Sökning: WFRF:(Torabi Moghadam Behrooz) >
An unsupervised app...
An unsupervised approach subgroups cancer types by distinct local DNA methylation patterns
-
- Torabi Moghadam, Behrooz (författare)
- Uppsala universitet,Institutionen för cell- och molekylärbiologi,Computational Biology and Bioinformatics
-
- Zamani, Neda (författare)
- Uppsala universitet,Institutionen för medicinsk biokemi och mikrobiologi,Genomics
-
- Gao, Jiangning (författare)
- Uppsala universitet,Institutionen för medicinsk biokemi och mikrobiologi,Genomics
-
visa fler...
-
- Meadows, Jennifer (författare)
- Uppsala universitet,Institutionen för medicinsk biokemi och mikrobiologi,Genomics
-
- Sundström, Görel (författare)
- Uppsala universitet,Institutionen för medicinsk biokemi och mikrobiologi,Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, Umeå, Sweden,Genomics
-
- Holmfeldt, Linda (författare)
- Uppsala universitet,Institutionen för immunologi, genetik och patologi,The Beijer Laboratory for gene and neurosciences
-
- Komorowski, Jan (författare)
- Uppsala universitet,Institutionen för cell- och molekylärbiologi,Institute of Computer Science, Polish Academy of Sciences, Warsaw, 01248, Poland,Computational Biology and Bioinformatics
-
- Grabherr, Manfred (författare)
- Genomics
-
visa färre...
-
(creator_code:org_t)
- Engelska.
- Relaterad länk:
-
https://urn.kb.se/re...
Abstract
Ämnesord
Stäng
- Cancer is one of the most common causes of death in humans. It can arise from many different cell types, and even cancers originating from the same tissue can constitute a heterogeneous group of diseases. While cytogenetics, the analysis of mutations and karyotypic alterations, has greatly improved the accuracy of diagnosis, it is likely that there are more categories in which cancers can be divided than is known today. Moreover, new biomarkers confirming existing classification schemes are desirable. Here, we interrogated the DNA methylation (DNAm) landscape as a novel indicator for discerning cancer subtypes.We developed and applied an unsupervised method, methylSaguaro, which is based on the combination of a Hidden Markov Model and a Neural Net. We first compared the concept of hypothesizing patterns and grouping to statistical methods that require a priori hypotheses to perform enrichment tests. We then analyzed samples from four cancer groups, Gliomas, Chronic Lymphocytic Leukemia (CLL), Renal Cell Carcinomas (RCC), and Acute Myeloid Leukemia (AML). On gliomas and CLL, we confirmed known cancer groupings in DNAm that perfectly correspond to known mutations. On Renal Cell Carcinomas, our method disagrees with the histological classification on 4% of the samples, and finds a novel cluster, suggesting that there might be a novel subtype that was hitherto unknown. On AML, methylSaguaro spreads the samples out on a continuous spectrum, enriching one end with patients assessed as having “poor” risk based on cytogenetics, but indicating that DNAm patterns would suggest a different risk assessment. Since methylSaguaro reports both the patterns and the specific sites behind the signals, we analyzed regions and genes indicative of subtypes across the cancers, revealing 41 genes affected by alterations in more than one cancer. In summary, we expect that DNAm, coupled with a hypothesis-free analysis method, will add to the set of clinical instruments to diagnose, assess, and treat cancer.
Ämnesord
- NATURVETENSKAP -- Data- och informationsvetenskap -- Bioinformatik (hsv//swe)
- NATURAL SCIENCES -- Computer and Information Sciences -- Bioinformatics (hsv//eng)
Nyckelord
- unsupervised learning
- DNA methylation
- cancer subtyping
Publikations- och innehållstyp
- vet (ämneskategori)
- art (ämneskategori)
Hitta via bibliotek
Till lärosätets databas