Gene Curation Coalition Gene-Disease Mappings

1 Gene Curation Coalition Gene-Disease Mappings

2 Introduction

This project explores the GenCC database of gene-disease mappings [1]. The data contains information relating to the validity of gene-disease relationships, focusing on Mendelian diseases. The analyses carried out are intended to provide a general overview of the data, and to identify interesting patterns and trends in the data by consulting the Human Phenotype Ontology [2] and the MONDO disease ontology [3]. Specifically, we are interested in neural and digestive system disorders, how we can use the data to infer and compare the genetic basis of these disorders, and how related subclassed diseases are linked to each other.

3 Data and Methods

The project is divided into four parts. An overview and references of used data and tools is provided in Table 9 and Table 10, respectively. Detailed methodology is provided along the results where appropriate.

First (P1), we conduct a summary analysis of the GenCC data. We first check the data for duplicates and suspicious entries, and then examine gene-disease mapping statistics using the Pandas library for data analysis and the Seaborn and Matplotlib libraries for data visualisation. Using similar methods, we then investigate the confidence categories of the data, and the URLs provided as supporting assertion criteria to gauge the overall validity of the data. Finally, we plot a timeline of the number of submissions over time to identify any trends.

Second (P2), we investigate the Modes of Inheritance (MOI) of the diseases in the GenCC data. We use the Human Phenotype Ontology to obtain the MOI metadata, and then plot the number of unqiue genes in the GenCC data associated with each MOI. The Pronto library is used to work with the ontology data.

For the third part (P3), we use the MONDO disease ontology to select a subset of diseases from the GenCC data, and then plot the number of unique genes associated with each disease. We compare the number of genes associated with the diseases relating to nervous system disorder (NSD) and digestive system disorder (DSD). This is achieved by joining the GenCC with the MONDO disease ontology data. We also examine subclasses of these disorders (the Pronto library is again used for ontology analysis).

Finally (P4), we build a network of NSD diseases and their associated genes, and identify communities of diseases with shared genes. This is achieved using the NetworkX library for network analysis and the netgraph library for data visualisation, a library which builds on top of the NetworkX visualisation tools but provides further functionality to plot clustered networks. The clustering is performed using the algorithms.community.greedy_modularity_communities¹ NetworkX function with its default parameters, which uses the Clauset-Newman-Moore greedy modularity maximisation algorithm [4].

For reproducibility purposes we use Quarto, a scientific document authoring tool, to generate this report. All tabular and visual results presented in this report require all Python scripts to compile successfully before the report can be generated. Furthermore, we used a Python dependency management tool, Poetry, to manage the project dependencies and to document specific versions of the libraries used in this project (Listing 1). Lastly, we provide a zip-file of any data used for this project as these datasets are regularly updated and might not be available in the same form in the future. This compressed folder also contains all generated csv-files, tables, and plots used for this report, as well as a pickle-file of the network generated in P4.

3.1 High-Level Description of the GenCC Data

The GenCC data is provided as a tab-separated values (TSV) file, with 18504 rows and 30 columns. An overview of columns relevant for this project is provided in Table 12, Table 13.

3.1.1 Sanity Check

Before exploring gene-disease-mapping statistics, we first conducted a sanity check of the data. There are two entries with the same Unique Universal ID (UUID) (see Table 11 (1)). They are largely identical, except for the differences listed in Table 14. Due to one of the rows containing the submitter name but the other row containing the Mode of Inheritance (MOI) name and the submission classification, we decided to keep both rows, knowing that it will not affect the significance of any of our analyses. Another approach would be to merge the entries but we decided against this approach because we wanted to keep the original data as is, and there is a small change that the UUID for the second entry was generated incorrectly.

