Protocol of disease gene identification using graphlet interaction.

a. The small network was taken, and only 5 types of graphlet interaction isomers (I1 to I5) were considered as an example. The black node 1 and node 3 were known disease genes. The protocol showed how to rank other genes according to known disease genes. b. The first step, calculation of the graphlet interaction between known disease gene (1, 3) and all the other genes. GI was the abbreviation of graphlet interaction, measured by a vector which had 5 elements corresponding to the numbers of the 5 types of graphlet interaction isomers (Figure. 1b I1 to I5). The graphlet interactions from one disease gene were added. Sum1 and Sum3 were the summations of the graphlet interaction vectors from node 1 and node 3. c. The second step, normalization of the graphlet interaction. Every graphlet interaction was divided by the corresponding summation. GI12, GI13, GI14 and GI15 were divided by Sum1, and GI31, GI32, GI34 and GI35 were divided by Sum3. d. The third step, the graphlet interactions from the disease gene to every candidate gene were summated. Then, the elements of the summation were multiplied by the weights, and then added. The score of the node was obtained. For example, to get score of node 2, the normalized GI12 and GI32 were added and the summation vector [0.83 0 0 1.67 1] was obtained. The score was 0.83+0+0+1.67+1 = 3.5 (the weight of every element was 1 here).

a. 本研究选取小型网络，并以5种图元相互作用异构体（graphlet interaction isomers，I1至I5）作为示例。其中黑色节点1与节点3为已知疾病基因，本流程展示了如何基于已知疾病基因对其余基因进行排序。
b. 第一步：计算已知疾病基因（节点1、节点3）与其余所有基因之间的图元相互作用（graphlet interaction，GI）。GI为图元相互作用的缩写，其以包含5个元素的向量进行量化，每个元素分别对应5种图元相互作用异构体的数量（见图1b中的I1至I5）。将单个疾病基因对应的图元相互作用向量进行求和，其中Sum1与Sum3分别为节点1与节点3的图元相互作用向量的求和结果。
c. 第二步：对图元相互作用进行归一化处理。将每一组图元相互作用除以其对应的求和结果：GI12、GI13、GI14与GI15除以Sum1，GI31、GI32、GI34与GI35除以Sum3。
d. 第三步：将疾病基因至每个候选基因的图元相互作用进行求和，随后将求和向量的各元素与对应权重相乘并累加，最终得到该节点的评分。以节点2的评分计算为例：将归一化后的GI12与GI32相加，得到求和向量[0.83, 0, 0, 1.67, 1]，将各元素直接求和（此处各元素权重均为1），最终得分为0.83+0+0+1.67+1=3.5。