be450_0413

be450_0413 - Self-organizing-map-based molecular Self based...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Self-organizing-map-based molecular Self based signature representing the development of hepatocellular carcinoma hepatocellular carcinoma Iizuka, N., et al. FEBS Letters, 2005. 579(5): p. 1089. Iizuka N., FEBS 2005. 579 Erin Bridgeford Erin Bridgeford Nancy Guillén BE.450 April 13, 2005 Microarrays to investigate Microarrays to problems in cell biology problems ♦ Data from transcription state of the cell Data under certain conditions under ♦ Each experiment produces lots of data ♦ Finding single change in gene expression ♦ Look at overall patterns of gene expression ♦ Hypothesis driven vs. fishing expedition Image removed due to copyright reasons. GeneChip® Microarray GeneChip Courtesy of Affymetrix. Used with permission. ♦ Probe is 25-mer oligonucleotide for high specificity ♦ Multiple probes for each expression or genotype measurement ♦ Optimized probe set Analysis of gene expression profiling data profiling ♦ Due to high volume of data, systematic Due methods for organization are required to convert data into a manageable set convert ♦ Strategies grouped in two categories: • Discrimination or supervised learning • Clustering or unsupervised learning (k-means, means, self-organizing-maps) self maps) ♦ Underlying biological phenomena might Underlying get lost in abstraction get Self-organizing maps for Self organizing clustering of expression data clustering ♦ SOM is a similarity graph, and a clustering diagram SOM ♦ Converts complex, nonlinear statistical relationships Converts between high-dimensional data items into simple between dimensional geometric relationships on a low-dimensional geometric dimensional display. ♦ SOM has a series of partitions with a predefined SOM geometrical configuration and, initially, their reference vectors are random reference ♦ Genes or samples are mapped to the relevant Genes partitions, depending on which reference vector they are most similar to ♦ Demo Gene expression profiles in hepatocellular carcinoma hepatocellular ♦ Microarray studies aimed at translating molecular studies information into clinical practice information ♦ Studies for breast cancer and large-B-cell lymphoma ♦ Studies generally include cohort of patients followed Studies for years after treatment for ♦ Link gene clusters with good or poor prognosis Link (survival, recurrence) ♦ HCC outcome complicated by the fact that cirrhosis HCC (pre-neoplastic) compromises liver functionality (pre ♦ Heterogeneous nature of human HCC ♦ Results have to provide rationale for a molecular Results classification of the tumor to be able to predict outcomes and guide treatments outcomes Hepatocellular carcinoma and hepatitis Hepatocellular carcinoma B (HBV) and C (HCV) viruses (HBV) ♦ Mutagenic effect of virus ♦ Chronic inflammation and disease Chronic leads to malignant neoplastic event neoplastic ♦ Molecular basis not well understood What are they trying to do with this? What ♦ Goal Understand the relation between development and Understand dedifferentiation of HCC dedifferentiation ♦ Hypothesis Disease progression: chronic HCV infection → well Disease well differentiated HCC → moderately differentiated HCC moderately → poorly differentiated HCC ♦ Approach Perform a comprehensive analysis of gene expression Perform levels and identify discriminatory genes for each stage to elucidate the molecular basis of HCC using a global picture of expression patterns picture Materials and Methods: Sample Selection Materials Samples taken from 76 HCC patients ♦ 50 seropositive for HCVAb 50 seropositive for HCVAb ♦ 26 seronegative for HCV 26 seronegative ♦ All seronegative for HBV surface antigen All seronegative ♦ Histopathology on HCV+ samples Histopathology • 7 well-differentiated HCC (group G1) • 35 moderately-differentiated HCC (G2) • 8 poorly-differentiated (G3) Control Groups Control ♦ Two control groups: ♦ Group L0 comprised of 6 nontumorous, Group nontumorous histologically normal liver samples from patients histologically normal with benign or metastatic liver tumors metastatic • • • • 1 focal nodular hyperplasia 2 hemangiomas hemangiomas 3 metastatic tumors (2 from colon cancer, 1 gastric) metastatic All seronegative for HCV Ab and HBVsAg All seronegative for Ab and HBVsAg ♦ Group L1: Five HCV-iinfected nontumorous nfected nontumorous