To achieve Figure 4, we corrected the submission_as_date field of three entries, one of which specified the year 3016 (Table 11 (2)) (i.e. a likely typo; corrected to 2016), and an alleged submission entry from 1930 (Table 11 (3)) which we removed. The last entry (Table 11 (4)) dates back to 2011 but the supporting assertion criteria URL explicitly references a document released in 2022. This and the fact that it would have been the only pre-2015 entry along with the 1930 entry was justification enough for us to also remove this entry.

4 Results

4.1 Part One – Summary Analysis of GenCC Data

4.1.1 Gene-Disease Mapping Statistics

The GenCC data contains 4888 unique genes and 6370 unique diseases. The number of genes associated with an individual disease ranges from 1 to 149. The top 10 diseases with the most number of associated genes are provided in Table 1 and Figure 1. The disease with most associated genes is complex neurodevelopmental disorder (149 genes), followed by Leigh syndrome (113), and retinitis pigmentosa (90).

The overall distribution of the number of genes associated with a disease is heavily skewed toward the left (Figure 2), showing that few gene associations per disease are the norm in this data (median: 1.0, std-dev: 4.47, see Table 15).

Table 1: Top 10 diseases with the most number of associated genes (GenCC).

Disease Name	Number of Genes
complex neurodevelopmental disorder	149
Leigh syndrome	113
retinitis pigmentosa	90
schizophrenia	88
nonsyndromic genetic hearing loss	85
Tourette syndrome	76
mitochondrial disease	76
hearing loss, autosomal recessive	72
syndromic intellectual disability	66
male infertility with azoospermia or oligozoospermia due to single gene mutation	53

Figure 1: Top 10 diseases with most associated unique genes in the GenCC data.

Figure 2: Distribution of the number of genes associated with an individual disease (GenCC data).

4.1.2 Evidence Confidence Categories

The GenCC data contains 8 confidence categories which indicate the level of evidence supporting the validity of a data entry, shown in Table 2 and Figure 3. The barchart shows that most entries fall into either the definite, strong, or supportive category, with the latter registering the most entries (5330). The fewest entries have refuted or disputed evidence, potentially because such entries might be removed from the data by the Gene Coalition as soon as they are identified to be invalid.

Table 2: Number of GenCC entries for each confidence category.

Confidence Category	GenCC Entries Count
Definitive	4178
Strong	4720
Supportive	5330
Moderate	1791
Limited	2030
Disputed Evidence	182
Refuted Evidence	27
No Known Disease Relationship	246

Figure 3: Number of GenCC entries for each confidence category.

4.1.3 Supporting Assertion Criteria URLs

The GenCC data has a column submitted_as_assertion_criteria_url to provide provenance information to support the entry. We investigated this considering two standardised resource identifiers: PubMedID (PMID) and Digital Object Identifier (DOI) (Table 3).²

Most fields in this column provide a URL outwith the PMID or DOI categories (Other). This does not mean that the URL does not link to a valid resource. In a more open-ended project, we would investigate the URLs that fall into the Other category further.

Table 3: Overview of GenCC Assertion Criteria URLs containing standardised identifiers (PMID or DOI), of remaining URLs (Other), or NaN.

PMID	DOI	Other	NaN
2507	533	14216	1248

4.1.4 Timeline of Submission Counts

The GenCC data also provides the date of submission entries (submitted_as_date column). We can use this data to plot a timeline of the number of submissions over time (Figure 4).

Figure 4: Number of submissions to GenCC, grouped by year. Note that outliers were removed (Section 3.1.1).

The plot shows that the number of submissions per year is commonly in the range of roughly 50 to 2200 submissions. Two notable exceptions are the years 2020 (3625 submisions) and 2021 (7534 submissions). We identified that the extreme spike for 2021 is caused by Orphanet submitting their rare disease gene evidence data in one go [5]. The trend in 2020 could be explained by the official launch of the GenCC database happening in that year [6].

4.2 Part Two – Modes of Inheritance

Errata:

The number of genes per mode of inheritance is miscounted, likely an error with filtering for unique genes (Figure 5).
The interpretation of the distribution is incorrect: autosomal means all chromosomes whereas X and y are single with X being larger than y.

