Supplemental Materials for:
Gene
Expression Overlap affects
Karyotype Prediction in Pediatric ALL
(Martin et al., 2007)
Table of Contents
Description
Methods
Experimental Protocol
Computational Analysis
VxInsight Analysis
Tables
Table S1.1. VxInsight gene list separating group A.
Table S1.2. VxInsight gene list separating group B.
Table S1.3. VxInsight gene list separating group C.
Table S1.4. VxInsight gene list separating group X.
Table S1.5. VxInsight gene list separating group Y.
Table S1.6. VxInsight gene list separating group Z.
Table S1.7. VxInsight gene list separating group T1.
Table S1.8. VxInsight gene list separating group T2.
Table S2.1. Performance evaluation for karyotype classification predictions.
Table S3.1. Combined gene/EST list for predicting t(12;21)[TEL-AML1].
Table S3.2. Combined gene/EST list for predicting t(4;11)[AF4-MLL].
Table S3.3. Combined gene/EST list for predicting t(1;19)[E2A-PBX1].
Table S3.4. Combined gene/EST list for predicting t(9;22)[BCR-ABL].
Table S3.5. Combined gene/EST list for predicting hperdiploid.
Table S4.1. Overlap of combined gene lists with gene lists from Yeoh et al.
Table S4.2. Genes in common with the Yeoh et al. χ2 list for t(12;21)[TEL-AML1].
Table S4.3. Genes in common with the Yeoh et al. χ2 list for t(4;11)[AF4-MLL].
Table S4.4. Genes in common with the Yeoh et al. χ2 list for t(1;19)[E2A-PBX1].
Table S4.5. Genes in common with the Yeoh et al. χ2 list for t(9;22)[BCR-ABL].
Table S4.6. Genes in common with the Yeoh et al. χ2 list for hyperdiploid>50.
Titles and legends to figures
Figure S1.1. VxInsight on Full Dataset.
Figure S1.2. Distribution of ALL patients according to VxInsight Cluster.
Figure S2.1. Distribution of genes from combined and χ2 lists for t(12;21)[TEL-AML1].
Figure S2.2. Distribution of genes from combined and χ2 lists for t(4;11)[AF4-MLL].
Figure S2.3. Distribution of genes from combined and χ2 lists for t(1;19)[E2A-PBX1].
Figure S2.4. Distribution of genes from combined and χ2 lists for t(9;22)[BCR-ABL].
Figure S2.5. Distribution of genes from combined and χ2 lists for hyperdiploid>50.
Figures
References
Description
This supplement contains descriptions and links to descriptions of the methods used in our study. These methods were used to produce the experimental data and analysis in the main text as well as the tables and figures in this supplement. The tables in this supplement include gene lists that differentiate the VxInsight clusters (Tables S1.1-S1.8); a table assessing the accuracy of our classification results (Table S2.1); combined gene lists for differentiating the karyotypes (Tables S3.1-S3.5); a table comparing our combined gene lists with the gene lists provided in Yeoh et al. (1)(Table S4.1); and gene lists giving the overlap of our gene lists with the lists produced in Yeoh et al. (Tables S4.2-S4.6). The figures in this supplement include visualizations for VxInsight (Figures S1.1-S1.2) and figures showing the distribution of genes from combined and χ2 lists for Yeoh et al. (Figures S2.1-S2.5).
Methods
Experimental Protocol
Informed consent from the parents, guardians, and patients was obtained for sample acquisition. Pre-treatment bone marrow (BM) or peripheral blood (PB) samples were acquired at the time of diagnosis from pediatric ALL patients registered to the Pediatric Oncology Group (POG) ALinC15 and ALinC16 therapeutic trials (8602, 9005, 9006, 9201, 9405, 9406, and 9605). The details of these clinical trials have been previously reported(2-4). The diagnosis of ALL was made on morphologic evaluation of bone marrow or peripheral blood and was confirmed by central review. Conventional cytogenetic banding and fluorescence in-situ hybridization were used to detect cytogenetic abnormalities in all patients at the POG Cytogenetics Reference Laboratory (University of Alabama, Birmingham), with the exception of the t(12;21) [TEL-AML1] translocation, which was detected using reverse-transcriptase polymerase chain reaction (RT-PCR) (5).
