# Integrating Orthology and Disease association databases

[TOC]

When studying the function of a gene in disease models, basic science research often starts by determining the effects of the gene in vitro models by knocking out the gene in cells or overexpressing the gene to understand what downstream effects it may have. After characterization in vitro, scientists move on to study the gene in live animal systems. One of the most common animal models that are employed throughout the field of biological research is the mouse. Not only are mice very similar to humans biologically, but they can also acquire many of the same diseases as humans. In addition, mice can be genetically manipulated to mimic most diseases found in humans.

> *source: https://www.clinicalomics.com/topics/informatics-topic/bioinformatics/essential-mouse-genes-are-related-to-many-human-disease-genes/

In this project, our biological feature was homologous genes of mice and humans. Our goal was to create a database that combines gene Id of known orthologs in humans and mice, ortholog quality scores, and diseases associated with the human ortholog. In our analysis, we compared homologous genes of mice and humans to determine whether mice can be used as certain models for various diseases. We determined which diseases are associated with certain genes to determine which specific human diseases transgenic mice could be used as models for. In addition, we tried to find out if there is any relationship between the type of ortholog and percentage identical ortholog score, Mouse Gene-order conservation score, and number of associated diseases.

## Methods
### Ensemble Biomart Table
#### Data Acquisition

Data for the mouse and human homologous gene scores were obtained from Ensembl BioMart.
Ensembl BioMart, which is a web-based tool that allows for data integration and acquisition based on the type of genetic or protein sequence analysis desired.
Because we were interested in comparing human genes to mouse genes, we selected the following filters: Mouse gene name, Gene name, Mouse homology type, Mouse gene name, Mouse orthology confidence [0 low, 1 high], %id. target Mouse gene identical to query gene, mouse gene order conservation score, last common ancestor with mouse, query protein or transcript ID, and %id. query gene identical to target Mouse gene.

The first few rows are attached below
|Mouse gene name |Mouse homology type | Mouse orthology confidence [0 low, 1 high]| %id. target Mouse gene identical to query gene| %id. query gene identical to target Mouse gene| Mouse Gene-order conservation score|Last common ancestor with Mouse |Gene name |Query protein or transcript ID |
|:---------------|:-------------------|------------------------------------------:|----------------------------------------------:|----------------------------------------------:|-----------------------------------:|:-------------------------------|:---------|:------------------------------|
|mt-Nd1 |ortholog_one2one | 0| 77.0440| 77.0440| 50|Euarchontoglires |MT-ND1 |ENSP00000354687 |
|mt-Nd2 |ortholog_one2one | 1| 57.0605| 57.3913| 75|Euarchontoglires |MT-ND2 |ENSP00000355046 |
|mt-Co1 |ortholog_one2one | 1| 90.8382| 90.6615| 100|Eutheria |MT-CO1 |ENSP00000354499 |
|mt-Co2 |ortholog_one2one | 1| 71.3656| 71.3656| 100|Mammalia |MT-CO2 |ENSP00000354876 |
|mt-Atp8 |ortholog_one2one | 0| 45.5882| 46.2687| 100|Eutheria |MT-ATP8 |ENSP00000355265 |
|mt-Atp6 |ortholog_one2one | 1| 75.6637| 75.6637| 100|Eutheria |MT-ATP6 |ENSP00000354632 |

The table has three types of orthologs. One2one types have one mouse ortholog for a human gene. One2many types have at least two orthologous mice genes for a human gene. Many2many types have at least 2 orthologous mice genes for multiple human genes.

Two of the variable that we were most interested include %id. target Mouse gene identical to query gene and %id. query gene identical to target Mouse gene. In this case, the query is the human sequence and the target is the mouse sequence. Target % ID refers to the percent of mouse sequence that matches the human sequence whereas query % ID refers to the percent of human sequence that matches to mouse sequence.

