EUVIMEDEUVIMED entdecken
Crossref

Odporna klasyfikacja nowotworów na bazie binaryzacji profili mutacji somatycznych na poziome ścieżek genowych

Karolina Widzisz, Mateusz Kania, Joanna Żyła, Andrzej Polański

Digital Library of the Silesian University of Technology (Silesian University of Technology) · 01.01.2024

DOI: 10.34918/89989 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Digital Library of the Silesian University of Technology (Silesian University of Technology), erschienen am 01.01.2024. Verfasst von Karolina Widzisz und 3 weiteren Autor:innen.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study targets the problem of unsupervised clustering based on somatic mutation profiles for various cancer types.We classified somatic mutations in tumor DNA sequencing data by their presence in specific pathways using binary-encoding.We have compared eight unsupervised learning methods, i.e. the Bernoulli Mixture Model, K-means initialized BMM, K-means, K-means++, C-means, Manhattan distance C-means, Fuzzy C-means, and Hierarchical Clustering (HC) on binary-encoded somatic mutation profiles from TCGA.Next, the mutation level was transformed to the 1568 signalling pathways, used for cluster evaluation.We compared the clustering of 45 tumor pairs using performance metrics: ARI, Fowlkes-Mallows, Jaccard, Error Rate, Balanced Accuracy, and Matthews Correlation.Then we validated the results using χ 2 test with Benjamini-Hochberg correction and V-Cramer effect size.The analyses showed that binary-encoded data is robust to the clustering algorithm used.The HC algorithm's statistical inference identified 5 unique significant pathways for original labels and 186 for clustering labels in LGG and LUSC.These pathways are related to proteins associated with cancer progression and metastasis, like KRAS, MET, PTK2, and NF-kB, immune system activity, cell differentiation, apoptosis, and lung tissue.In conclusion, the binary-encoded clustering approach provides robust clustering based on somatic mutations, distinguishing cancers by their mutation profile in signalling pathways.

Schlagwörter

cancer classificationsomatic mutationsbinary encodingsignaling pathwaysunsupervised clusteringclustering algorithm comparisonTCGAglioblastomalung squamous cell carcinomapathway enrichment

Zitieren

APA

Widzisz, K., Kania, M., Żyła, J., & Polański, A. (2024). Odporna klasyfikacja nowotworów na bazie binaryzacji profili mutacji somatycznych na poziome ścieżek genowych. Digital Library of the Silesian University of Technology (Silesian University of Technology). https://doi.org/10.34918/89989

Vancouver

Widzisz K, Kania M, Żyła J, Polański A. Odporna klasyfikacja nowotworów na bazie binaryzacji profili mutacji somatycznych na poziome ścieżek genowych. Digital Library of the Silesian University of Technology (Silesian University of Technology). 2024. doi:10.34918/89989

BibTeX
@article{Widzisz2024,
  author = {Widzisz, Karolina and Kania, Mateusz and Żyła, Joanna and Polański, Andrzej},
  title = {Odporna klasyfikacja nowotworów na bazie binaryzacji profili mutacji somatycznych na poziome ścieżek genowych},
  journal = {Digital Library of the Silesian University of Technology (Silesian University of Technology)},
  year = {2024},
  doi = {10.34918/89989}
}
Kostenlos starten →