Limits...
A differentiation-based phylogeny of cancer subtypes.

Riester M, Stephan-Otto Attolini C, Downey RJ, Singer S, Michor F - PLoS Comput. Biol. (2010)

Bottom Line: To date, there is no objective systematic method to categorize tumor subtypes by maturation.We validate our methodology with expression data of leukemia, breast cancer and liposarcoma subtypes and then apply it to a broader group of sarcomas.This ranking of tumor subtypes resulting from the application of our methodology allows the identification of genes correlated with differentiation and may help to identify novel therapeutic targets.

View Article: PubMed Central - PubMed

Affiliation: Computational Biology Center, Memorial Sloan-Kettering Cancer Center, New York, NY, USA.

ABSTRACT
Histopathological classification of human tumors relies in part on the degree of differentiation of the tumor sample. To date, there is no objective systematic method to categorize tumor subtypes by maturation. In this paper, we introduce a novel computational algorithm to rank tumor subtypes according to the dissimilarity of their gene expression from that of stem cells and fully differentiated tissue, and thereby construct a phylogenetic tree of cancer. We validate our methodology with expression data of leukemia, breast cancer and liposarcoma subtypes and then apply it to a broader group of sarcomas. This ranking of tumor subtypes resulting from the application of our methodology allows the identification of genes correlated with differentiation and may help to identify novel therapeutic targets. Our algorithm represents the first phylogeny-based tool to analyze the differentiation status of human tumors.

Show MeSH

Related in: MedlinePlus

A phylogeny of liposarcoma subtypes.(a) The figure shows the consensus tree of liposarcoma subtypes. The tree is rooted with expression data of human mesenchymal stem cells (hMSC), and expression data of normal fat cells is included as well. The differentiation pathway from hMSC to normal fat cells is represented in purple. The pink points represent common ancestors of (sets of) subtypes. The boxed numbers specify bootstrap values of branches. The tree indicates that dedifferentiated liposarcoma is most similar to stem cells, followed by pleomorphic, myxoid, round-cell, and finally well-differentiated liposarcoma. (b) The figure shows a schematic representation of the correlation of adipogenesis to liposarcoma differentiation. In [6], human mesenchymal stem cells were differentiated in vitro to produce fat cells, and gene expression was measured for five different time points during the differentiation. The expression data of four different liposarcoma subtypes was then compared to the data obtained from the differentiation time course. This comparison identified dedifferentiated liposarcoma as the subtype most similar to stem cells, followed by pleomorphic, myxoid/round-cell, and well-differentiated liposarcoma. The correspondence between the results of our algorithm applied to gene expression datasets and these experimentally derived results serves as a validation of our methodology. Adapted from [6].
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2865519&req=5

pcbi-1000777-g004: A phylogeny of liposarcoma subtypes.(a) The figure shows the consensus tree of liposarcoma subtypes. The tree is rooted with expression data of human mesenchymal stem cells (hMSC), and expression data of normal fat cells is included as well. The differentiation pathway from hMSC to normal fat cells is represented in purple. The pink points represent common ancestors of (sets of) subtypes. The boxed numbers specify bootstrap values of branches. The tree indicates that dedifferentiated liposarcoma is most similar to stem cells, followed by pleomorphic, myxoid, round-cell, and finally well-differentiated liposarcoma. (b) The figure shows a schematic representation of the correlation of adipogenesis to liposarcoma differentiation. In [6], human mesenchymal stem cells were differentiated in vitro to produce fat cells, and gene expression was measured for five different time points during the differentiation. The expression data of four different liposarcoma subtypes was then compared to the data obtained from the differentiation time course. This comparison identified dedifferentiated liposarcoma as the subtype most similar to stem cells, followed by pleomorphic, myxoid/round-cell, and well-differentiated liposarcoma. The correspondence between the results of our algorithm applied to gene expression datasets and these experimentally derived results serves as a validation of our methodology. Adapted from [6].