Mode of Inheritance (MOI) is a classification of the way in which a genetic condition is inherited [7]. The GenCC data comprises 18504 entries of which 100% have a MOI Compact URI specified in the corresponding moi_curie column. 329 entries (%) have the placeholder MOI “HP:0000005”, “mode of inheritance”, which describes an unknown or unspecified MOI.

Joining this data with the human phonotype ontology (Table 10), we can obtain the corresponding MOI metadata (Table 4).

Table 4: MOI CURIEs present in the GenCC data, and their associated descriptions.

MOI CURIE	Name
HP:0000005	Mode of inheritance
HP:0000006	Autosomal dominant inheritance
HP:0000007	Autosomal recessive inheritance
HP:0001417	X-linked inheritance
HP:0001419	X-linked recessive inheritance
HP:0001423	X-linked dominant inheritance
HP:0001427	Mitochondrial inheritance
HP:0001442	Typified by somatic mosaicism
HP:0001450	Y-linked inheritance
HP:0010984	Digenic inheritance
HP:0012274	Autosomal dominant inheritance with paternal imprinting
HP:0012275	Autosomal dominant inheritance with maternal imprinting
HP:0032113	Semidominant inheritance

Merging the data from Table 4 with the GenCC data, we can investigate how many unique genes are associated with each MOI (Figure 5).

Figure 5: Number of GenCC entries per MOI. The plot provides counts where the bar is not visible due to a much larger maximum count in the data.

Autosomal dominant inheritance and autosomal recessive inheritance are the most common MOI (2464, 2963, respectively), with next most common category, mode of inheritance, being featured much less in the GenCC data (329 times). Autosomal dominant inheritance implies that only one altered gene copy is needed for a condition to to be inherited [8]. This significantly increases the likelihood of a child inheriting the condition if one parent carries it. Hence, such conditions are prevalent in the population.

Autosomal recessive conditions needs both copies of the gene to be mutated for the condition to manifest in a child. The higher count of recessive conditions in Figure 5 could stem from their increased severity, motivating more extensive research and consequently more associated gene data being analysed and discovered.

Regarding the scarcity of other categories like Y-linked inheritance, this could be explained with their rarity in the population. These MOIs are less common, reducing the occurrence of gene mutations associated with these inheritance patterns.

4.3 Part Three – Disease Groupings and Their Associated Genes

For the next analysis we used the MONDO disease ontology [3] to select a specific subset of GenCC diseases and find their associated genes. Specifically, we investigated the nervous system disorder (MONDO:0005071; NSD). NSD features 5587 subclasses in the ontology, excluding the root node. The first 10 subclasses are listed in Table 5, showing that the ontology is structured in a hierarchical manner, with NSD being an umbrella term for more specific disease definitions. We repeated this analysis with digestive system disorder (1446 subclasses in total; Table 16 in Appendix; DSD).

Table 5: Top 10 subclasses in the MONDO disease ontology for NSD.

MONDO ID	MONDO Name
MONDO:0002320	congenital nervous system disorder
MONDO:0002602	central nervous system disorder
MONDO:0002977	autoimmune disorder of the nervous system
MONDO:0003569	cranial nerve neuropathy
MONDO:0003620	peripheral nervous system disorder
MONDO:0004466	neuronitis
MONDO:0004618	diplegia of upper limb
MONDO:0005283	retinal disorder
MONDO:0005287	developmental disability
MONDO:0005391	restless legs syndrome

The top subclasses in the NSD category include various specific conditions, including autoimmune, developmental, and localised neuropathies, reflecting a spectrum from broad neurological issues to specific anatomical or functional disruptions. For DSD, the subclasses largely focus on organ-specific diseases such as neoplasms, autoimmune problems, and localised issues like peptic ulcers, demonstrating a more concentrated emphasis on the digestive tract.

