Systems Biology · Network Bioinformatics · Functional Enrichment · Biomedical Data Integration
Portfolio version of the Master's Thesis “Organización funcional y relevancia biomédica de los genes del espliceosoma humano: un enfoque de Biología de Sistemas”, graded 9.6/10.
This repository documents and preserves the computational work behind the thesis as a technical portfolio project. The study combines curated spliceosome gene sets, protein–protein interaction analysis, network topology, functional enrichment, tissue-expression associations, human phenotype associations and clinical evidence.
Start here: START_HERE.md · Results gallery · Methods · Results by analysis
Data provenance: numerical values, genes, thresholds and reported findings are taken from the supplied thesis/project material. No new biological result has been added to the portfolio presentation.
The project starts from a manually curated set of 87 human spliceosome-related protein-coding genes, classified into 39 structural-core genes, 20 auxiliary factors and 28 regulatory factors.
The PPI network was generated in STRING v12.0 for Homo sapiens using a combined confidence threshold of ≥ 0.700 and a zero-order network. ISL1 remained in the complete gene set but had no interaction with the rest of the set at that threshold, leaving 86 connected genes for the topological analysis.
The resulting PPI network contained 1,832 edges, a mean interaction score of 0.784, and a STRING PPI-enrichment result of p < 1 × 10⁻¹⁶, compared with 96 expected interactions for a random set of the same size.
The analysis then integrated functional enrichment, HPA tissue associations, HPO phenotype associations and clinical evidence from OMIM, ClinVar and PubMed. SF3A2 had the highest PPI degree (69), whereas SRSF1 had the highest PPI betweenness (268.84).
| Area | Demonstrated work |
|---|---|
| Programming | R · data wrangling · automated result export · figure generation |
| Network biology | PPI networks · bipartite networks · Jaccard projections · Louvain communities |
| Graph metrics | Degree · betweenness · closeness · eigenvector centrality · B/D ratio |
| Functional genomics | g:Profiler · GO · Reactome · KEGG · HPO |
| Biomedical data | HPA · HPO · OMIM · ClinVar · PubMed |
| Data integration | Gene-symbol / Entrez mapping · heterogeneous biological annotations |
| Analysis | Result |
|---|---|
| Curated spliceosome set | 87 genes |
| Functional classification | 39 core · 20 auxiliary · 28 regulatory |
| Connected PPI network | 86 nodes · 1,832 edges |
| Mean STRING interaction score | 0.784 |
| STRING PPI enrichment | p < 1 × 10⁻¹⁶ |
| Highest PPI degree | SF3A2 — 69 |
| Highest PPI betweenness | SRSF1 — 268.84 |
| Topological reference subset | 15 genes |
| HPA global bipartite network | 76 genes · 346 tissues · 12,527 connections |
| HPO global bipartite network | 27 genes · 516 phenotypes · 836 connections |
| HPO Louvain communities | 14 |
| Jaccard threshold | J ≥ 0.20 groups · J ≥ 0.25 global |
The visual outputs are deliberately placed before the detailed methodology: the repository is designed to work as a portfolio as well as an academic archive.
STRING v12.0 network from the 87-gene input set. The thesis reports 86 connected nodes and 1,832 edges; ISL1 remained isolated at the ≥0.700 confidence threshold. Independently confirmed by STRING's own Network Stats panel:
The global HPA analysis produced 12,527 gene–tissue bipartite connections between 76 genes and 346 tissues. The global similarity projection used J ≥ 0.25 and Louvain community detection.
The global HPO analysis comprised 27 mapped genes, 516 phenotypes and 836 bipartite connections, followed by Jaccard projection at J ≥ 0.25 and Louvain community detection, yielding 14 communities.
More figures: Results gallery →
This project demonstrates the ability to move from a biological question to structured computational analysis, integrate heterogeneous biomedical resources, quantify network structure, generate interpretable visual outputs and document the boundary between automated analysis and external-platform results.
It is representative of work in network biology, systems biology, computational biomedicine and data-driven target/disease research.
The thesis asks whether the human spliceosome can be characterized as an interconnected biological system in which genes occupy distinguishable topological, functional, tissue and clinical positions.
The computational strategy integrates multiple views of the same gene set:
- Network topology: which nodes are highly connected and which act as intermediaries?
- Functional organization: which biological processes, cellular components and pathways are enriched?
- Tissue context: where are the genes associated with the most connected HPA terms?
- Phenotypic context: which human phenotypes are shared across genes?
- Clinical evidence: which highly central genes have documented disease associations?
flowchart LR
A[Curated 87-gene set] --> B[STRING v12.0]
A --> C[g:Profiler]
A --> D[HPA annotations]
A --> E[HPO associations]
B --> F[86-node PPI network]
F --> G[Degree + Betweenness]
G --> H[15-gene reference subset]
C --> I[GO / Reactome / KEGG / HPO enrichment]
D --> J[Gene–tissue bipartite network]
E --> K[Gene–phenotype bipartite network]
J --> L[Jaccard projection]
K --> M[Jaccard projection]
L --> N[Louvain communities]
M --> O[Louvain communities]
H --> P[OMIM + ClinVar + PubMed]
The final study set contains 87 protein-coding genes manually classified into:
- Structural core: 39 genes.