Mentions: Our liposarcoma dataset includes 180 surgical samples that have been pathologically classified as 61 dedifferentiated, 52 well differentiated, 26 pleomorphic, 18 round cell, and 23 myxoid liposarcomas [27], [30]. Samples that were likely misclassified were filtered in previous studies, which is a pre-processing step critical for the outcome of the algorithm. For an FDR of the ANOVA filter of 0.01 after correction with the Benjamini-Hochberg method, we find 13,429 probe sets that are differentially expressed among the liposarcoma subtypes. Those sets are then used to construct an unrooted phylogenetic tree. To root the tree, we use expression data of mesenchymal stem cells and fully differentiated adipocytes. The resulting consensus tree is shown in Fig. 4a. The tree topology is stable with bootstrap values larger than 85%. Based on the consensus tree, the subtypes can be ordered by increasing dissimilarity from stem cells as dedifferentiated, pleomorphic, myxoid/round-cell, and well-differentiated liposarcoma (Fig. 4a). This order coincides with experimental results based on the gene expression observed during in vitro differentiation published earlier (Fig. 4b) [6]. By setting the p-value threshold of the Holm correction to 0.01, we obtain 7,290 differentially expressed probe sets; these probe sets generate a tree topology that is identical to the case described above with bootstrap values larger than 91.5% (data not shown). When rooting with embryonic stem cells, the branching between embryonic stem cells and the rest of the tree is less stable since the expression of embryonic stem cells differs considerably from all other samples (data not shown). To increase the stability of the tree, it is preferable to root with an outgroup that is relatively closely related to the investigated samples (in this case, mesenchymal stem cells; see also the section “Systematic analysis of methods and parameters”) [32]. Again we test how well the tree fits the distance matrix and observe a mean average percent standard deviation of 11.3%, which has been reported to be a good fit for the Pearson correlation distance [22]. Therefore, our methodology is also able to rank liposarcoma subtypes in the correct order according to their dissimilarity to stem cells.


A differentiation-based phylogeny of cancer subtypes.

Riester M, Stephan-Otto Attolini C, Downey RJ, Singer S, Michor F - PLoS Comput. Biol. (2010)

A phylogeny of liposarcoma subtypes.(a) The figure shows the consensus tree of liposarcoma subtypes. The tree is rooted with expression data of human mesenchymal stem cells (hMSC), and expression data of normal fat cells is included as well. The differentiation pathway from hMSC to normal fat cells is represented in purple. The pink points represent common ancestors of (sets of) subtypes. The boxed numbers specify bootstrap values of branches. The tree indicates that dedifferentiated liposarcoma is most similar to stem cells, followed by pleomorphic, myxoid, round-cell, and finally well-differentiated liposarcoma. (b) The figure shows a schematic representation of the correlation of adipogenesis to liposarcoma differentiation. In [6], human mesenchymal stem cells were differentiated in vitro to produce fat cells, and gene expression was measured for five different time points during the differentiation. The expression data of four different liposarcoma subtypes was then compared to the data obtained from the differentiation time course. This comparison identified dedifferentiated liposarcoma as the subtype most similar to stem cells, followed by pleomorphic, myxoid/round-cell, and well-differentiated liposarcoma. The correspondence between the results of our algorithm applied to gene expression datasets and these experimentally derived results serves as a validation of our methodology. Adapted from [6].
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2865519&req=5