Mouse orthology confidence refers to how well the orthologue pairs meet certain thresholds. These thresholds depend on the most-common recent ancestor of the species pair.

Orthologous genes are likely to be part of a block of genes that were inherited from a common ancestor. Gene order observation score is calculated by fetching two upstream and downstream genes and comparing them. If a gene is by itself without orthologous neighbors, it is given a 0 score.

#### Data Simplification
The BIOMART orthology file has two columns that were not relevant to the task at hand.
The last common ancestor with mouse and Query protein columns were removed using the following command in r

homology <- select(homology, c("Gene name", "Mouse homology type", "Mouse gene name", "Mouse orthology confidence [0 low, 1 high]", "%id. target Mouse gene identical to query gene", "%id. query gene identical to target Mouse gene"))

After simplification, the table looked like this:
Gene name |Mouse homology type |Mouse gene name | Mouse orthology confidence [0 low, 1 high]| %id. target Mouse gene identical to query gene| %id. query gene identical to target Mouse gene| Mouse Gene-order conservation score|
|:---------|:-------------------|:---------------|------------------------------------------:|----------------------------------------------:|----------------------------------------------:|-----------------------------------:|
|MT-ND1 |ortholog_one2one |mt-Nd1 | 0| 77.0440| 77.0440| 50|
|MT-ND2 |ortholog_one2one |mt-Nd2 | 1| 57.0605| 57.3913| 75|
|MT-CO1 |ortholog_one2one |mt-Co1 | 1| 90.8382| 90.6615| 100|
|MT-CO2 |ortholog_one2one |mt-Co2 | 1| 71.3656| 71.3656| 100|
|MT-ATP8 |ortholog_one2one |mt-Atp8 | 0| 45.5882| 46.2687| 100|
>

### DisGeNET

#### Data Acquisition

Data for which genes are associated with which human diseases were obtained from DisGeNET. DisGeNET is an online database that is designed to address many questions concerning the genetics of human diseases. It is one of the largest repositories for human gene-disease associations and integrates data from several other curated repositories. From DisGeNET, we were able to obtain a curated table containing geneSymbol and diseases.

| geneId|geneSymbol | DSI| DPI|diseaseId |diseaseName |diseaseType |diseaseClass |diseaseSemanticType | score| EI| YearInitial| YearFinal| NofPmids| NofSnps|source |
|------:|:----------|-----:|-----:|:---------|:------------------------|:-----------|:------------|:--------------------------------|-----:|-----:|-----------:|---------:|--------:|-------:|:---------|
| 1|A1BG | 0.700| 0.538|C0019209 |Hepatomegaly |phenotype |C23;C06 |Finding | 0.30| 1.000| 2017| 2017| 1| 0|CTD_human |
| 1|A1BG | 0.700| 0.538|C0036341 |Schizophrenia |disease |F03 |Mental or Behavioral Dysfunction | 0.30| 1.000| 2015| 2015| 1| 0|CTD_human |
| 2|A2M | 0.529| 0.769|C0002395 |Alzheimer's Disease |disease |C10;F03 |Disease or Syndrome | 0.50| 0.769| 1998| 2018| 3| 0|CTD_human |
| 2|A2M | 0.529| 0.769|C0007102 |Malignant tumor of colon |disease |C06;C04 |Neoplastic Process | 0.31| 1.000| 2004| 2019| 1| 0|CTD_human |
| 2|A2M | 0.529| 0.769|C0009375 |Colonic Neoplasms |group |C06;C04 |Neoplastic Process | 0.30| 1.000| 2004| 2004| 1| 0|CTD_human |
| 2|A2M | 0.529| 0.769|C0011265 |Presenile dementia |disease |C10;F03 |Mental or Behavioral Dysfunction | 0.30| 1.000| 1998| 2004| 3| 0|CTD_human |

This database contains many unimportant columns which need to be removed before integration with the BIOMART table.
Each gene in the DisGeNET was linked to multiple diseases. Using right.join() to integrate both tables would lead to deletions of numerous gene-disease interactions since it was not a one-to-one mapping.