samples from 5 HCC patients • Two chronic hepatitis • 3 liver cirrhosis ♦ Concerns in sample selection: • No normal samples of liver as baseline • No samples from HCV+ patients without HCC Materials and Methods: DNA Microanalysis Materials ♦ Resected specimens specimens divided in two groups: divided • One frozen immediately One after surgery for later RNA extraction extraction • One preserved in 10% One formaldehyde and embedded in paraffin embedded Used to demonstrate that non-necrotic that necrotic tissues were source of RNA RNA ♦ RNA extraction RNA performed performed ♦ Quality control of RNA: • Look for genomic DNA Look contamination contamination • Check for RNA decay by Check agarose gel agaros gel electrophoresis If ratio of 28S/18S rRNA iis rRNA s around 2.0, suggests RNA had not decayed before or during extraction during Reduced 28S/18S ratios indicate poor quality RNA Materials and Methods: Microarray Analysis Microarray ♦ Synthesis of cDNA and cRNA (see Iizuka et al, Cancer Research Synthesis cDNA and cRNA (see Iizuka et 62, 2002) 62, ♦ Oligonucleotide microarray screening • huU95A DNA Chips (12,600 probes that correspond to 8900 huU95A named genes) for initial screen named Image removed due to copyright reasons. Materials and Methods: Gene Selection Materials ♦ At first pass, 3559 genes selected • Expression levels were greater than 40 arbitrary units Expression (arbitrary units = intensity/brightness of probed spot over brightness of local background) over ♦ Fisher ratio applied to evaluate which genes Fisher could help discriminate among the groups: • Measures the difference between two means Measures normalized by the average variance (ie, estimates normalized estimates signal-to-noise ratio). signal noise • Larger Fisher ratio suggests a stronger likelihood for a Larger gene’s ability to discriminate between groups. gene Gene selection cont’d Gene ♦ Random permutation test perfomed to validate Random perfomed to Fisher ratio : • Looks to find undesired structure in random data. • If original result is due to chance, then randomly If relabelling data should achieve similar ratios relabelling • Genes with P<0.005 were selected ♦ Different numbers of genes for each group Different deemed discriminatory: deemed • • • • L0 to L1: 152 genes L1 to G1: 191 genes G1 to G2: 54 genes G2 to G3: 40 genes Materials and Methods: Identifying Discriminatory Genes Discriminatory ♦ Percentage of genes identified by chance Percentage (false discovery rate) calculated (false ♦ Ratio of false positives/total positives ♦ A high FDR value can still be meaningful Materials and Methods: Comparing Classes Materials ♦ For class comparison, ♦ Self-organizing map For minimum distance • Algorithm used for Algorithm clustering data clustering classifier designed • Provides visualization Provides with top 40 genes of multi-dimensional of dimensional from each class: from • Finds centers of Finds classes and measures between those centers and a test image’s and center center data data Materials and Methods: Some Concerns Materials ♦ No indication that laser capture microdissection (LCM) or No microdissection (LCM) any more precise method of tissue selection was used: analyzing stroma and vasculature as well stroma ♦ Technology does not always generate reproducible or Technology consistent results even with optimized samples. • Per Stearns, “The current state of the art provides 5−10% Per 10% variation in signal intensities among replicate array elements on variation the same microarray, and 10−30% variation among microarray 30% corresponding array elements on different microarrays.1” corresponding 1Stears Stears et al, Trends in Microarray Analysis, Nature Medicine, 9 (140-145), 2003 Microarray Materials and Methods: Concerns, cont’d Materials ♦ Several manipulations of data required to Several estimate and select genes of interest estimate ♦ Each step can introduce assumptions/bias of Each authors in selection authors ♦ May select out biologically relevant data ♦ Assumption of certain percentage of false Assumption positives positives Table 1: Clinicopathologic characteristics of 50 HCV-positive HCCs Clinicopathologic positive HCCs Well (G1)* Moderately (G2)* Poorly (G3)* 4 3 24 11 6 2 65.3 + 7.0 - 65.4 + 7.1 - 67.2 + 9.5 - Primary lesion Single tumor Multiple tumors 6 1 15 20 2 6 Capsule formation Present Absent 4 3 29 6 6 2 Tumor size (cm)** 2.0 + 