Table 6: Top 10 diseases subclassing NSD with most unique genes associated in GenCC.

MONDO ID	Disease Name	Gene Count
MONDO:0100038	complex neurodevelopmental disorder	149
MONDO:0009723	Leigh syndrome	113
MONDO:0019200	retinitis pigmentosa	90
MONDO:0005090	schizophrenia	88
MONDO:0019497	nonsyndromic genetic hearing loss	85
MONDO:0007661	Tourette syndrome	76
MONDO:0019588	hearing loss, autosomal recessive	72
MONDO:0000508	syndromic intellectual disability	66
MONDO:0001071	intellectual disability	49
MONDO:0100062	developmental and epileptic encephalopathy	45

NSDs like complex neurodevelopmental disorder and Leigh syndrome exhibit a notably higher gene count in the GenCC data, suggesting a larger number of genetic factors involved in these disorders. DSDs such as hereditary nonpolyposis colon cancer and permanent neonatal diabetes mellitus have fewer associated genes, suggesting a comparatively more focused genetic involvement in these conditions. This is further illustrated in Table 6 (Table 17 for DSD).

Indeed, comparing the total number of genes associated with the disorders in the GenCC data, we see that diseases associated with NSD have a higher total gene count (2211) than DSD (217). Table 7 presents the top 10 genes by number of entries for the GenCC data filtered for NSD and DSD.

Table 7: Top 10 genes by number of entries for NSD and DSD GenCC data.

(a) NSD

Gene	Number of restricted GenCC entries
SCN4A	17
MECP2	16
SCN1A	15
POMGNT1	15
COL6A3	15
TTN	15
ARX	15
PLP1	14
MYO7A	14
CACNA1A	14

(b) DSD

Gene	Number of restricted GenCC entries
KCNJ11	17
ABCC8	15
EPCAM	11
GCK	11
INS	10
GUCY2C	9
MLH1	9
MSH6	8
HNF4A	8
MSH2	8

4.4 Part Four – Building a Simple GenCC Disease Network

This part of the project focused on building a simple network of diseases falling under the NSD category, and of their associated genes. Table 8 a) presents the top 10 disease pairs with the highest shared genes counts. The top pair, nonsyndromic genetic hearing loss (NGHL) and hearing loss, autosomal recessive (HLAR), share 59 genes. This is not surprising as NGHL, a subtype of hearing loss without additional medical issues or symptoms, can also be inherited in an autosomal recessive manner [9].

Table 8 b) shows the communities of diseases identified in the network built from the results of the previous analysis, only adding disease pairs with three or more common genes to the network. The network is visualised in Figure 6, featuring 18 communities, with the largest community C1 containing 12 nodes. 93 edges connect 76 nodes; the network is not fully connected so that most communities are isolated from each other.

Within a community, nodes are generally highly connected, indicating a high number of shared genes between community-specific diseases. For instance, having a closer look at how diseases are related within the largest community C1 (Table 18), the common factor among these disorder seems involve intellectual disability. This implicates shared pathways related to genetic links or disruptions in brain functions, caused by a highly overlapping set of genes.

Table 8: NSD Network Details

(a) Disease pairs sharing the most genes in GenCC.

Disease Pair	Gene_Count
MONDO:0019588,MONDO:0019497	54
MONDO:0019587,MONDO:0019497	29
MONDO:0009723,MONDO:0016815	28
MONDO:0019609,MONDO:0019234	13
MONDO:0019587,MONDO:0019588	11
MONDO:0100038,MONDO:0015802	11
MONDO:0000508,MONDO:0014699	10
MONDO:0000508,MONDO:0100038	10
MONDO:0019200,MONDO:0018998	9
MONDO:0019200,MONDO:0015993	8

(b) Communities in the NSD network, ordered by size.