#### Data Simplification
Our goal was to convert the raw table into a form that has a one-to-one relationship between genes and associated diseases. To achieve this, we wrote a python script that converts the table into 3 columns. The first one is the gene name, the second one contains a list of all diseases associated with the gene and the third column contains the number of diseases associated with the gene.

#python code for disease annotations
import csv
import pandas as pd
infile = open("curated_gene_disease_associations.tsv")
infile.readline()

temp = 'A1BG'
disdict = {'A1BG':[]}
for line in infile:
    line = line.strip()
    data = line.split("\t")
    if data[1] != temp:
        disdict[data[1]] = [data[5]]
    else:
        disdict[data[1]].append(data[5])
    temp= data[1]

with open("test5.tsv",'w', newline='') as out_file:
    tsv_writer = csv.writer(out_file, delimiter='\t')
    for key, value in disdict.items():
        tsv_writer.writerow([key, value, len(value)])

DisGenet table after simplification

|Gene name |Associated_diseases | Number of diseases|
|:---------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------:|
|A1BG |['Hepatomegaly', 'Schizophrenia'] | 2|
|A2M |["Alzheimer's Disease", 'Malignant tumor of colon', 'Colonic Neoplasms', 'Presenile dementia', 'Mental Depression', 'Depressive disorder', 'Hepatolenticular Degeneration', '"Kidney Failure, Acute"', 'Liver Cirrhosis', '"Liver Cirrhosis, Experimental"', 'Lung diseases', 'Lung Neoplasms', 'Nephrotic Syndrome', 'Hepatocellular Adenoma', '"Fibrosis, Liver"', 'Malignant neoplasm of lung', 'Familial Alzheimer Disease (FAD)', '"Alzheimer Disease, Late Onset"', 'Acute Confusional Senile Dementia', '"Alzheimer\'s Disease, Focal Onset"', '"Alzheimer Disease, Early Onset"', 'Hepatic Form of Wilson Disease', 'Acute Kidney Insufficiency', 'Liver carcinoma', 'Acute kidney injury', 'alpha-2-Macroglobulin Deficiency'] | 26|

### Data Integration
Before Integration, we need to remove all rows with NA values in them. This is accomplished by the following command.

disease <- na.omit(disease)
homology <- na.omit(homology)

Now our BIOMART orthology table and DisGenET gene-disease interaction table have a common column called gene name. We can use right.join() to integrate the tables.

homo1 <- right_join(x=homology, y=disease, by="Gene name")

## Results
### Integrated Database
The result is a Table that contains Ortholog gene data, Orthology scores, and diseases associated with the genes. This table has 25125 observations. The first few rows of the table are attached below.
|Gene name |Mouse homology type |Mouse gene name | Mouse orthology confidence [0 low, 1 high]| %id. target Mouse gene identical to query gene| %id. query gene identical to target Mouse gene| Mouse Gene-order conservation score|Associated_diseases | Number of diseases|
|:---------|:-------------------|:---------------|------------------------------------------:|----------------------------------------------:|----------------------------------------------:|-----------------------------------:|:--------------------------------------------------------------------------------------------------------------------------------------|------------------:|
|USP9Y |ortholog_one2one |Usp9y | 0| 82.8180| 82.7856| 25|['Partial chromosome Y deletion', '"Spermatogenic Failure, Nonobstructive, Y-Linked"', 'Male sterility due to Y-chromosome deletions'] | 3|
|DAZ2 |ortholog_one2many |Dazl | 0| 30.6452| 57.3825| 0|['Partial chromosome Y deletion', '"Spermatogenic Failure, Nonobstructive, Y-Linked"', 'Male sterility due to Y-chromosome deletions'] | 3|
|RBMY1A1 |ortholog_many2many |Gm21708 | 0| 31.2500| 40.7895| 0|['Partial chromosome Y deletion', '"Spermatogenic Failure, Nonobstructive, Y-Linked"', 'Male sterility due to Y-chromosome deletions'] | 3|

