利用LINCS L1000資料中藥物擾動及小髮夾RNA對細胞之影響預測藥物與標的之交互作用

Abstract

一顆藥物從成千上萬的化合物中被篩選出來到最後審查通過上市需要耗費龐大的資金及平均至少十年以上的時間，如何加速整個新藥開發的過程一直是醫學上很迫切的問題。不論是透過提升前期藥物研發效率或是針對已上市的藥物找到別的可能應用的疾病，亦即所謂老藥新用都可有效加速整體藥物發展的過程。在本篇研究中我們建立了一套預測分析流程，透過分析藥物擾動以及小髮夾核糖核酸擾動所產生基因表現資料之間的關聯去預測藥物可能的作用標的，以期能加速藥物研發時確認藥物作用標的及探討作用機制的過程。透過利用已知藥物及其作用標的的資料進行驗證，本方法在接收者操作特徵曲線下面積的指標達到0.71的結果，展現出其在預測藥物可能作用標的的潛力，此外，我們也將此方法預測出的結果透過功能分析，進一步應用到老藥新用上，並舉癌症及免疫疾病為例，對於已知藥物對此兩疾病的重要機制進行篩選，找出我們認為有可能拿來應用的候選藥物，而在預測的潛力藥物結果中，亦有許多得到實驗研究證實且已進入臨床試驗階段，另一方面展現此方法在老藥新用上的應用性。

The journey of a drug, from being selected in the laboratory to finally be sold on the market, is tedious, money-consuming and full of risks. It is an urgent need to shorten the process of drug discovery and development. Either accelerating the initial phase – drug discovery or repurposing existing drugs for new indications could be beneficial to achieve the goal. In this study, we have developed an analysis pipeline for predicting potential targets of drugs based on only perturbational profiles in L1000 data. Through analyzing the associations between compounds and short hairpin RNAs (shRNAs), the potential drug-target interactions could be inferred. The performance of the prediction results was evaluated by a known dataset extracted from Drug Repurposing Hub, and an average area under the receiver operating characteristic curve (AUC) of 0.71 has been achieved. Finally, we further applied our approach to explore opportunities for drug repurposing in cancer and inflammatory diseases through functional analysis. Several putative anti-cancer drugs and anti-inflammatory drugs revealed from the prediction are supported by preclinical or clinical studies, which demonstrated the efficacy of the proposed approach.