Community Number	Number of Diseases
C1	12
C2	7
C3	6
C4	6
C5	5
C6	5
C7	5
C8	4
C9	4
C12	3
C13	3
C10	3
C11	3
C14	2
C15	2
C16	2
C17	2
C18	2

Figure 6: Network graph of NSD-associated diseases in the GenCC data, clustered into communities using Clauset-Newman-Moore greedy modularity maximization [4]. Nodes represent diseases, edges represent common genes in GenCC. Nodes with less than three shared genes are excluded.

5 Discussion

This project looked into the GenCC database, exploring gene-disease mappings and leveraging support from the MONDO disease ontology and Human phenotype ontology. We found that the majority of diseases are linked to a limited number of genes, with a trimmed mean average of 1.67 genes per disease. The validity of the data is predominantly supported by strong, or definite, supportive evidence. Most entries lacked an associated PMID or DOI as assertion URL. The temporal analysis revealed spikes in data submissions in 2020 and 2021, with the latter surge attributed to the submission of rare disease gene evidence by Orphanet and the former likely associated with the official database launch.

The investigation into Modes of Inheritance (MOIs) confirmed that autosomal dominant and autosomal recessive patterns were the most prevalent, aligning with their high occurrence in the population. Utilising the MONDO disease ontology, we compared NSD and DSD, finding that NSD exhibited a significantly higher number of subclasses and associated genes than DSD. This discrepancy suggested a more concentrated genetic involvement in the subclassed conditions of DSD, potentially indicating a more organ-specific focus for DSD compared to the broader implications of NSDs.

A network focusing on NSD diseases and their associated genes was clustered into 18 distinct communities, with the largest community centering around intellectual disability. This community highlighted shared pathways related to genetic links or disruptions in brain functions.

There are several opportunities for future work. First, investigation into URLs falling into the Other assertion_criteria_url category could yield further provenance information to support the entry to ascertain that the provided evidence categories are accurate. Further research could compare the locations of disease-specific genes within and between clusters of the network to gain insight into which shared genes between diseases have what effect in the overall picture of NSD.

References

[1]

M. T. DiStefano et al., “The Gene Curation Coalition: A global effort to harmonize genedisease evidence resources,” Genetics in Medicine, vol. 24, no. 8, pp. 1732–1742, Aug. 2022.

[2]

S. Köhler et al., “The Human Phenotype Ontology in 2021,” Nucleic Acids Research, vol. 49, no. D1, pp. D1207–D1217, Jan. 2021.

[3]

N. A. Vasilevsky et al., “Mondo: Unifying diseases for the world, by the world,” Health Informatics, Preprint, Apr. 2022.

[4]

A. Clauset, M. E. J. Newman, and C. Moore, “Finding community structure in very large networks,” Physical Review E, vol. 70, no. 6, p. 066111, Dec. 2004.

[5]

“Orphanet submitter information and submissions.” https://search.thegencc.org/submitters/GENCC_000110, Nov-23AD.

[6]

“Gene Curation Coalition Launches GenCC Database - ClinGen | Clinical Genome Resource.” https://clinicalgenome.org/docs/gencc/, Dec-2020.

[7]

G. Alliance and T. N. Y.-M.-A. C. for G. and N. Screening Services, “INHERITANCE PATTERNS,” in Understanding Genetics: A New York, Mid-Atlantic Guide for Patients and Health Professionals, Genetic Alliance, 2009.

[8]

R. G. Lewis and B. Simpson, “Genetics, Autosomal Dominant,” in StatPearls, Treasure Island (FL): StatPearls Publishing, 2023.

[9]

A. E. Shearer, M. S. Hildebrand, A. M. Schaefer, and R. J. Smith, “Genetic Hearing Loss Overview,” in GeneReviews, M. P. Adam, J. Feldman, G. M. Mirzaa, R. A. Pagon, S. E. Wallace, L. J. Bean, K. W. Gripp, and A. Amemiya, Eds. Seattle (WA): University of Washington, Seattle, 1993.

