Finding side effects of drugs, before reaching the animal and clinical test, can decrease the cost and time of developing new drugs. Also, a lot of side effects are reported after going to market and using it by patients (post-marketing). There are some systems to report these events but most of the time, side effects stay hidden. Thus, a lot of computational methods have been developed to predict the side effects of drugs before reaching the clinical test or going to market. These methods are based on different features of drugs like chemical or biological features. Most of these methods can just complete the side effect list of approved drugs or require a great deal of biological information about the drugs. So, it is impossible to find the side effects of a new drug candidate very early. In this paper, we propose a network and domain-based algorithm (NDDSA) that can predict the side effects of new drug candidates by assigning a score to each side effect. Unlike other methods, it needs just the chemical formula of the new drug. We show that NDDSA performs better than the best present method in case of predicting ADRs for a new drug and has acceptable results in case of predicting potential side effects for approved drugs, while it is much faster than the other methods.

在进行动物和临床试验之前发现药物的副作用，可以减少开发新药物的成本和时间。而且，据报道，上市后患者使用了很多副作用（上市后）。有一些系统可以报告这些事件，但是在大多数情况下，副作用是隐藏的。因此，已经开发出许多计算方法来预测药物的副作用，然后再进行临床测试或推向市场。这些方法基于药物的不同特征，例如化学或生物学特征。这些方法中的大多数只能完成已批准药物的副作用列表，或者需要大量有关药物的生物学信息。因此，不可能尽早发现新药候选者的副作用。在本文中，我们提出了一种基于网络和域的算法（NDDSA），该算法可以通过为每个副作用分配分数来预测新药候选者的副作用。与其他方法不同，它仅需要新药的化学式。我们表明，在预测新药的ADR时，NDDSA的效果要比目前最好的方法好，而在预测已批准药物的潜在副作用的情况下，NDDSA的效果可以接受，而比其他方法要快得多。