RNA was isolated from sterile, viable cryopreserved leukemic cell suspensions that had been stored at -135oC. After thawing, cells were washed once with Hanks’ balanced salt solution and total RNA was isolated with Qiagen RNeasy. Wright’s stained cytospins were also prepared from a small aliquot of the thawed cells and the percentage of leukemic blasts in the cell suspension was confirmed by light microscopy; only samples with 80-100% leukemic blasts were retained in the study. Total RNA was quantified with the RiboGreen assay (Molecular Probes). All RNA samples retained in the study were of high quality (A260nm/A280nm ratios > 1.8, with no RNA degradation (28S rRNA:18S rRNA > 2) or DNA contamination as assayed on RNA NanoChips (Agilent)).
To enhance the likelihood of detection of low abundance genes that may have been missed without linear amplification, cRNA was prepared from total RNA in the same manner as for expression profiling of hematopoietic stem and progenitor cells (6). Briefly, mRNA in the total RNA sample was reverse transcribed into cDNA, followed by re-transcription using two rounds of linear amplification as reported by Ivanova et al. (6), with the following modifications: 10 g/ml of linear acrylamide (Ambion) was used as a co-precipitant in steps that used alcohol precipitation and 2.5 g of total RNA from patient leukemic samples was used as the starting material. Fifteen micrograms of the resulting cRNA was fragmented at 95°C for 35 minutes and then hybridized in MES buffer (2-[N-Morpholino] ethanesulfonic acid) containing 0.5 mg/ml acetylated bovine serum albumin (Sigma) at 45°C for 20 hours to HG_U95Av2 oligonucleotide microarrays (Affymetrix). The hybridized probe arrays were washed and stained with the EukGE-WS2 fluidics protocol (Affymetrix), including streptavidin phycoerythrin conjugate (SAPE, Molecular Probes) followed by an antibody amplification step (Anti-streptavidin, biotinylated) (Vector Labs) and a second streptavidin phycoerythrin conjugate staining. The HG_U95Av2 chips were scanned and analyzed at 488 nm following the Affymetrix Microarray Suite (MAS) Version 5.0 Software. The data from each array was scaled to minimize variation due to sample preparation, hybridization conditions, staining, or array manufacturing lot.
The criteria used for final inclusion of patient samples in this study included: a minimum of 80% leukemic blasts, sufficient (2.5 g) total RNA of high quality, high quality cRNA, probe array image inspection, B2 oligonucleotide staining (used for Array grid alignment), and high quality of array hybridization for cRNA that has undergone double amplification (GAPDH value > 1800; >10% expressed genes; and 3’/5’ amplification ratios < 4). Of 311 pediatric ALL cases initially selected for the case-control study, 254 patient samples were ultimately retained. Unfortunately, one case-control cohort (hypodiploidy) had insufficient sample numbers, preventing our effective modeling of that group. Samples were excluded due to insufficient or poor quality RNA (13 cases); < 80% blasts (24 cases); or poor quality array hybridization (20 cases). Thus, high quality gene expression data was obtained on 254/274 (93%) cases that underwent array hybridization. All Affymetrix microarray signal and CEL data, together with covariate clinical, cytogenetic, and annotated experimental information is available at the National Cancer Institute Cancer Array Informatics website ( Experiment ID 1015897590271440).
Computational Analysis
The 254 patient dataset was first pre-processed by the removal of control probe sets (AFFX accession IDs), and probe sets with no “present” calls, as determined by the Affymetrix MAS 5.0 statistical software. After this process, 8,943 of the original 12,625 Affymetrix HG_U95Av2 probe sets were retained for analysis. Next a base-10 logarithmic transformation of the gene expression data was performed. This logarithmic transformation scales the data to assist in visualizations, remedies right-skewed distributions and makes error components additive (7). Log transformation was used for all data analysis methods except VxInsight, which used Savage scores instead (8).