0.8 - 5.0 + 3.2 - 6.0 + 7.0 - Stage* I II IIIA/IV 6 1 0 10 17 8 2 3 3 Microscopic venous invasion* (-) (+) 7 0 21 14 3 5 Alpha-feto protein (ng/ml) < or = 100 > 100 6 1 24 11 3 5 Non-tumorous liver Normal or chronic hepatitis Liver cirrhosis 2 5 15 20 2 6 Factors Sex Male Female Age (year)** P-value P = 0.8007 P = 0.9612 (G1 vs. G2) P = 0.6595 (G1 vs. G3) P = 0.5406 (G2 vs. G3) P = 0.0568 P = 0.3339 P = 0.0007 (G1 vs. G2) P = 0.0279 (G1 vs. G3) P = 0.6397 (G2 vs. G3) P = 0.0656 P = 0.0381 P = 0.1504 P = 0.7569 Fisher's exact test, Student's t test and Mann-Whitney's U test were used to elucidate differences in backgrounds between each group. * Tumor differentiation, stage, and microscopic venous invasion were determined on the basis of TNM classification of UICC. G1-G3 tumors are equal to types I-III of Edmondson and Steiner classification, respectively. ** Mean + S.D. - Figure by MIT OCW. Table 1: Clinicopathologic characteristics Clinicopathologic Tumor size • Significantly larger in groups G2 and G3 Significantly compared with G1 compared • Significance determined by Mann-Whitney U Whitney test Tumor invasiveness Tumor • No vessel involvement in group G1 • Significantly more frequent vessel involvement Significantly in G2 and G3 in Clinicopathologic characteristics, Clinicopathologic characteristics, cont’ cont ♦ Tumor stage • Tended to be more advanced from G1 to G3 P = 0.066 by Fisher’s exact test: (borderline 0.066 exact value?) value?) ♦ Based on clinicopathologic characteristics, Based clinicopathologic characteristics, authors posit that HCC develops sequentially from L0 to L1 up to G3 sequentially Figure 1: Discriminatory Genes in Development of HCC Figure Image removed due to copyright reasons. Please see: Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma." FEBS Letters 579, no. 5 (February 14, 2005): 1089-100. Figure 1a: Discriminatory Genes Genes Reading the array: Reading • Each row is a gene • Each column is a group/sample • Red is upregulated, green Red upregulated green downregulated downregulated Figure 1a: ♦ 152 differentially expressed 152 genes genes ♦ Criteria for selection: Criteria • Downregulated genes: fold genes: change of L1 vs L0 <1 vs 85 downregulated downregulated • Upregulated genes: fold change genes: of L1 vs L0 >1 vs L0 67 upregulated upregulated Figure 1a: L1 vs L0 Image removed due to copyright reasons. Please see: Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma." FEBS Letters 579, no. 5 (February 14, 2005): 1089-100. Table 2: Downregulated genes from L1 compared with L0 Table ♦ Dystrophin: anchors anchors cytoskeleton to cell membrane. • Absence increases cell Absence permeability • May get lysis, more May lysis more inflammation inflammation ♦ Fibronectin: • • • • Tissue repair Embryogenesis Blood clotting Cell migration/adhesion ♦ Many genes with unknown functions Fisher Ratio GB Number Description Symbol Locus Function Eighteen genes downregulated in L1 in comparison with L0 Xp21.2 Cytoskeleton 6p24.3 Unknown ZNF337 20p11.1 Unknown X123 9q13-q21 Unknown DMD 50.45 M18533 Dystrophin 23.02 AF035316 Homolog to tubulin beta chain 20.65 AL049942 Zinc finger protein 337 18.34 L27479 Friedreich ataxia region gene X123 Extracellular matrix Fibronectin (Alt. Splice 1) 16.63 16.13 U19765 Zinc finger protein 9 ZNF9 3q21 Transcription/retroviral nucleic acid binding protein 14.91 X55503 Metallothionein IV MTIV 16q13 Detoxification 13.71 AL046394 Poly(rC) binding protein 3 PCBP3 21q22.3 RNA-binding protein/ post-transcriptional control 12.56 AB007886 KIAA0426 gene product KIAA0426 6p22.2-p21.3 Unknown 12.41 AL050139 Hypothetical protein FLJ13910 FLJ13910 2p11.1 Unknown Top-40 Discriminatory Genes in L0 and L1 Figure by MIT OCW. 22 genes upregulated in L1 compared with L0 upregulated ♦ Many are inflammatory in nature • Secondary to HCV infection? Secondary • Possible relationship to other oncogenic process? Possible oncogenic process? Upregulation of Ras suspcious? Upregulation of Ras • Normal liver baseline or HCV-iinfected liver only would be useful nfected comparison comparison Table 2 40.49 AI362017 Cystatin C CST3 20p11.21 Cysteine protease inhibitor 21.66 L13977 Prolylcarboxypeptidase (angiotensinase C) PRCP 11q14 Metabolism/lysosomerelated protein 20.59 D32053 Lysyl-tRNA synthetase KARS 16q23-q24 Protein biosynthesis 13.70 AF038962 Voltage-dependent anion channel 3 VDAC3 8p11.2 Transport of