Translating Gene Signatures Into a Pathologic Feature: Tumor Necrosis Predicts Disease Relapse in Operable and Stage I Lung Adenocarcinoma.

PURPOSE: The high 5-year disease relapse rate in patients with stage I lung adenocarcinoma indicates the need for additional risk stratification parameters. This study assessed whether gene signatures translate into a pathologic feature for better disease stratification. MATERIALS AND METHODS: The mutual interdependence and risk stratification power of three gene signatures, cell cycle, hypoxia, and mammalian target of rapamycin (mTOR), were investigated in nine cohorts of patients with lung adenocarcinoma and five cohorts of patients with lung squamous cell carcinoma. The translation from gene signatures to a pathologic feature, tumor necrosis, was validated in The Cancer Genome Atlas lung adenocarcinoma cohort. The translation of signature score to pathway activity was further investigated by integrative analyses using The Cancer Genome Atlas and The Cancer Protein Atlas lung adenocarcinoma data sets. RESULTS: The results showed that the three gene signatures were mutually interdependent in lung adenocarcinoma but not in lung squamous cell carcinoma. The signature activities were higher in necrosis-positive tumors than in necrosis-negative tumors. The signature score correlated with the expression level of the representative protein that implicated the activity of each pathway. These signatures stratified patients with operable and stage I lung adenocarcinomas into different risk groups independent of age and stage. Furthermore, the signatures translated to a pathologic feature, tumor necrosis, which predicted shorter overall and relapse-free survival in patients with operable and stage I lung adenocarcinomas. CONCLUSION: This study showed that gene signatures could translate into a pathologic feature, tumor necrosis, with risk stratification ability in patients with operable and stage I lung adenocarcinomas.

Early Assessment of Colorectal Cancer by Quantifying Circulating Tumor Cells in Peripheral Blood: ECT2 in Diagnosis of Colorectal Cancer.

Circulating tumor cells (CTCs) in peripheral blood is an indication of poor prognosis for patients with different cancer types. However, most of the available technologies for detecting CTCs show low sensitivity and specificity. Therefore, we attempted to find an alternative marker for CTCs of colorectal cancer. We have directly extracted RNA from CTCs contained in 1.5 mL peripheral blood from 90 colorectal cancer patients and 151 healthy donors, and screened these samples for candidate marker genes by nested real-time quantitative polymerase chain reaction (PCR). From genes selected from a public database of microarray analyses, we successfully identified epithelial cell transforming sequence 2 oncogene (ECT2) as a gene that exhibits high differential expression ratios (p < 0.01). ECT2 displays good sensitivity and specificity, with an area under the curve (AUC) value of 0.821. This marker gene also has a high detection rate in patients with serum carcinoembryonic antigen (CEA) concentrations below the diagnostic threshold of 5 ng/mL. The expression of ECT2 can therefore serve as an alternative measurement that can compensate for the inadequacy of the current CEA test in the diagnosis and monitoring of colorectal cancer patients.

Signal-on Protein Detection via Dye Translocation between Aptamer and Quantum Dot.

A unique interaction between the cyanine dye and negatively charged quantum dot is used to construct a signal-on biaptameric quantum dot (QD) Förster resonance energy transfer (FRET) beacon for protein detection and distinct aptamer characterization. The beacon comprises a pair of aptamers, one intercalated with the cyanine dye (YOYO-3) and the other conjugated to a negatively charged, carboxyl-QD. When the target protein is present, structural folding and sandwich association of the two aptamers take place. As a consequence, YOYO-3 is displaced from the folded aptamer and transferred to the unblocked QD surface to yield a target concentration-dependent FRET signal. As a proof-of-principle, we demonstrate the detection of thrombin ranging from nanomolar to submicromolar concentrations and confirm the dye translocation using cylindrical illumination confocal spectroscopy (CICS). The proposed beacon provides a simple, rapid, signal-on FRET detection for protein as well as a potential platform for distinct aptamer screening.

Selection of aptamers targeting the sialic acid receptor of hemagglutinin by epitope-specific SELEX.

A new SELEX scheme is proposed for the selection of aptamers targeting a specific epitope of a native protein. Anti-sialic acid receptor (SAR) aptamers that inhibit H1 hemagglutination at a low picomole dose are selected accordingly.

Overcoming EGFR T790M-based Tyrosine Kinase Inhibitor Resistance with an Allele-specific DNAzyme.

Epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKIs) are the main therapeutic agents used to treat non-small-cell lung cancer patients harboring EGFR-activating mutations. However, most of these patients will eventually develop resistance, 50% of which are due to a secondary mutation at T790M in the EGFR. In this paper, we describe the development of an allele-specific DNAzyme, DzT, that can specifically silence EGFR T790M mutant messenger RNA while leaving wild-type EGFR intact. Allele-specific silencing of EGFR T790M expression and downstream signaling by DzT triggered apoptosis in non-small-cell lung cancer cells harboring this mutant. Adding a cholesterol-triethylene glycol group on the 3'-end of DzT (cDzT) improved drug efficacy, increasing inhibitory effect on cell viability from 46 to 79% in T790M/L858R-harboring H1975(TM/LR) non-small-cell lung cancer cells, without loss of allele specificity. Combined treatment with cDzT and BIBW-2992, a second-generation EGFR-tyrosine kinase inhibitor, synergistically inhibited EGFR downstream signaling and suppressed the growth of xenograft tumors derived from H1975(TM/LR) cells. Collectively, these results indicate that the allele-specific DNAzyme, DzT, may provide an alternative treatment for non-small-cell lung cancer that is capable of overcoming EGFR T790M mutant-based tyrosine kinase inhibitor resistance.Molecular Therapy-Nucleic Acids (2014) 3, e150; doi:10.1038/mtna.2014.3; published online 4 March 2014.

Synergistic inhibition of lung cancer cell invasion, tumor growth and angiogenesis using aptamer-siRNA chimeras.

Early metastasis is one of the major causes of mortality among patient with lung cancer. The process of tumor metastasis involves a cascade of events, including epithelial-mesenchymal transition, tumor cell migration and invasion, and angiogenesis. To specifically suppress tumor invasion and angiogenesis, two nucleolin aptamer-siRNA chimeras (aptNCL-SLUGsiR and aptNCL-NRP1siR) were used to block key signaling pathways involved in lung cancer metastasis that are pivotal to metastatic tumor cells but not to normal cells under ordinary physiologic conditions. Through nucleolin-mediated endocytosis, the aptNCL-siRNA chimeras specifically and significantly knocked down the expressions of SLUG and NRP1 in nucleolin-expressing cancer cells. Furthermore, simultaneous suppression of SLUG and NRP1 expressions by the chimeras synergistically retarded cancer cell motility and invasive ability. The synergistic effect was also observed in a xenograft mouse model, wherein the combined treatment using two chimeras suppressed tumor growth, the invasiveness, circulating tumor cell amount, and angiogenesis in tumor tissue without affecting liver and kidney functions. This study demonstrates that combined treatment of aptNCL-SLUGsiR and aptNCL-NRP1siR can synergistically suppress lung cancer cell invasion, tumor growth and angiogenesis by cancer-specific targeting combined with gene-specific silencing.

A new nucleic acid-based agent inhibits cytotoxic T lymphocyte-mediated immune disorders.

BACKGROUND: Stevens-Johnson syndrome (SJS), toxic epidermal necrolysis (TEN), and graft-versus-host disease (GVHD) are distinct immune reactions elicited by drugs or allogeneic antigens; however, they share a pathomechanism with the activation of cytotoxic T lymphocytes (CTLs). CTLs produce cytotoxic proteins, cytokines, chemokines, or immune alarmins, such as granulysin (GNLY), leading to the extensive tissue damage and systemic inflammation seen in patients with SJS/TEN or GVHD. Currently, there is no effective therapeutic agent specific for CTL-mediated immune disorders. OBJECTIVES: By targeting GNLY(+) CTLs, we aimed to develop a nucleic acid-based agent consisting of an anti-CD8 aptamer with GNLY small interfering RNA (siRNA). METHODS: We performed systematic evolution of ligands using exponential enrichment to select and identify effective anti-CD8 aptamers. We developed an aptamer-siRNA chimera using a "sticky bridge" method by conjugating the aptamer with siRNA. We analyzed the inhibitory effects of the aptamer-siRNA chimera on CTL responses in patients with SJS/TEN or GVHD. RESULTS: We identified a novel DNA aptamer (CD8AP17s) targeting CTLs. This aptamer could be specifically internalized into human CTLs. We generated the CD8AP17s aptamer-GNLY siRNA chimera, which showed a greater than 79% inhibitory effect on the production of GNLY by drug/alloantigen-activated T cells. The CD8AP17s aptamer-GNLY siRNA chimera decreased cytotoxicity in in vitro models of both SJS/TEN (elicited by drug-specific antigen) and GVHD (elicited by allogeneic antigens). CONCLUSIONS: Our results identified a new nucleic acid-based agent (CD8 aptamer-GNLY siRNA chimera) that can significantly inhibit CTL-mediated drug hypersensitivity, such as that seen in patients with SJS/TEN, as well as the alloreactivity seen in patients with GVHD. This study provides a novel therapeutic strategy for CTL-mediated immune disorders.