We analyzed the processed gene expression data in two steps. First, we used unsupervised methods, including Principal Component Analysis (9, 10) (PCA) and VxInsight (11-14), a data visualization environment with a clustering algorithm based on force-directed graph layout. Unsupervised methods allow the discovery and visualization of unknown groups and relationships in the data. Our data was visualized using the first three principal components of PCA, as well as the layout given by VxInsight. Second, the dataset was subjected to various supervised learning techniques. Supervised techniques require the division of the dataset into a training and test set. The 254 patient dataset was divided into a 167 patient training set and an 87 patient test set. The training set was selected at random but was balanced so that the distribution of known karyotypes (including t(12;21)[TEL-AML1], t(4;11)[AF4-MLL], t(1;19)[E2A-PBX1], t(9;22)[BCR-ABL], and hyperdiploid>50) was representative of the entire dataset.
The supervised learning techniques were designed to perform two related tasks. First, gene selection methods were used to discover genes differentially expressed in different groups of patients. The methods applied in our study included Support Vector Machine (SVM) Recursive Feature Elimination (15), adaptive boosting (ADA) (16), Discriminant Analysis (DA)(17), the threshold number of misclassification measure (TNoM) (18, 19), and Fuzzy Inference (20). As differentially expressed genes were discovered, classifiers were designed using the selected genes for the purpose of predicting when patients from the test set belonged to a particular group. Classifiers used included Support Vector Machines (15, 21, 22) (SVMs) and Bayesian networks (23-25).
In addition to the division of the dataset into a training and test set, we were careful to perform the supervised analysis in an unbiased way. In particular, we performed gene selection and designed our classifiers only on the training set. Once the classifiers were finalized we made predictions only once on the test set. Thus we eliminated any possibility of inadvertently biasing our gene lists or classifiers. Furthermore, having selected genes from the supervised learning analysis using training and test sets, the cases were analyzed using the log odds ratio (17) (as well as other methods) to examine statistically whether or not classifier performance was significant. Further details on these methods can be found under computational methods at our website
In addition to the above methods, we also implemented a technique for combining gene lists based on a weighted voting scheme. In this scheme, we consider genes to be candidates and gene lists to be votes. In other words, each method suggests, in order of preference, which genes should be elected. Our method for combining the gene lists ranks the candidate genes according to the geometric mean of the voting order in each list, where each method is allowed the same number of votesas the length n of the shortest list (e.g. n = 21), and all other genes are given a vote of the n+1 (e.g. 21). The lengths of the shortest lists were n = 26 for t(12;21)[TEL-AML1]; n = 18 for t(4;11)[AF4-MLL]; n = 18 for t(1;19)[E2A-PBX1]; n = 21 for t(9;22)[BCR-ABL]; and n = 25 for hyperdiploid>50.
The combined gene lists were used to collate our results as well as to compare our results to the gene lists computed in Yeoh et al.(1). In Yeoh et al. gene lists are generated using five different methods: correlation-based feature selection (CFS); self-organizing maps and discriminate analysis with variance (SOM/DAV); Wilkins’ method; the T-statistic; and the χ2 statistic. Explanations of these methods can be found in the supplementary material for Yeoh et al. at Each of the methods used in Yeoh et al. generate gene lists with 40 to 50 genes per list. The comparison of our combined gene lists with the gene lists produced in Yeoh et al. was performed by computing the intersections of our combined lists with of the lists in Yeoh et al.. These intersections were calculated using Venn diagrams in Genespring ( for each of the karyotypes. From these intersections we computed the percentage overlap between our lists and the lists produced in Yeoh et al. (Table S4.1).
