Research topics

 

My research interest is in the development of statistical machine learning methods for analyzing real-world data. My previous research mainly focused on bio/chem data, but my current/future research is not limited to these areas. If you are interested in working with me, contact me by the following mail address.

 

saigo@inf.kyushu-u.ac.jp

 

Topic 1: Prediction of protein functions

 

All the living things including humans consists of proteins, and all the proteins consist of amino acid sequences. In the era of Next Gen Sequencing, there are many amino acid sequences available in the databases, but many of them are left annotated.

In my research, I have developed a kernel function that can capture similarity among amino acid sequences based on well studied amino acid substitution matrix and sequence alignment.

http://www.i.kyushu-u.ac.jp/~saigo/research/SaigoResearch_html_m2f9d014.gif

 

Topic 2: QSAR (Quantitative Structure Activity Relationship)

 

QSAR (a.k.a. ligand based virtual screening) is one of steps in developing drugs in pharmaceutical companies. In order to decrease the number of candidates subjected to bio/chem experiments, computational methods makes an important role.

In my research, I am aiming at developing prediction methods that can not only equipped with high prediction accuracy, but also equipped with interpretability to humans. Since in drug development processes, it is important to work with bio/chem scientists, and convince them with explainable prediction model.

 

http://www.i.kyushu-u.ac.jp/~saigo/research/SaigoResearch_html_6e575ebe.gif

 

Topic 3: HIV drug resistance prediction

 

The latest treatment of HIV/AIDS reads the sequence of HIV viruses in a patient. Each virus has a different genetic information, and therefore behave differently against the given drugs. Therefore it is important to know the genetic information of viruses, then predict the amount of drug resistance prior to selection of a therapy.

In my research, I have developed a prediction method that can consider multiple mutations in the amino acid sequence of a HIV-1 virus. Consideration of multiple mutation turned out to be able to model the increase of drug resistance by the accumulation of mutations around the active site of HIV-1 protein.

 

http://www.i.kyushu-u.ac.jp/~saigo/research/SaigoResearch_html_42cd4903.gif