SCUBE3 regulation of early lung cancer angiogenesis and metastatic progression.

Signal peptide-CUB-EGF-like domain-containing protein 3 (SCUBE3) is strongly expressed in extremely invasive lung carcinoma. We showed in our previous study that SCUBE3 triggers the transforming growth factor-ß pathway and subsequently promotes tumor angiogenesis and the epithelial-mesenchymal transition (EMT). However, the role of SCUBE3 in early tumor expansion hasn't been fully demonstrated in vivo. The present study used dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) to temporally assess tumor angiogenesis in SCUBE3-knockdown and control non-small-cell lung carcinoma (NSCLC) cancer cells in the early tumor stage (weeks 1-3). We further evaluated the metastatic potential of the SCUBE3-knockdown and control tumor cells using a circulating tumor cell (CTC) assay. The differences in gene expression profile between these cell lines were determined using microarray analysis. The results show that SCUBE3 knockdown was associated with lower vascular permeability in the tumor and effectively inhibited the metastatic potential of NSCLC, as evidenced by the decreased CTCs in the mice bearing SCUBE3-knockdown tumors. Microarray analysis revealed that several genes involved in angiogenesis and EMT were down-regulated in SCUBE3-knockdown tumors, including matrix metalloproteinases (MMPs) 2, 9, and 14, (MMP-2, MMP-9, and MMP-14, respectively), fibronectin (FN-1), lysyl oxidase (LOX), hairy/enhancer-of-split related with YRPW motif protein 1 (HEY1), early growth response protein 1 (EGR1), and interleukin 8 (IL-8). Together these data suggest that SCUBE3 is a potential target for pharmacological intervention. The findings of the present study also show that differences in vascular permeability precede the CTCs detection, indicating that DCE-MRI may be a sensitive biomarker for assessing tumor invasiveness.

Aptamer microarray as a novel bioassay for protein-protein interaction discovery and analysis.

Aptamer microarray is investigated as a novel bioassay for protein-protein interaction (PPI) discovery and analysis. Assaying a mixture of fluorescence-labeled thrombin and Escherichia coli proteins with an aptamer microarray, we found that thrombin and an unknown protein of E. coli (protein X) formed a complex of PPI, which was captured by an anti-thrombin aptamer probe. The PPI observed on the microarray was double-checked by protein microarrays and confirmed by aptamer-baited co-immunoprecipitation (Co-IP) assays. Characterizing the Co-IP products, we identified protein X as an E. coli Dps protein (DNA-binding protein from starved cells). A SDS-PAGE analysis suggested that Dps should be a substrate for thrombin, a trypsin-like serine protease. A dose-response microarray experiment predicted an apparent dissociation constant of 1.33 µM for the PPI. Moreover, an on-microarray competition assay revealed that the capture of the PPI by the anti-thrombin aptamer probe would be blocked by an E. coli aptamer via complementary base pairing. Thus, a network of protein-protein, protein-DNA, and DNA-DNA interactions and their interaction orders could be addressed in addition to simple PPI discovery.

Unraveling virus identity by detection of depleted probes with capillary electrophoresis.

With the emergence of new viral infections and pandemics, there is a need to develop faster methods to unravel the virus identities in a large number of clinical samples. This report describes a virus identification method featuring high throughput, high resolution, and high sensitivity detection of viruses. Identification of virus is based on liquid hybridization of different lengths of virus-specific probes to their corresponding viruses. The probes bound to target sequences are removed by a biotin-streptavidin pull-down mechanism and the supernatant is analyzed by capillary electrophoresis. The probes depleted from the sample appear as diminished peaks in the electropherograms and the remaining probes serve as calibrators to align peaks in different capillaries. The virus identities are unraveled by a signal processing and peak detection algorithm developed in-house. Nine viruses were used in the study to demonstrate how the system works to unravel the virus identity in single and double virus infections. With properly designed probes, the system is able to distinguish closely related viruses. The system takes advantage of the high resolution feature of capillary electrophoresis to resolve probes that differ by length. The method may facilitate virus identity screen from more candidate viruses with an automated 4-color DNA sequencer.