VxInsight Analysis
The VxInsight second clustering algorithm resulted in nine distinct clusters, labeled T1, T2, A, B, C, X, Y, and Z, as shown in Figure S1.1. (We note that the two nearby clusters were grouped to form cluster X.) In the VxInsight clustering, T1 and T2 account for the T-ALL, while A, B, C, X, Y, and Z accounted for the B-ALL cases. Analysis of the B-ALL clusters showed some trends, but no cytogenetic abnormalityprecisely defined any specific cluster. Cases with a t(12;21)[TEL-AML1]or hyperdiploidy, both conferring low risk and good outcomes,tended to cluster together and were seen primarily in clustersC and Z as well as the top component of the X cluster. On the VxInsight terrain map, these 3 clusterregions (C, Z, and X) are co-located, indicating they are related. Similarly, the t(1;19)[E2A-PBX1] cases clustered in Y and hada poorer outcome than those in more distant clusters A and B. Finally, itis of interest to note that ALL cases with t(9;22)[BCR-ABL] did not cluster and instead appeared to be distributed among virtually all B precursorclusters. Results similar to our own were reportedby Fine et al. (26). The distribution of karyotypes within the VxInsight clusters is shown in Figure S1.1 and Figure S1.2. Gene tables differentiating the various clusters are found in Tables S1.1-S1.8.
Tables
Table S1.1. VxInsight gene list separating group A.
Rank / F-score / p-value / Affy Probe Set ID / Gene / Description1 / 106.12 / 0.005 / 37188_at / PCK2 / phosphoenolpyruvate carboxykinase 2 (mitochondrial)
2 / 103.5 / 0.015 / 33342_at / RNUT1 / RNA, U transporter 1
3 / 76.73 / 0.011 / 35701_at / HRAS / v-Ha-ras Harvey rat sarcoma viral oncogene homolog
4 / 71.29 / 0.012 / 36193_at / ARFIP2 / ADP-ribosylation factor interacting protein 2 (arfaptin 2)
5 / 70.59 / 0.019 / 40084_at / TFCP2 / transcription factor CP2
6 / 69.59 / 0.013 / 38895_i_at / NCF4 / neutrophil cytosolic factor 4, 40kDa
7 / 68.14 / 0.012 / 39780_at / PPP3CB / protein phosphatase 3 (formerly 2B), catalytic subunit, beta isoform (calcineurin A beta)
8 / 63.2 / 0.029 / 33430_at / C18orf10 / chromosome 18 open reading frame 10
9 / 63.1 / 0.006 / 35911_r_at / OR1F2P / olfactory receptor, family 1, subfamily F, member 2
10 / 61.34 / 0.044 / 34255_at / DGAT1 / diacylglycerol O-acyltransferase homolog 1 (mouse)
11 / 60.55 / 0.02 / 39009_at / LSM3 / LSM3 homolog, U6 small nuclear RNA associated (S. cerevisiae)
12 / 60.16 / 0.011 / 1382_at / RPA1 / replication protein A1, 70kDa
13 / 60.1 / 0.077 / 35695_at / CHS1 / Chediak-Higashi syndrome 1
14 / 58.74 / 0.04 / 40676_at / ITGB3BP / integrin beta 3 binding protein (beta3-endonexin)
15 / 57.25 / 0.127 / 40472_at / LOC254531 / PlSC domain containing hypothetical protein
16 / 56.08 / 0.011 / 37479_at / CD72 / CD72 antigen
17 / 55.25 / 0.028 / 41198_at / GRN / granulin
