Microarray Gene Expression Analysis of Fixed Archival Tissue Permits Molecular Classification and Identification of Potential Therapeutic Targets in Diffuse Large B-Cell Lymphoma Kim Linton, Christopher Howarth, Mark Wappett, Gillian Newton, Cynthia Lachel, Javeed Iqbal, Stuart Pepper, Richard Byers, Wing (John) Chan, John Radford The Journal of Molecular Diagnostics Volume 14, Issue 3, Pages 223-232 (May 2012) DOI: 10.1016/j.jmoldx.2012.01.008 Copyright © 2012 American Society for Investigative Pathology and the Association for Molecular Pathology Terms and Conditions
Figure 1 Scatter plots showing the correlation between biological replicates for log2 signal intensity calls (red, absent; yellow, marginal; green, present) on Plus 2.0 microarrays for FT samples (R2 = 0.706) (A) and FFPE samples (R2 = 0.585) (B). C: Scatter plot of log2 signal intensity data showing the raw correlation between FT and paired FFPE data across all samples and arrays (R2 = 0.57). The Journal of Molecular Diagnostics 2012 14, 223-232DOI: (10.1016/j.jmoldx.2012.01.008) Copyright © 2012 American Society for Investigative Pathology and the Association for Molecular Pathology Terms and Conditions
Figure 2 Standard Affymetrix algorithms were used to calculate the signal intensity of probe sets. Associated P values from all probe signal intensity data for a given probe set were then used to assign an Absent/Present/Marginal call. A: Scatter plot showing the signal distribution for all probe sets called Present in FT and FFPE arrays (blue), and the 9% and 12% of probe sets present exclusively in either FFPE (pink) or FT (yellow). B and C: Probe set distance distribution plots illustrating signal intensity (x axis) relative to probe set distance from the 3′ end of the transcript (y axis) for FFPE probe sets (B) and FT probe sets (C). The Journal of Molecular Diagnostics 2012 14, 223-232DOI: (10.1016/j.jmoldx.2012.01.008) Copyright © 2012 American Society for Investigative Pathology and the Association for Molecular Pathology Terms and Conditions
Figure 3 Heat maps illustrating unsupervised hierarchical clustering of 20 cases of DLBCL (12 GCB and 8 ABC subtypes); horizontal columns represent individual genes, vertical columns individual cases. Clustering of all present-filtered probe sets for FFPE arrays (A) and all present-filtered probe sets for FT arrays (B); distinct clusters containing ABC and GCB subtypes are observed for both FFPE and FT arrays, but separation is tightest for FFPE arrays, with only one ABC case clustering together with the GCB tumors. Clustering using 90 of the 100 Alizadeh genes cross-mapped to Plus 2.0 for FFPE arrays (C) and FT arrays (D); separation improves and is again superior in FFPE arrays where classification accuracy reaches 100%. Bayesian DLBCL Classifier in FFPE (E) and FT (F). The Journal of Molecular Diagnostics 2012 14, 223-232DOI: (10.1016/j.jmoldx.2012.01.008) Copyright © 2012 American Society for Investigative Pathology and the Association for Molecular Pathology Terms and Conditions
Figure 4 The average signal intensity values for all probe sets representing each of six NF-κB genes in ABC and GCB subtypes. Significant up-regulation of at least five genes can be seen in ABC-DLBCL. The Journal of Molecular Diagnostics 2012 14, 223-232DOI: (10.1016/j.jmoldx.2012.01.008) Copyright © 2012 American Society for Investigative Pathology and the Association for Molecular Pathology Terms and Conditions
Figure 5 Top 21 ABC/GCB discriminatory genes in FFPE gene expression profiles. The Journal of Molecular Diagnostics 2012 14, 223-232DOI: (10.1016/j.jmoldx.2012.01.008) Copyright © 2012 American Society for Investigative Pathology and the Association for Molecular Pathology Terms and Conditions