[读论文]Modeling polypharmacy side effect

本文是2020.12.02的笔记，
于2022.03.08从本人onenote迁移到CSDN

相关连接

论文网站： https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6022705/#
项目网站：http://snap.stanford.edu/decagon/
github： https://github.com/mims-harvard/decagon

论文内容

1. Introduction
   多药合用（复方用药，polypharmacy）效果好。
   但增加了副作用的风险。又很难预测副作用。很难每种药排列组合都做实验。
   前人做了一些研究，但他们研究的是“某些药一起，会不会有副作用”
   但更重要的是，“某些药一起，会有哪些副作用？” （即本文研究的内容）

   我们把问题建模为multi modal graph, 节点是protein or drug， 其中drug drug interaction都有label标明副作用类型。
   我们的motivation来自于，首先做了exploratory analysis，发现两件事儿： 一起开的药比随机的要有更多共同靶蛋白，把蛋白蛋白相互作用纳入模型，对药药副作用的预测很重要。
   前人做的类似的预测，一般是一个graph feature extraction, 一个 edge prediciton , 分开训练。
   我们的优点是：准确率最高能高69%，平均20%； 他们只预测有没有副作用，我们预测具体是哪种副作用。

2. Dataset：
   来自网络的公开数据集，包含protein protein, protein drug, drug drug , drug alone.

3. Data driven motivation
   观察生成的图（网络），发现三点：
   a. （有点像二八定律）>53%的副作用种类发生在<3%的药物组合中；而最常见副作用的发生要高出一个数量级。（所以这种少见副作用的预测很难的，要建立能够share information的model）
   b. 同一药物组合的多种副作用常常同时出现（由统计学方法得知）。（所以要利用副作用之间的依赖关系）
   c. 考察了drug pair 和他们有没有共同靶蛋白，有没有副作用之间的关系，发现三点。作者总结认为靶蛋白这个事儿很重要。
   i. 68%的药物组合其实没有共同靶蛋白，所以要考察蛋白蛋白关系，才能把他们连接起来，不然他们连不起来呀。
   ii. random的药物组合有共同靶蛋白的概率比常常一起开的要要少。
   iii. 不通副作用的药物组合有没有共同靶蛋白的概率还不太一样。

4. 网络 Decagon
   a. A encoder: a GCN to produce embeddings for nodes in G
   输入：graph G， node feature vectors xi (nx 矩阵）
   输出： d维embedding for each node. (nd 矩阵)
   b. A decoder: a tensor factorization model using embeddings to predict edge existance and type

网络结构

loss是交叉熵函数

上图摘自论文。上半部分是图一，原文说明如下：

Fig. 1.
An example graph of polypharmacy side effects derived from genomic and patient population data. A multimodal graph consists of protein–protein interactions, drug–protein targets and drug–drug interactions encoded by 964 different polypharmacy side effects (i.e. edge types ri, i = 1, …, 964). Side information is integrated into the model in the form of additional protein and drug feature vectors. Highlighted network neighbors of Ciprofloxacin (node C) indicate this drug targets four proteins and interacts with three other drugs. The graph encodes information that Ciprofloxacin (node C) taken together with Doxycycline (node D) or with Simvastatin (node S) increases the risk of bradycardia side effect (side effect type r2), and its combination with Mupirocin (M) increases the risk of gastrointestinal bleed side effect r1. We use the graph representation to develop Decagon, a graph convolutional neural model of polypharmacy side effects. Decagon predicts associations between pairs of drugs and side effects (shown in red) with the goal of identifying side effects, which cannot be attributed to either individual drug in the pair

来自 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6022705/#

下半部分是图三，原文如下：
Fig. 3.
Overview of Decagon model architecture.
(A) An Decagon encoder. Shown is a per-layer update for a single graph node (a drug node representing Ciprofloxacin based on the small example input graph in Fig. 1). Hidden state activations from neighboring nodes Ncr are gathered and then transformed for each relation type r individually (i.e. gastrointestinal bleed, bradycardia and drug target relation). The resulting representation is accumulated in a (normalized) sum and passed through a non-linear activation function (i.e. ReLU) to produce hidden state of node vc in the (k + 1)-th layer, h(k+1)c. This per-node update is computed in parallel with shared parameters across the whole graph.
(B) For every relation, Decagon decoder takes pairs of embeddings (e.g. hidden node representations zc and zs representing Ciprofloxacin and Simvastatin) and produces a score for every (potential) edge in the graph. Shown is the decoder for poypharmacy side effects relation types. © A batch of neural networks that compute embeddings of six drug nodes in the input graph. In Decagon, neural networks differ from node to node but they all share the same set of relation-specific trainable parameters [i.e. the parameters of the encoder and decoder; see Equations (1) and (2)]. That is, rectangles with the same shading patterns share parameters, and thin rectangles with black and white shading pattern denote densely connected neural layers

来自 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6022705/#