FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W1570446714> ?p ?o ?g. }

• W1570446714 endingPage "no" @default.
• W1570446714 startingPage "no" @default.
• W1570446714 abstract "B-cell non-Hodgkin lymphomas (NHL) can be distinguished from each other according to the morphological and immunophenotypic features of malignant cells compared to their normal counterparts, as well as by the presence of specific chromosomal aberrations, considered to be key events in lymphomagenesis. Unsupervised clustering algorithms applied to gene expression profiling (GEP) datasets have identified new NHL subgroups, while supervised analyses enabled the characterization of specific differences among subsets (Iqbal et al, 2009). Array-based comparative genomic hybridization (Array-CGH) is a technique that allows genome-wide exploration for unbalanced copy number (CN) changes, namely DNA gains and losses (Maciejewski & Mufti, 2008). Array-CGH has been applied to NHL samples, identifying recurrent genomic changes (Iqbal et al, 2009). Recently, an application of supervised, and sequentially both supervised and unsupervised analysis on arrayCGH data demonstrated the possibility of differentiating between NHL subtypes based upon DNA CN changes (Ferreira et al, 2008; Takeuchi et al, 2009). In order to evaluate an unsupervised method with the best sensitivity for arrayCGH data and to assess mature B-NHL genetic similarity, we successively employed HC and non-negative matrix factorization (NMF) (Lee & Seung, 2001) with Kullback-Leibler divergence on a dataset of genomic profiles of 109 mature B-cell lymphoid neoplasms. The series included 25 diffuse large B-cell lymphoma (DLBCL) (20 clinical samples and five cell lines), 30 multiple myeloma (MM) (13 clinical samples and 17 cell lines), 20 post-transplant lymphoproliferative disorders (PTLD) samples (of which 13 cases were classified as DLBCL), 26 mantle cell lymphoma (MCL) cases (22 clinical samples and four cell lines), and eight follicular lymphoma (FL) samples (Rinaldi et al, 2006a,b; Lombardi et al, 2007). The Affymetrix GeneChip Human Mapping 10k arrays were used for all cases, as previously described (Rinaldi et al, 2006a,b). After discarding redundant probes showing an identical segmented estimated CN values across all samples in two or more consecutive probes, unsupervised analysis was performed using both NMF and hierarchical clustering (HC). Figure 1A shows the frequency of CN changes in the whole series. NMF unsupervised clustering identified the rank 4 as the most stable cluster structure (Fig 1B). Cluster 1 was characterized mainly by gains of chromosomes 3 (FOXP1, NFKBIZ, BCL6) and 7 (CDK6), of 6p (TNF), 13q (MIR17HG) and 18q (NFATC1, MALT1, BCL2), in combination with losses at 4q, 6q (PRDM1 TNFAIP3), 13q (MIR15/MIR16) and 17p (TP53) (Fig 1C). Cluster 2 presented mainly gains of 3q (BCL6, NFKBIZ), and losses at 1p, 6q (PRDM1, TNFAIP3), 9p (CDKN2A), 11q (ATM), 13q (MIR15/MIR16) and 17p (TP53). Cluster 3 had mainly gains of chromosomes 5, 12 and of the odd numbered chromosomes. Cluster 4 had gains of 1q, 8q (MYC), 11q (CCND1), 18q (NFATC1, MALT1, BCL2), 20q and losses of chromosome 13 (RB1, MIR15/MIR16), and at 1p and 17p (TP53). The NMF unsupervised clustering algorithm was able to divide relatively well the mixed group of B-cell lymphoid tumours based upon the pattern of DNA CN changes. Three clusters were clearly enriched in one of the main B-cell NHL types, while one appeared as a mixed group (Table I): 74% (17/23) of cluster 1 were DLBCL (both de novo and post-transplant-DLBCL), 56% (22/39) of cluster 2 MCL, 86% (19/22) of cluster 4 were MM. Moreover, 85% (22/26) of MCL were in cluster 2, and 63% (19/30) of MM were in cluster 4. DLBCL cell lines were classified in cluster 1 (80%, 4/5), together with the vast majority of DLBCL clinical specimens, while 14/17 (82%) MM cell lines were comprised in cluster 4 alongside 5/13 MM clinical specimens. The DLBCL clinical specimens did not cluster based upon their cell of origin, as determined based upon the algorithm of Hans et al (2004): 50% of both germinal centre B-cell-like and non-germinal centre B-cell-like DLBCL cases were together in cluster 1, possibly due to the relatively small number of samples analysed and to the fact that differences between the two DLBCL subtypes were smaller than the differences between DLBCL, MCL and MM. HC separated only the MM cases relatively