adenine nucleotides 11.90 AL008726 Protective protein for beta-galactosidase (cathepsin A) PPGB 20q13.1 Lysosomal protein/ enzyme activator 11.71 J03909 Interferon, gamma-inducible protein 30 IFI30 19p13.1 Lysosomal thiol reductase/IFN-inducible 11.32 Z69043 Signal sequence receptor, delta SSR4 Xq28 Translocatation of newly synthesized polypeptides 11.17 AL080080 Thioredoxin-related transmembrane protein TXNDC 14q21.3 Redox reaction 11.15 M63138 Cathepsin D CTSD 11p15.5 Lysosomal aspartyl protease/proteolysis 11.12 L09159 Ras homolog gene family, member A ARHA 3p21.3 Oncogenesis/actin cytoskeleton Twenty-two Genes Upregulated in L1 in Comparison with L0 Figure by MIT OCW. Figure 1b: Differential Expression of G1 compared with L1 compared Figure 1b: L1 to G1 L1 to G1: 191 genes L1 differentially expressed differentially • 95 upregulated in G1 95 upregulated Types include signal transduction, transcription, and RNA processing, RNA ATOX1 increase noted in previous study with HCVprevious related HCC • 96 downregulated in G1 96 downregulated Includes tumor suppressor/apoptotic genes (BCL2, IGFB3), Cell proliferation genes (FOS and IGFBP4 – may also be may associated with apoptosis) associated Image removed due to copyright reasons. Please see: Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma." FEBS Letters 579, no. 5 (February 14, 2005): 1089-100. Figure 1c: G1 to G2 G1 considered wellG1 differentiated, G2 differentiated, moderately differentiated moderately ♦ 54 genes differentiallyexpressed ♦ 25 genes upregulated iin 25 upregulated n G2 G2 ♦ 15 genes 15 downregulated in G2 downregulated • Many related to protein Many modification, transcription, and translation translation Image removed due to copyright reasons. Please see: • Many IFN-related related genes genes OAS2 (antiviral protein) protein) STAT1 (transcription pathway) pathway) PSME1 (proteolysis) PSME1 Suggestive of decreased immune response response Earlier paper noted IFN-iinducible genes IFN nducible in HCV-related HCC in related but not HBV-related but rel HCC HCC Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma." FEBS Letters 579, no. 5 (February 14, 2005): 1089- 100. Figure 1d: G2 to G3 Figure G3 considered poorly differentiated • More vascular invasion, larger tumor Image removed due to copyright More size than G1 reasons. Please see: ♦ 40 genes differentially expressed Iizuka, N., et al. "Self-organizing♦ 10 genes upregulated in G3 10 upregulated map-based molecular signature • LGALS9 (galectin; associated with associated representing the development of cell adhesion, growth regulation, hepatocellular carcinoma." FEBS apoptosis, metastasis) apoptosis, • TGFB1 (may trigger invasiveness of Letters 579, no. 5 (February 14, TGFB1 2005): 1089-100. HCC cells via integrin) integrin ♦ 30 genes downregulated in G3 30 downregulated • SDCI (cell adhesion, metastasis) SDCI Another study found decreased levels found in HCC with high metastatic potential) metastatic Figures 1e-h: 40 most discriminatory genes Figures h: for each transition for Image removed due to copyright reasons. Please see: Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma." FEBS Letters 579, no. 5 (February 14, 2005): 1089-100. Most discriminatory genes determined by looking at Most greatest differential expression (highest Fisher ratios) from one transition to another (L0 to L1, etc) from Figure 1e-h Figure ♦ Almost no overlap between discriminatory genes Almost in all groups: 17 out of 437 (0.39%) in ♦ False discovery rate (FDR) (i.e., genes identified False by chance) of all groups were all extremely low • • • • L0 vs L1: FDR of 0% L0 vs L1 vs G1: 0% L1 vs G1 vs G2: 0.24% G1 vs G2 vs G3: 0.29% G2 vs ♦ Overall trend towards smaller numbers of Overall significant genes as identified by this technology as the cancer becomes more advanced (i.e., dedifferentiation continues) dedifferentiation • Possible significance of this? Significance of selected genes Significance To verify significance of selected genes, To authors constructed the Minimum Distance classifier classifier • Quick recap: The minimum distance Quick classifier finds centers of classes and measures between those centers and a test image’s center. The distance is defined as an image center. index of similarity so that the minimum distance is identical to the maximum similarity. Figure 2 Figure Authors’ classification results: Image removed due