A. General

This is a package for GO term similarity comparison . There are multiple methods from differect categories of bibliography concerning GO term similarity .

B. Set Up

1. Java - 11

• Install : sudo apt-get install openjdk-11-jdk
• Add to path : export JAVA_HOME=/usr/lib/jvm/java-11-openjdk-amd64

2. OWL Tools :

• Source : owltools

1. Get from GitHub : git clone https://github.com/owlcollab/owltools.git
2. Move to OWLTools-Parent directory .
3. Install using maven : Run mvn clean install .
4. Add owltools to path : export PATH=$PATH:directory/to/where/you/downloaded/owltools/OWLTools-Runner/target

3. Python3 Modules

• ontobio : pip install ontobio
• networkx : pip install networkx[default,extra]

C. Methods

Edge Based Methods

1. Simple Edge Counting Method

• Source : Development and application of a metric on semantic nets
• Implementation : ./EdgeBasedSimilarity/edgeBasedMethods.py

1.1 Between Term Similarity

• Find the minimum lenght path .
• Each transition costs 1 .

1.2 Between Entities

• For each combination of terms from each entity find the minimum path of terms .
• Sum all the distances and divide by product of number of elements of each set .

2. Weight Edges using the maximum distance from LCA to the root

• Source : A Knowledge-Based Clustering Algorithm Driven by Gene Ontology
• Implementation : ./EdgeBasedSimilarity/edgeBasedMethods.py

2.1 Between Term Similarity

• Find the LCA of terms .
• Find the maximum distance of LCA from the root .
• Use the formula from :

$$ W_{path(t_1,t_2)} = \sum_{i=0}^{n}{(w)^i} $$

• w is a hyperparameter for weight . (.815)

2.2 Between Entities

3. Semantic Value Method

• Source : A new method to measure the semantic similarity of GO terms
• Implementation : ./EdgeBasedSimilarity/edgeBasedMethods.py

3.1 Between Term Similarity

• Construct the subgraph for each term $t$ involving all of its ancestors $N_t$ in ontology graph .

$$ G_t = (t, N_t , E_t) $$

• Calculate s-value for a term in each subgraph using the formula :

$$ S_{t}(t_i) = \begin{equation} \begin{cases}S_{t}(t) = 1 \\ S_t(t_i) = \max{{w_e\cdot S_{t}(t_i')|t_i' \in childrenof(t_i) \in N_t}} \ if \ t \ \neq \ t_i\end{cases} \end{equation} $$

• Calculate term similarity using the formulas :

$$ S_{GO}(t_1, t_2) = \frac{\sum_{t_i \in N_{t_1}\cap N_{t_2}}{(S_{t_1}(t_i) + S_{t_2}(t_i)}}{SV(t_1) + SV(t_2)} $$

$$ SV(t) = \sum_{t_i \in N_t}{S_t(t_i)} $$

4. Shortest Semantic Differentiation Distance - SSDD

• Source : A novel insight into Gene Ontology semantic similarity
• Implementation : ./EdgeBasedSimilarity/edgeBasedMethods.py

4.1 Between Term Similarity

---

Information Content Methods

1. Resnik Semantic Similarity

2. Jiang & Conrath Semantic Similarity

3. Lin Semantic Similarity

4. Relevance Semantic Similarity

5. SimIC

---

Hybrid Methods

1. Integrated Semantic Similarity

2. Hybrid Relative Specificity Similarity

• Source : Improving the Measurement of Semantic Similarity between Gene Ontology Terms and Gene Products: Insights from an Edge- and IC-Based Hybrid Method

2.1 Between Term Similarity

• Find Most Informative Common Ancestor - MICA .
• Calculate Information Content Distance from root .

$$ \alpha_{IC} = -\ln{(p_{IC}(MICA)} $$

• Find Most Informative Leaf for each term .
• Calculate Average Information Content Distance from each MIL and term .

$$ \beta_{IC} = \frac{-\ln{(p_{IC}(t_1))} + \ln{(p_{IC}(MIL_1))} - \ln{(p_{IC}(t_2))} + \ln{(p_{IC}(MIL_2))} }{2} $$

• Find Minimum Path from each term to MICA .

---

Graph Embedding Methods

1. go2vec