- Auxiliary factors: 20 genes.
- Regulatory factors: 28 genes.
The thesis describes inclusion and exclusion criteria based on spliceosome participation, predominantly nuclear localisation, evidence of a direct splicing-related role and representation in at least two consulted databases.
The complete reference dataset is available in data/reference/.
The PPI network was created in STRING v12.0 with Homo sapiens, combined confidence ≥ 0.700, and a zero-order network restricted to the supplied protein set.
Network topology was evaluated as an undirected network using degree centrality and betweenness centrality. A B/D ratio was also calculated as a complementary measure of relative intermediation.
SF3A2 ranked first in degree with 69. SRSF1 ranked first in betweenness with 268.84. A 15-gene reference subset was selected by combining centrality with functional diversity.
The enrichment analysis used g:Profiler, Homo sapiens as organism, and GO, Reactome, KEGG and HPO resources. Statistical significance was evaluated using the platform's g:SCS multiple-testing correction, with adjusted p < 0.05 as the significance threshold.
At the global level, the reported leading biological-process terms include RNA splicing, mRNA splicing via the spliceosome, mRNA processing and RNA processing.
HPA terms exported through g:Profiler were transformed from the intersections field into a long-format structure and used to construct gene–tissue bipartite networks and Jaccard similarity projections.
The global network contained 76 genes, 346 tissues and 12,527 bipartite connections. The highest reported gene degrees were HNRNPK and HNRNPM (305 each), SRSF3 (298), HNRNPA2B1 (285), HNRNPU (284) and SNRNP70 (283).
The complete supplied HPO genes_to_phenotype.txt association resource was integrated after mapping gene symbols to NCBI Entrez identifiers with org.Hs.eg.db.
The global HPO network contained 27 mapped genes, 516 phenotypes and 836 bipartite connections, with 14 Louvain communities after Jaccard projection.
The highest reported gene degrees were SF3B4 (82), HNRNPK (70), HNRNPH1 (59), SF3B2 (54) and CRNKL1 (53).
For the 15 genes selected as the topological reference subset, the thesis integrated evidence from OMIM, ClinVar and PubMed. Evidence was grouped into hereditary disease, cancer-related evidence, indirect functional evidence and no clear pathogenic association.
spliceosome-network-analysis-R/
├── README.md
├── START_HERE.md
├── CITATION.cff
├── NOTICE.md
├── data/
│ ├── reference/
│ ├── gprofiler/
│ ├── networkanalyst/
│ └── hpo/
├── scripts/
│ ├── analysis/
│ └── data/
├── results/
│ ├── HPA/
│ └── HPO/
├── figures/
│ ├── PPI/
│ ├── HPA/
│ └── HPO/
└── docs/
├── portfolio/
└── tfm/
The main R workflow is scripts/analysis/network_analysis.R.
Rscript scripts/analysis/network_analysis.RThe supplied R workflow performs the HPA/HPO analysis pipeline, including identifier mapping, bipartite network construction, centrality calculations, Jaccard projections, Louvain communities, Excel result generation and figures.
The original STRING acquisition and NetworkAnalyst calculations are preserved as documented parts of the academic project. The supplied R script does not recreate the original STRING interaction retrieval through an API or rerun the NetworkAnalyst web calculation. The repository therefore distinguishes between source/external-platform outputs, R-based processing and final interpretation.
See REPRODUCIBILITY.md for the detailed boundary.
| File | Purpose |
|---|---|
| START_HERE.md | Fast route through the repository |
| RESULTS_GALLERY.md | Main visual outputs |
| RESULTS_SUMMARY.md | Compact quantitative summary |
| RESULTS_BY_ANALYSIS.md | Results organised by analytical branch |
| METHODS.md | Compact technical methods |
| PROJECT_OVERVIEW.md | Project context and study design |
| REPRODUCIBILITY.md | Inputs, thresholds and reproducibility boundaries |
| REPOSITORY_PLAN.md | Architecture and maintenance logic |
| network_analysis.R | Main R workflow |
| TFM.pdf | Complete Master's Thesis |
Network centrality describes topological position, not biological causality. The thesis discusses database bias, dependence on the initial gene-selection criteria, the static representation of a dynamic spliceosome and the absence of experimental validation.
Disease or phenotype association should therefore not be read as proof that a gene is causally responsible for a pathology.
Student: Antonio José Bailén de la Cruz
Programme: Master's in Biotechnology and Biomedicine
Department: Biología Experimental
Date: 02/07/2026
See NOTICE.md for provenance and reuse information.