pcbi-1000777-g004: A phylogeny of liposarcoma subtypes.(a) The figure shows the consensus tree of liposarcoma subtypes. The tree is rooted with expression data of human mesenchymal stem cells (hMSC), and expression data of normal fat cells is included as well. The differentiation pathway from hMSC to normal fat cells is represented in purple. The pink points represent common ancestors of (sets of) subtypes. The boxed numbers specify bootstrap values of branches. The tree indicates that dedifferentiated liposarcoma is most similar to stem cells, followed by pleomorphic, myxoid, round-cell, and finally well-differentiated liposarcoma. (b) The figure shows a schematic representation of the correlation of adipogenesis to liposarcoma differentiation. In [6], human mesenchymal stem cells were differentiated in vitro to produce fat cells, and gene expression was measured for five different time points during the differentiation. The expression data of four different liposarcoma subtypes was then compared to the data obtained from the differentiation time course. This comparison identified dedifferentiated liposarcoma as the subtype most similar to stem cells, followed by pleomorphic, myxoid/round-cell, and well-differentiated liposarcoma. The correspondence between the results of our algorithm applied to gene expression datasets and these experimentally derived results serves as a validation of our methodology. Adapted from [6].
Mentions: Our liposarcoma dataset includes 180 surgical samples that have been pathologically classified as 61 dedifferentiated, 52 well differentiated, 26 pleomorphic, 18 round cell, and 23 myxoid liposarcomas [27], [30]. Samples that were likely misclassified were filtered in previous studies, which is a pre-processing step critical for the outcome of the algorithm. For an FDR of the ANOVA filter of 0.01 after correction with the Benjamini-Hochberg method, we find 13,429 probe sets that are differentially expressed among the liposarcoma subtypes. Those sets are then used to construct an unrooted phylogenetic tree. To root the tree, we use expression data of mesenchymal stem cells and fully differentiated adipocytes. The resulting consensus tree is shown in Fig. 4a. The tree topology is stable with bootstrap values larger than 85%. Based on the consensus tree, the subtypes can be ordered by increasing dissimilarity from stem cells as dedifferentiated, pleomorphic, myxoid/round-cell, and well-differentiated liposarcoma (Fig. 4a). This order coincides with experimental results based on the gene expression observed during in vitro differentiation published earlier (Fig. 4b) [6]. By setting the p-value threshold of the Holm correction to 0.01, we obtain 7,290 differentially expressed probe sets; these probe sets generate a tree topology that is identical to the case described above with bootstrap values larger than 91.5% (data not shown). When rooting with embryonic stem cells, the branching between embryonic stem cells and the rest of the tree is less stable since the expression of embryonic stem cells differs considerably from all other samples (data not shown). To increase the stability of the tree, it is preferable to root with an outgroup that is relatively closely related to the investigated samples (in this case, mesenchymal stem cells; see also the section “Systematic analysis of methods and parameters”) [32]. Again we test how well the tree fits the distance matrix and observe a mean average percent standard deviation of 11.3%, which has been reported to be a good fit for the Pearson correlation distance [22]. Therefore, our methodology is also able to rank liposarcoma subtypes in the correct order according to their dissimilarity to stem cells.

Bottom Line: To date, there is no objective systematic method to categorize tumor subtypes by maturation.We validate our methodology with expression data of leukemia, breast cancer and liposarcoma subtypes and then apply it to a broader group of sarcomas.This ranking of tumor subtypes resulting from the application of our methodology allows the identification of genes correlated with differentiation and may help to identify novel therapeutic targets.

View Article: PubMed Central - PubMed

Affiliation: Computational Biology Center, Memorial Sloan-Kettering Cancer Center, New York, NY, USA.

ABSTRACT
Histopathological classification of human tumors relies in part on the degree of differentiation of the tumor sample. To date, there is no objective systematic method to categorize tumor subtypes by maturation. In this paper, we introduce a novel computational algorithm to rank tumor subtypes according to the dissimilarity of their gene expression from that of stem cells and fully differentiated tissue, and thereby construct a phylogenetic tree of cancer. We validate our methodology with expression data of leukemia, breast cancer and liposarcoma subtypes and then apply it to a broader group of sarcomas. This ranking of tumor subtypes resulting from the application of our methodology allows the identification of genes correlated with differentiation and may help to identify novel therapeutic targets. Our algorithm represents the first phylogeny-based tool to analyze the differentiation status of human tumors.

Show MeSH
Related in: MedlinePlus