to copyright reasons. a) 92% accuracy b) 98% accuracy c) 84% accurad) 100% accuracy Please see: Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma." FEBS Letters 579, no. 5 (February 14, 2005): 1089-100. • High accuracy reported between classes • Are results/authors’ conclusions of pre- and posttransition discrimination and grouping of molecular transition signatures reasonable? Arrangements of samples by SOM Arrangements ♦ 61 samples mapped according to 61 expression levels of the top 40 genes for each transition (total = 160 genes) each ♦ G2 tumors classified into two subtypes: • Without venous invasion • With venous invasion ♦ Tumor size assigned to samples ♦ p53 abnormality data applied to 22 of the p53 HCC samples Figure 3: Visualization of sample arrangement by SOM: development and classification by Image removed due to copyright reasons. Please see: Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma." FEBS Letters 579, no. 5 (February 14, 2005): 1089-100. Figure 3 results Figure ♦ 8x5 cells on hexagonal grid – 40 clusters 8x5 ♦ Clusters showed a sigmoidal curve in the Clusters sigmoidal curve order L0, L1, G1, G2, G3 order ♦ G2 w/o venous invasion closer to G1 ♦ G2 w/ venous invasion closer to G3 Figure 4: Tumor size and p53 status Image removed due to copyright reasons. Please see: Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma.“ FEBS Letters 579, no. 5 (February 14, 2005): 1089-100. Figure 4 results Figure ♦ Tumor size not always consistent with Tumor differentiation state differentiation ♦ HCCs become progressively less differentiated become as they enlarge as ♦ HCCs with WT-p53 located within or close to G1 p53 clusters clusters ♦ Most HCCs with mutant p53 located at most Most HCCs with distant points from L0, L1 and G1 clusters distant ♦ Genetic abnormality of p53 is a feature of late Genetic stage HCC stage Arrangement of HCV - / HCC HCC samples by SOM samples Image removed due to copyright reasons. Please see: Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma.“ FEBS Letters 579, no. 5 (February 14, 2005): 1089-100. Figure S3 results Figure ♦ SOM for HCV –/ HCCs failed to arrange SOM HCCs failed samples sequentially according to differentiation state differentiation ♦ Changes in identified discriminatory genes Changes are specific for HCV +/ HCCs HCCs Validation of microarray data by microarray data quantitative RT-PCR quantitative ♦ One discriminatory gene for each transition One selected at random to validate microarray data microarray data by analysis with real time RT-PCR by • • • • CD74 for L0→L1 IGFBP3 for L1→G1 STAT1 for G1→G2 TGFB1 for G2→G3 ♦ Abundance of each transcript calculated the as Abundance the mean copy number per 100 ng RNA for each ng RNA tissue ♦ Data compared by Student’s t test or Mann– Whitney U test and Pearson’s correlation correlation coefficient. coefficient. Validation of microarray data by microarray data quantitative RT-PCR quantitative Image removed due to copyright reasons. Please see: Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma." FEBS Letters 579, no. 5 (February 14, 2005): 1089-100. Comparison of expression patterns as measured by microarray and RT-PCR microarray Image removed due to copyright reasons. Please see: Iizuka, N., et al. "Self-organizing-map-based molecular signature representing the development of hepatocellular carcinoma." FEBS Letters 579, no. 5 (February 14, 2005): 1089-100. RT-PCR validation results RT ♦ Expression patterns of CD74, IGFBP3, Expression STAT1 and TGFB1 reproduced by real time quantitative RT-PCR time ♦ Is the data accurately reproduced ? Conclusions ♦ Differential genetic expression for each Differential stage of development with characteristic molecular signature molecular ♦ No overlap for discriminatory genes for No each transition each ♦ Patterns valid for HCV+/HCC only ♦ Provide additional biomarkers • Diagnosis and treatment Questions and concerns about the paper Questions ♦ Sample and tissue concerns • Isolated cancer cells? • Lack of HCV infected tissue from non Lack cancerous patients cancerous ♦ Introduced bias from statistical Introduced manipulations manipulations ♦ Variability and lack of reproducibility ♦ Choosing SOM to present data Choosing ♦ Occult HBV infection because of lack of Occult data for core antigen and viral DNA data Further discussions ...
View Full Document

Ask a homework question - tutors are online