18 / 55.16 / 0.026 / 40486_g_at / TRIM44 / tripartite motif-containing 44
19 / 53.67 / 0.018 / 41057_at / THEM2 / thioesterase superfamily member 2
20 / 53.19 / 0.007 / 34359_at / C6orf74 / chromosome 6 open reading frame 74
21 / 53.18 / 0.043 / 37303_at / ADPRTL1 / ADP-ribosyltransferase (NAD+; poly (ADP-ribose) polymerase)-like 1
22 / 52.16 / 0.049 / 36626_at / HSD17B4 / hydroxysteroid (17-beta) dehydrogenase 4
23 / 51.67 / 0.004 / 36276_at / CNTN2 / contactin 2 (axonal)
24 / 50.75 / 0.014 / 41308_at / CTBP1 / C-terminal binding protein 1
25 / 49.81 / 0.015 / 39965_at / RAC3 / ras-related C3 botulinum toxin substrate 3 (rho family, small GTP binding protein Rac3)
26 / 49.48 / 0.041 / 40487_at / TRIM44 / tripartite motif-containing 44
27 / 48.82 / 0.029 / 39043_at / ARPC1B / actin related protein 2/3 complex, subunit 1B, 41kDa
28 / 48.57 / 0.057 / 467_at / OSTF1 / osteoclast stimulating factor 1
29 / 48.46 / 0.01 / 37898_r_at / TFF3 / trefoil factor 3 (intestinal)
30 / 47.27 / 0.017 / 38104_at / DECR1 / 2,4-dienoyl CoA reductase 1, mitochondrial
31 / 47.22 / 0.087 / 36091_at / SCAP2 / src family associated phosphoprotein 2
32 / 46.14 / 0.029 / 399_at / STK25 / serine/threonine kinase 25 (STE20 homolog, yeast)
33 / 45.86 / 0.02 / 34970_r_at / OPLAH / 5-oxoprolinase (ATP-hydrolysing)
34 / 45.74 / 0.029 / 39743_at / FLJ20580 / hypothetical protein FLJ20580
35 / 45.41 / 0.029 / 35843_at / NEK9 / NIMA (never in mitosis gene a)- related kinase 9
36 / 45.21 / 0.039 / 1250_at / PRKDC / protein kinase, DNA-activated, catalytic polypeptide
37 / 45.16 / 0.103 / 33250_at / HLA-DPA3 / major histocompatibility complex, class II, DP alpha 2 (pseudogene)
38 / 44.47 / 0.063 / 32245_at / METTL3 / methyltransferase like 3
39 / 43.99 / 0.042 / 37845_at / HEM1 / hematopoietic protein 1
40 / 43.99 / 0.037 / 1599_at / CDKN3 / cyclin-dependent kinase inhibitor 3 (CDK2-associated dual specificity phosphatase)
41 / 43.83 / 0.03 / 33727_r_at / TNFRSF6B / tumor necrosis factor receptor superfamily, member 6b, decoy
42 / 43.41 / 0.043 / 35820_at / GM2A / GM2 ganglioside activator /// GM2 ganglioside activator
43 / 42.69 / 0.038 / 39896_at / DHX16 / DEAH (Asp-Glu-Ala-His) box polypeptide 16
44 / 42.62 / 0.046 / 40509_at / ETFA / electron-transfer-flavoprotein, alpha polypeptide (glutaric aciduria II)
45 / 42.37 / 0.024 / 35986_at / MYST1 / MYST histone acetyltransferase 1
46 / 41.89 / 0.042 / 34765_at / KIAA0020 / KIAA0020
47 / 41.83 / 0.038 / 40063_at / NDP52 / nuclear domain 10 protein
48 / 41.8 / 0.033 / 40415_at / ACAA1 / acetyl-Coenzyme A acyltransferase 1 (peroxisomal 3-oxoacyl-Coenzyme A thiolase)
49 / 41.78 / 0.021 / 1553_r_at / CYP2A7 / cytochrome P450, family 2, subfamily A, polypeptide 7
50 / 41.63 / 0.132 / 37251_s_at / GPM6B / glycoprotein M6B
Table S1.2. VxInsight gene list separating group B.
