研發人體免疫缺失病毒治療的最佳化藥物配方篩選系統

Abstract

本計畫為「研發人體免疫缺失病毒治療的最佳化藥物配方篩選系統」之一年 期研究計畫，主要目的是發展一套可依據病人生理特性，協助醫生決定使用人體 免疫缺失病毒(HIV)治療的最佳藥物配方之電腦篩選系統。本篩選系統需建立三 個藥物治療模型，分別是雞尾酒療法、進入抑制劑及成熟抑制劑。藥物治療模型 之研發程序分為三個重要步驟：(1)蒐集並建立病毒序列和病人生理表現之資料 庫，以便於擷取作為HIV 個人化醫療的生物醫學特徵；(2)整合HIV 病毒序列特 性與生物醫學特徵來設計藥物治療模型，並將模型設計轉換成數個參數最佳化問 題；(3)發展一套以高效能繼承式雙目標基因演算法(IBCGA)作為核心的建模與預 測演算法。 本最佳化藥物配方篩選系統主要分為三個子系統，各子系統間相輔相成，簡 述如下： 1. 病毒序列及病患生理現象的資料庫，並將相對應之基因型和生理型資料整 合。 2. 雞尾酒療法、進入抑制劑及成熟抑制劑等三個整合的藥物治療模型。 3. 個人化醫療的決策系統，藉由利用HIV 病毒與生理現象資料庫及三個藥物治 療模型而來。 我們結合IBCGA 與支持向量機分類器，已建立進入抑制劑之藥物治療的初步 模型與病毒序列資料庫，並驗證此模型能有效地進行藥物篩選。此成果與經驗可 幫助另二個藥物篩選模型的建立，進一步加速整合篩選系統的建立與驗證。

This is a one-year project: Developing an Optimized Recipe Selection System of HIV Therapy. The objective is to develop a recipe selection system of HIV therapy according to physiological properties of patients to help doctors decide the medicine usage. The recipe selection system needs to establish three medicine therapy models, HAART, entry inhibitor, and maturation inhibitor. The procedure of developing the therapy models consists of the three important steps: 1) establish databases of HIV sequences and physiological properties of patients to extract potentially medical features for individualized HIV therapy; 2) integrate characteristics of HIV sequences and medical features to design the therapy models, and transform the model design into a number of parameter optimization problems, and 3) develop a set of modeling and prediction algorithms based on a high-performance inheritable bi-objective optimization genetic algorithm (IBCGA) to solve these optimization problems. The optimized recipe selection system of HIV therapy mainly consists of three subsystems which are fully cooperated, described below. 1) The databases of integrating HIV sequences and physiological properties of patients. 2) Three integrated therapy models of HAART, entry inhibitor, and maturation inhibitor. 3) The decision system of individualized HIV therapy by utilizing the three therapy models and databases of HIV and physiological properties. We have established a prototype of the ‘entry inhibitor’ therapy model and a database of HIV sequences based on IBCGA and a support vector machine based classifier. The prototype model has been verified that can screen out effective recipes. The achievement and experience can help we design and establish the other two therapy models of HAART and maturation inhibitor. Therefore, the optimized recipe selection system can be fast established and validated.