[10]

W. McKinney, “Pandas: Powerful Python data analysis toolkit,” 2021.

[11]

E. Bisong, “Matplotlib and Seaborn,” in Building Machine Learning and Deep Learning Models on Google Cloud Platform, Berkeley, CA: Apress, 2019, pp. 151–165.

[12]

M. Larralde et al., “Althonos/pronto: V2.5.5.” Zenodo, Aug-2023.

[13]

A. Hagberg, D. Schult, P. Swart, and J. M. Hagberg, “Exploring Network Structure, Dynamics, and Function using NetworkX,” 2008.

[14]

P. J. N. Brodersen, “Netgraph: Publication-quality Network Visualisations inPython,” Journal of Open Source Software, vol. 8, no. 87, p. 5372, Jul. 2023.

[15]

J. J. Allaire, C. Teague, C. Scheidegger, Y. Xie, and C. Dervieux, “Quarto.” Jan-2022.

[16]

S. Eustace and The Poetry contributors, “Poetry: Python packaging and dependency management made easy.” Nov-2023.

Appendix

Table 9: Overview of tools used in the project.

Library/Tool	Description	Used for Part
Pandas [10]	Data analysis library for Python	P1,P2,P3,P4
Matplotlib [11]	Data visualisation library for Python	P1,P2,P3,P4
Seaborn [11]	Data visualisation library for Python	P1,P2,P3,P4
Pronto [12]	Python library for working with ontologies	P2,P3
NetworkX [13]	Python library for network analysis	P4
netgraph [14]	Python library for network visualisation	P4
Quarto [15]	Document authoring tool	P1,P2,P3,P4
Poetry [16]	Python dependency management tool	P1,P2,P3,P4

Table 10: Overview of data used in the project.

Data	Description	Date of Download (YYYY-MM-DD)
GenCC [1]	Gene-disease mapping data	2023-11-07
Human Phenotype Ontology [2]	Ontology of human phenotypes	2023-11-07
MONDO Disease Ontology [3]	Ontology of disease definitions	2023-11-07

Table 11: UUIDs of suspicious entries in the GenCC data

Number	UUID
1	GENCC_000112-HGNC_15516-OMIM_615777-HP_0000007-GENCC_100002
2	GENCC_000107-HGNC_28883-OMIM_617222-HP_0000007-GENCC_100002
3	GENCC_000114-HGNC_20074-MONDO_0015159 -HP_0000007-GENCC_100001
4	GENCC_000101-HGNC_9588-OMIM_605309-HP_0000006-GENCC_100002

Table 12: Statistical descriptions relevant GenCC data columns (1). Too long “top” values are ommitted.

	uuid	gene_curie	gene_symbol	disease_curie	submitted_as_classification_id
count	18504	18504	18504	18394	18504
unique	18503	4888	4888	6370	9
top	ommitted	HGNC:2200	COL2A1	MONDO:0100038	GENCC:100009
freq	2	44	44	159	5330

Table 13: Statistical column descriptions of the GenCC data (2). Too long “top” values are ommitted.

	moi_curie	submitted_as_classification_name	submitted_as_date	submitted_as_assertion_criteria_url
count	18504	16372	18504	17256
unique	13	18	6904	989
top	HP:0000007	Supportive	2021-09-14 00:00:00	omitted
freq	9122	5330	5330	5330

Table 14: Value differences of the presumably duplicated entries in the GenCC data.

	self	other
submitted_as_moi_name	nan	Autosomal recessive inheritance
submitted_as_submitter_name	TGMI G2P	nan
submitted_as_classification_name	nan	Strong
submitted_as_date	2019-03-27 00:00:00	2019-03-27 22:15:19
submitted_as_submission_id	1000115516	2173
submitted_run_date	2023-09-11	2020-11-25

Table 15: Statistics of the distribution of the number of genes associated with a disease (GenCC), rounded to two decimal places. The top 10 entries were removed to calculate the trimmed mean.