Rank / F-score / p-value / Affy Probe Set ID / Gene / Description1 / 108.64 / 0.003 / 32854_at / FBXW1B / F-box and WD-40 domain protein 1B
2 / 99.76 / 0.006 / 39224_at / CENTD1 / centaurin, delta 1
3 / 90.32 / 0.016 / 41625_at / THRAP1 / thyroid hormone receptor associated protein 1
4 / 85.92 / 0.007 / 35289_at / GPR21 / G protein-coupled receptor 21
5 / 82.46 / 0.017 / 35268_at / AXOT / axotrophin
6 / 82.44 / 0.013 / 38082_at / KIAA0650 / KIAA0650 protein
7 / 82.03 / 0.019 / 36827_at / ACBD3 / acyl-Coenzyme A binding domain containing 3
8 / 81.96 / 0.025 / 39759_at / QKI / quaking homolog, KH domain RNA binding (mouse)
9 / 80.54 / 0.01 / 34879_at / DPM1 / dolichyl-phosphate mannosyltransferase polypeptide 1, catalytic subunit
10 / 80.35 / 0.004 / 38659_at / SHOC2 / soc-2 suppressor of clear homolog (C. elegans)
11 / 79.56 / 0.019 / 38462_at / NDUFA5 / NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 5, 13kDa
12 / 79.42 / 0.013 / 38837_at / DJ971N18.2 / hypothetical protein DJ971N18.2
13 / 77.75 / 0.048 / 36144_at / SYT11 / synaptotagmin XI
14 / 77.4 / 0.024 / 37731_at / EPS15 / epidermal growth factor receptor pathway substrate 15
15 / 76.52 / 0.032 / 38685_at / STX12 / syntaxin 12
16 / 75.08 / 0.041 / 38765_at / DICER1 / Dicer1, Dcr-1 homolog (Drosophila)
17 / 72.93 / 0.03 / 38056_at / KIAA0195 / KIAA0195 gene product
18 / 71.6 / 0.034 / 38764_at / DICER1 / Dicer1, Dcr-1 homolog (Drosophila)
19 / 71.4 / 0.021 / 41651_at / KIAA1033 / KIAA1033 protein
20 / 71.22 / 0.032 / 1814_at / TGFBR2 / transforming growth factor, beta receptor II (70/80kDa)
21 / 70.82 / 0.015 / 34370_at / ARCN1 / archain 1
22 / 70.79 / 0.017 / 36474_at / KIAA0776 / KIAA0776
23 / 70.71 / 0.025 / 38041_at / GALNT1 / UDP-N-acetyl-alpha-D-galactosamine:polypeptide N-acetylgalactosaminyltransferase 1 (GalNAc-T1)
24 / 70.35 / 0.021 / 33418_at / RAB3GAP / RAB3 GTPase-ACTIVATING PROTEIN
25 / 70.2 / 0.025 / 33805_at / CAP350 / centrosome-associated protein 350
26 / 70.09 / 0.032 / 34654_at / MTMR1 / myotubularin related protein 1
27 / 70.09 / 0.02 / 35279_at / TAX1BP1 / Tax1 (human T-cell leukemia virus type I) binding protein 1
28 / 69.7 / 0.039 / 34800_at / LRIG1 / leucine-rich repeats and immunoglobulin-like domains 1
29 / 69.62 / 0.03 / 34825_at / TTRAP / TRAF and TNF receptor associated protein
30 / 69.01 / 0.035 / 39964_at / RP2 / retinitis pigmentosa 2 (X-linked recessive)
31 / 68.94 / 0.013 / 39389_at / CD9 / CD9 antigen (p24)
32 / 68.82 / 0.021 / 40610_at / ZFR / zinc finger RNA binding protein
33 / 67.47 / 0.021 / 706_at / Glucocorticoid Receptor, Beta
34 / 67.31 / 0.07 / 33761_s_at / KIAA0493 / KIAA0493 protein
35 / 67 / 0.039 / 33893_r_at / KAB / KARP-1-binding protein
36 / 66.44 / 0.017 / 35793_at / G3BP2 / Ras-GTPase activating protein SH3 domain-binding protein 2
37 / 66.39 / 0.045 / 40839_at / UBL3 / ubiquitin-like 3
38 / 66.33 / 0.036 / 35258_f_at / SFRS2IP / splicing factor, arginine/serine-rich 2, interacting protein
39 / 65.38 / 0.028 / 32857_at / RASSF3 / Ras association (RalGDS/AF-6) domain family 3
40 / 65.28 / 0.034 / 40591_at / CDC27 / cell division cycle 27
41 / 64.78 / 0.042 / 33381_at / NCOA3 / nuclear receptor coactivator 3
42 / 64.13 / 0.037 / 35205_at / COBRA1 / cofactor of BRCA1
43 / 63.96 / 0.052 / 32872_at / MRNA; cDNA DKFZp564I083 (from clone DKFZp564I083)
44 / 63.79 / 0.038 / 35153_at / NBS1 / Nijmegen breakage syndrome 1 (nibrin)
45 / 63.78 / 0.048 / 39695_at / DAF / decay accelerating factor for complement (CD55, Cromer blood group system)
46 / 63.58 / 0.033 / 39691_at / SH3GLB1 / SH3-domain GRB2-like endophilin B1
47 / 63.31 / 0.031 / 34877_at / JAK1 / Janus kinase 1 (a protein tyrosine kinase)
48 / 62.87 / 0.032 / 38818_at / SPTLC1 / serine palmitoyltransferase, long chain base subunit 1
49 / 62.23 / 0.046 / 40865_at / TDG / thymine-DNA glycosylase
50 / 62.16 / 0.063 / 38505_at / EIF2C2 / eukaryotic translation initiation factor 2C, 2
Table S1.3. VxInsight gene list separating group C.
