xuyang Yuan1, 2, Atsushi Kubo3, Yutaka Satou3, Kenta Nakai1, 2, 4<br />
1. Department of Computer Science, Graduate School of Information<br />
Science and Technology,<br />
University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan.<br />
2. Human Genome Center, Institute of Medical Science, University of Tokyo, 4-6-1 Shirokanedai,<br />
Minato-ku, Tokyo 108-8639, Japan.<br />
3. Department of Zoology, Graduate School<br />
of Science, Kyoto University, Kitashirakawa -Oiwake-cho,<br />
Sakyo-ku, Kyoto 606-8502, Japan.