well, with one cluster clearly containing the ‘13q-deleted’ cases and the other cluster containing MM cases with chromosomal gains of the odd numbered chromosomes. It did not succeed in separating the other NHL subtypes. Thus, we might conclude that, when applied for the analysis of arrayCGH data, NMF showed an advantage over HC, successfully dividing a mixed group of B-cell lymphoid tumours based upon the pattern of DNA CN changes. Copy number frequencies and unsupervised clustering of 109 B cell mature lymphoid neoplasms. (A) Frequency of gains/losses across the whole group of 109 B-non-Hodgkin lymphomas. Frequency of gains (positive y-axis) and frequency of losses (negative y-axis) versus genome position (x-axis). (B) Non-negative matrix factorization (NMF) consensus plots for rank 1–8 (from left to right, top to bottom, except last plot). Bottom-right: Plot of the cophenetic correlation coefficient for rank 1–8, indicating the most stable cluster at its maximum at rank = 4, which is also evident from the consensus plots. (C) Copy number changes distribution among the four NMF clusters. For each plot, frequency of gains (positive y-axis) and frequency of losses (negative y-axis) versus genome position (x-axis) are shown. NMF is a relatively new algorithm, which could be regarded as an Independent Components Analysis (ICA) variant but restricted to positive values (Lee & Seung, 2001). Its simple idea of restricting the factors to be positive, leads to desirable features of NMF. First of all, any genomic profile is represented by strictly positive features. This facilitates interpretation of the underlying components, in contrast to Principal Component Analysis (PCA) and other ICA-based methods where it is difficult to interpret the overall effect of higher order components. Secondly, NMF generally yields sparse and localized features (Lee & Seung, 2001), whereas component vectors of PCA/ICA extend across the entire space. Sparse and localized features better suit the application of genomic profiling because genomic alterations appear to occur in a localized but random fashion. Thirdly, the iterative nature of the NMF algorithm produces stability plots that are useful to determine the optimal number of clusters. HC is based on distance metrics, organizing objects into a tree structure. Weak points of HC are the high sensitivity to the metrics used for the match assessment and the inflexible structure of dendrogram, which lead to subjective evaluation of cluster numbers (Brunet et al, 2004). Although, at least in our hands, NMF appeared more sensitive with respect to arrayCGH data analysis, both HC and NMF did not reach a complete distinction among mature B-NHL, due the presence of recurrent lesions common to different lymphoma subtypes, and, possibly, to the inability of arrayCGH to detect the subtype-specific balanced chromosomal translocations. In conclusion, the identification of individual lymphoma subtypes was obtained using arrayCGH data and the NMF unsupervised clustering. The latter appeared more suitable for genomic data than HC. The application of NMF to larger arrayCGH datasets obtained with higher resolution platforms will probably identify new groups of patients, possibly sharing common altered pathways that might benefit from specific therapeutic approaches. Work supported by: Oncosuisse grant OCS 01517-02-2004; Cantone Ticino (‘Computational life science/Ticino in rete’ program); Fondazione per la Ricerca e la Cura sui Linfomi (Lugano, Switzerland); Cofin 2004, MIUR, Rome, Italy; Novara-AIL Onlus, Novara, Italy; Associazione Italiana Ricerca sul Cancro (AIRC); Italian Ministry of Health; Italian Ministry of University and Research grant FIRB RBAU01935A. E.C. is recipient of an European Society for Medical Oncology (ESMO) Fellowship Grant. The authors declare no conflicts of interests." @default.
• W1570446714 created "2016-06-24" @default.
• W1570446714 creator A5016760892 @default.
• W1570446714 creator A5039188583 @default.
• W1570446714 creator A5041745367 @default.
• W1570446714 creator A5042629318 @default.
• W1570446714 creator A5046270022 @default.
• W1570446714 creator A5066518317 @default.
• W1570446714 creator A5074436993 @default.
• W1570446714 creator A5079771818 @default.
• W1570446714 creator A5088819677 @default.
• W1570446714 creator A5090350899 @default.
• W1570446714 date "2010-04-28" @default.
• W1570446714 modified "2023-09-24" @default.