### Relationship between type of ortholog and orthology scores
**Box plot of the effects of type of ortholog and %id. query gene identical to target Mouse gene score**
**Violin plot to visualize the relationship between the type of ortholog and Mouse Gene-order conservation score**

### Relationship between type of ortholog and number of associated diseases.
**Box plot of type of ortholog and number of associated diseases.**

#Python code for  the generated graphs
import matplotlib.pyplot as plt
from statistics import median
infile = open("homo1lol.tsv")
infile.readline()

one2one = []
one2many= []
many2many =[]

one2onescore=[]
one2manyscore = []
many2manyscore = []
for line in infile:
    line = line.strip()
    data = line.split("\t")
    if data[1] =='ortholog_one2one':
        one2one.append(float(data[5]))
        one2onescore.append(float(data[6]))
    if data[1] =='ortholog_many2many':
        many2many.append(float(data[5]))
        many2manyscore.append(float(data[6]))
    if data[1] =='ortholog_one2many':
        one2many.append(float(data[5]))
        one2manyscore.append(float(data[6]))

x = [one2one,one2many,many2many]
plt.boxplot(x)
plt.show()

y = list([one2onescore,one2manyscore, many2manyscore])
fig,ax = plt.subplots()
ax.violinplot(y,showmeans=False, showmedians=True)
ax.set_title('violin plot')
ax.set_xlabel('x-axis')
ax.set_ylabel('Mouse Gene-order conservation score')
xticklabels = ['one2one', 'one2many', 'many2many',]
ax.set_xticks([1,2,3])
ax.set_xticklabels(xticklabels)
ax.yaxis.grid(True)
plt.show()

## Discussion
### Database limitations
Our integrated database does not include all the data types required for disease model selection. It is meant to simplify model selection by providing the quality scores of orthologous relationships and disease annotations. However, a high ontology quality score should not be the only attribute to base model selection on. The score does not contain information on other extremely important features like pathway interaction, gene regulation, diseases caused due to genetic variation to name a few.
### Effect of orthology type.
* From the first box plot (ortholog type vs %identical) we can observe that one2one orthologs have the highest percentage identical scores followed by one2many orthologs and many2many orthologs. This means that mice might not be a good model organism diseases associated with human genes with many2many orthologous relationships since they have lower average % identical scores.
* The second violin plot shows the distribution of gene order conservation scores among orthology types. Most one2one orthologs have a 100 score. The one2many orthologs are evenly distributed at both the extreme scores (100 and 0). This means approximately half of the one2many orthologs are part of orthologous gene blocks while the rest of them are not is such orthologous blocks. Many2many orthologs have the lowest average scores. This is consistent with the inference from the first graph.
* From the box plot(ortholog type vs number of diseases), it is clear ortholog type does not affect the number of diseases. This plot also reveals that few outliers have an extremely large number of annotated diseases.

### Hopes of what to expand
We were successful at integrating orthologous scores into our database. We failed at integrating protein blast identical percentage. This would help us compare and contrast proteins is essential for model selection. Other information we could integrate in the future are KEGG pathways associated with both human and mice genes, gene ontology analysis of associated genes, and included other model organisms from all over the phylogenetic tree into our database. Over time, we would add more model selection information from a different database to become a one-stop database for all model selection needs.

# Bibliography
* Kevin L Howe,et al. https://doi.org/10.1093/nar/gkaa942
* Gene-disease association data retrieved from DisGeNET v7.0 (https://www.disgenet.org/), Integrative Biomedical Informatics Group GRIB/IMIM/UPF .
* https://uswest.ensembl.org/info/genome/compara/Ortholog_qc_manual.html

NOTE: the input files from both database and the output table are submitted as supplementary files.