Statistic	Value
Median	1.00
Trimmed Mean@^10	1.67
Standard Deviation	4.47

Table 16: Top 10 subclasses in the MONDO disease ontology for DSD.

MONDO ID	MONDO Name
MONDO:0000385	benign digestive system neoplasm
MONDO:0000588	autoimmune disorder of gastrointestinal tract
MONDO:0000888	gastrointestinal mucositis
MONDO:0001673	diarrheal disease
MONDO:0002356	pancreas disorder
MONDO:0002515	hepatobiliary disorder
MONDO:0002516	digestive system cancer
MONDO:0003749	esophageal disorder
MONDO:0004247	peptic ulcer disease
MONDO:0004298	stomach disorder

Table 17: Top 10 diseases subclassing DSD with most unique genes associated in GenCC.

MONDO ID	Disease Name	Gene Count
MONDO:0018630	hereditary nonpolyposis colon cancer	17
MONDO:0018911	maturity-onset diabetes of the young	13
MONDO:0015967	monogenic diabetes	12
MONDO:0100164	permanent neonatal diabetes mellitus	12
MONDO:0016391	neonatal diabetes mellitus	10
MONDO:0005575	colorectal cancer	10
MONDO:0020525	transient neonatal diabetes mellitus	9
MONDO:0009831	malignant pancreatic neoplasm	8
MONDO:0005835	Lynch syndrome	7
MONDO:0018309	Hirschsprung disease	7

Table 18: Diseases in the largest community (C1).

ID	Name
MONDO:0000508	syndromic intellectual disability
MONDO:0018614	undetermined early-onset epileptic encephalopathy
MONDO:0014699	intellectual disability, autosomal dominant 40
MONDO:0005027	epilepsy
MONDO:0005260	autism
MONDO:0015802	autosomal dominant non-syndromic intellectual disability
MONDO:0100172	intellectual disability, autosomal dominant
MONDO:0100038	complex neurodevelopmental disorder
MONDO:0001071	intellectual disability
MONDO:0005258	autism spectrum disorder
MONDO:0019502	autosomal recessive non-syndromic intellectual disability
MONDO:0017615	benign familial infantile epilepsy

Listing 1: pyproject.toml configuration used for Poetry

[tool.poetry]
name = "pf-gene-disease-mapping-exploration"
version = "0.1.0"
description = ""
authors = []
readme = "README.md"

[tool.poetry.dependencies]
python = ">=3.11,<3.13"
pandas = "^2.1.2"
seaborn = "^0.13.0"
jupyter = "^1.0.0"
matplotlib = "^3.8.1"
tabulate = "^0.9.0"
pronto = "^2.5.5"
rich = "^13.6.0"
networkx = "^3.2.1"
netgraph = "^4.12.11"
jupyter-cache = "^1.0.0"


[build-system]
requires = ["poetry-core"]
build-backend = "poetry.core.masonry.api"

Footnotes

NetworkX greedy_modularity_communities documentation ↩︎
For the PMID, we did not require a strict PMID:ID format but also allowed the https://pubmed.ncbi.nlm.nih.gov/<PMID> format.↩︎

Community Number	Number of Diseases
C1	12
C2	7
C3	6
C4	6
C5	5
C6	5
C7	5
C8	4
C9	4
C12	3
C13	3
C10	3
C11	3
C14	2
C15	2
C16	2
C17	2
C18	2

Community Number	Number of Diseases
C1	12
C2	7
C3	6
C4	6
C5	5
C6	5
C7	5
C8	4
C9	4
C12	3
C13	3
C10	3
C11	3
C14	2
C15	2
C16	2
C17	2
C18	2

Community Number	Number of Diseases
C1	12
C2	7
C3	6
C4	6
C5	5
C6	5
C7	5
C8	4
C9	4
C12	3
C13	3
C10	3
C11	3
C14	2
C15	2
C16	2
C17	2
C18	2