Rank / F-score / p-value / Affy Probe Set ID / Gene / Description1 / 68.93 / 0.002 / 840_at / MYST3 / MYST histone acetyltransferase (monocytic leukemia) 3
2 / 53.59 / 0.016 / 1463_at / PTPN12 / protein tyrosine phosphatase, non-receptor type 12
3 / 53.38 / 0.012 / 35739_at / MTMR3 / myotubularin related protein 3
4 / 52.92 / 0.007 / 39809_at / HBP1 / HMG-box transcription factor 1
5 / 51.49 / 0.014 / 40140_at / RNF103 / ring finger protein 103
6 / 49.62 / 0.015 / 37497_at / HHEX / hematopoietically expressed homeobox
7 / 49.21 / 0.026 / 38148_at / CRY1 / cryptochrome 1 (photolyase-like)
8 / 49.1 / 0.025 / 33861_at / CNOT2 / CCR4-NOT transcription complex, subunit 2
9 / 48.1 / 0.012 / 40570_at / FOXO1A / forkhead box O1A (rhabdomyosarcoma)
10 / 47.34 / 0.044 / 39696_at / PEG10 / paternally expressed 10
11 / 43.91 / 0.031 / 33392_at / DKFZP434J154 / DKFZP434J154 protein
12 / 43.67 / 0.023 / 40128_at / ENTH / enthoprotin
13 / 43.22 / 0.019 / 34892_at / TNFRSF10B / tumor necrosis factor receptor superfamily, member 10b
14 / 43.09 / 0.018 / 1039_s_at / HIF1A / hypoxia-inducible factor 1, alpha subunit (basic helix-loop-helix transcription factor)
15 / 42.84 / 0.023 / 36949_at / Clone IMAGE:3869896, mRNA
16 / 42.09 / 0.029 / 38278_at / ARID5A / AT rich interactive domain 5A (MRF1-like)
17 / 41.65 / 0.07 / 35338_at / FURIN / furin (paired basic amino acid cleaving enzyme)
18 / 41.18 / 0.037 / 34740_at / FOXO3A / forkhead box O3A
19 / 40.94 / 0.054 / 36942_at / KIAA0174 / KIAA0174 gene product
20 / 40.8 / 0.028 / 41577_at / PPP1R16B / protein phosphatase 1, regulatory (inhibitor) subunit 16B
21 / 40.58 / 0.035 / 32025_at / TCF7L2 / transcription factor 7-like 2 (T-cell specific, HMG-box)
22 / 40.53 / 0.035 / 38666_at / PSCD1 / pleckstrin homology, Sec7 and coiled-coil domains 1(cytohesin 1)
23 / 40.41 / 0.024 / 32916_at / PTPRE / protein tyrosine phosphatase, receptor type, E
24 / 40.27 / 0.029 / 1556_at / RBM5 / RNA binding motif protein 5
25 / 40.26 / 0.037 / 36978_at / PSME4 / proteasome (prosome, macropain) activator subunit 4
26 / 39.92 / 0.028 / 35321_at / TLK2 / tousled-like kinase 2
27 / 38.91 / 0.06 / 38980_at / MAP3K7IP2 / mitogen-activated protein kinase kinase kinase 7 interacting protein 2