• W1570446714 title "Non-negative matrix factorization to perform unsupervised clustering of genome wide DNA profiles in mature B cell lymphoid neoplasms" @default.
• W1570446714 cites W1580240122 @default.
• W1570446714 cites W1978923828 @default.
• W1570446714 cites W2024355877 @default.
• W1570446714 cites W2052024543 @default.
• W1570446714 cites W2065602617 @default.
• W1570446714 cites W2113319573 @default.
• W1570446714 cites W2136787567 @default.
• W1570446714 cites W2155076036 @default.
• W1570446714 cites W2158267770 @default.
• W1570446714 doi "https://doi.org/10.1111/j.1365-2141.2010.08181.x" @default.
• W1570446714 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/20433677" @default.
• W1570446714 hasPublicationYear "2010" @default.
• W1570446714 type Work @default.
• W1570446714 sameAs 1570446714 @default.
• W1570446714 citedByCount "0" @default.
• W1570446714 crossrefType "journal-article" @default.
• W1570446714 hasAuthorship W1570446714A5016760892 @default.
• W1570446714 hasAuthorship W1570446714A5039188583 @default.
• W1570446714 hasAuthorship W1570446714A5041745367 @default.
• W1570446714 hasAuthorship W1570446714A5042629318 @default.
• W1570446714 hasAuthorship W1570446714A5046270022 @default.
• W1570446714 hasAuthorship W1570446714A5066518317 @default.
• W1570446714 hasAuthorship W1570446714A5074436993 @default.
• W1570446714 hasAuthorship W1570446714A5079771818 @default.
• W1570446714 hasAuthorship W1570446714A5088819677 @default.
• W1570446714 hasAuthorship W1570446714A5090350899 @default.
• W1570446714 hasConcept C104317684 @default.
• W1570446714 hasConcept C121332964 @default.
• W1570446714 hasConcept C141231307 @default.
• W1570446714 hasConcept C152671427 @default.
• W1570446714 hasConcept C154945302 @default.
• W1570446714 hasConcept C158693339 @default.
• W1570446714 hasConcept C41008148 @default.
• W1570446714 hasConcept C42355184 @default.
• W1570446714 hasConcept C54355233 @default.
• W1570446714 hasConcept C552990157 @default.
• W1570446714 hasConcept C62520636 @default.
• W1570446714 hasConcept C70721500 @default.
• W1570446714 hasConcept C73555534 @default.
• W1570446714 hasConcept C86803240 @default.
• W1570446714 hasConceptScore W1570446714C104317684 @default.
• W1570446714 hasConceptScore W1570446714C121332964 @default.
• W1570446714 hasConceptScore W1570446714C141231307 @default.
• W1570446714 hasConceptScore W1570446714C152671427 @default.
• W1570446714 hasConceptScore W1570446714C154945302 @default.
• W1570446714 hasConceptScore W1570446714C158693339 @default.
• W1570446714 hasConceptScore W1570446714C41008148 @default.
• W1570446714 hasConceptScore W1570446714C42355184 @default.
• W1570446714 hasConceptScore W1570446714C54355233 @default.
• W1570446714 hasConceptScore W1570446714C552990157 @default.
• W1570446714 hasConceptScore W1570446714C62520636 @default.
• W1570446714 hasConceptScore W1570446714C70721500 @default.
• W1570446714 hasConceptScore W1570446714C73555534 @default.
• W1570446714 hasConceptScore W1570446714C86803240 @default.
• W1570446714 hasLocation W15704467141 @default.
• W1570446714 hasLocation W15704467142 @default.
• W1570446714 hasOpenAccess W1570446714 @default.
• W1570446714 hasPrimaryLocation W15704467141 @default.
• W1570446714 hasRelatedWork W1966846766 @default.
• W1570446714 hasRelatedWork W2109085477 @default.
• W1570446714 hasRelatedWork W2147339185 @default.
• W1570446714 hasRelatedWork W2168039431 @default.
• W1570446714 hasRelatedWork W2273149514 @default.
• W1570446714 hasRelatedWork W2481706124 @default.
• W1570446714 hasRelatedWork W2572998321 @default.
• W1570446714 hasRelatedWork W4285239292 @default.
• W1570446714 hasRelatedWork W4298385783 @default.
• W1570446714 hasRelatedWork W83612550 @default.
• W1570446714 isParatext "false" @default.
• W1570446714 isRetracted "false" @default.
• W1570446714 magId "1570446714" @default.
• W1570446714 workType "article" @default.