Involvement of prohibitin upregulation in abrin-triggered apoptosis.

Abrin (ABR), a protein purified from the seeds of Abrus precatorius, induces apoptosis in various types of cancer cells. However, the detailed mechanism remains largely uncharacterized. By using a cDNA microarray platform, we determined that prohibitin (PHB), a tumor suppressor protein, is significantly upregulated in ABR-triggered apoptosis. ABR-induced upregulation of PHB is mediated by the stress-activated protein kinase/c-Jun NH2-terminal kinase (SAPK/JNK) pathway, as demonstrated by chemical inhibitors. In addition, ABR significantly induced the expression of Bax as well as the activation of caspase-3 and poly(ADP-ribose) polymerase (PARP) in Jurkat T cells, whereas the reduction of PHB by specific RNA interference delayed ABR-triggered apoptosis through the proapoptotic genes examined. Moreover, our results also indicated that nuclear translocation of the PHB-p53 complex may play a role in the transcription of Bax. Collectively, our data show that PHB plays a role in ABR-induced apoptosis, which may be helpful for the development of diagnostic or therapeutic agents.

Verifying expressed transcript variants by detecting and assembling stretches of consecutive exons.

We herein describe an integrated system for the high-throughput analysis of splicing events and the identification of transcript variants. The system resolves individual splicing events and elucidates transcript variants via a pipeline that combines aspects such as bioinformatic analysis, high-throughput transcript variant amplification, and high-resolution capillary electrophoresis. For the 14 369 human genes known to have transcript variants, minimal primer sets were designed to amplify all transcript variants and examine all splicing events; these have been archived in the ASprimerDB database, which is newly described herein. A high-throughput thermocycler, dubbed GenTank, was developed to simultaneously perform thousands of PCR amplifications. Following the resolution of the various amplicons by capillary gel electrophoresis, two new computer programs, AmpliconViewer and VariantAssembler, may be used to analyze the splicing events, assemble the consecutive exons embodied by the PCR amplicons, and distinguish expressed versus putative transcript variants. This novel system not only facilitates the validation of putative transcript variants and the detection of novel transcript variants, it also semi-quantitatively measures the transcript variant expression levels of each gene. To demonstrate the system's capability, we used it to resolve transcript variants yielded by single and multiple splicing events, and to decipher the exon connectivity of long transcripts.

Identification and characterization of oligonucleotides that inhibit Toll-like receptor 2-associated immune responses.

Toll-like receptors (TLRs) play important roles in the immune responses against invading microorganisms. Development of TLR antagonists is recognized as a promising direction in suppressing the associated inflammatory reactions of the TLRs. Aptamers are single-stranded RNA or DNA molecules isolated through an in vitro selection process. Using a novel molecular evolution strategy that combines immunoprecipitation (IP) with systematic evolution of ligands by exponential enrichment (SELEX), we developed an IP-SELEX selection method to facilitate the screening of high-affinity aptamers for the Toll-like receptor 2 (TLR2). Also, human TLR2 functional aptamers were identified and characterized using NF-kappaB reporter assays. Among the functional aptamers, the most effective, AP177, with a dissociation constant of 73 pM, was characterized with TLR2-expressing cells challenged with bacterial cells and purified ligands. The aptamer could effectively antagonize TLR2, significantly inhibit NF-kappaB activity, and suppress the secretion of the cytokines by >80%. In addition, the precise region within the functional aptamer that specifically bound TLR2 was resolved using aptamer microarray analysis. The results of functional assays showed that AP177 acted as a TLR2 antagonist and may hold therapeutic potential in the treatment of diseases related to dysregulated TLR2 immune responses.

Enhancement of aptamer microarray sensitivity through spacer optimization and avidity effect.

This work aims for ultrasensitive detection of target proteins in complex biological matrixes based on aptamer microarrays. Two extensively studied aptamers (HTQ and HTDQ) that bind distinct epitopes of thrombin are chosen for the microarray study. Although HTQ and HTDQ have nanomolar to subnanomolar affinities, it is found that either aptamer when applied directly has difficulty in detecting a few nanomoles per liter thrombin in the presence of a 10- or 100-fold (w/w) excess of serum total protein (STP). By investigating dodecyl (12-carbon) and oligodeoxythymidine (oligo(dT)) spacers, we observe both spacers enhance the microarray signal response, but oligo(dT) is strikingly better than dodecyl. Moreover, we discover that a microarray spot coprinted with the two distinct aptamers (HTQ and HTDQ) functions like a bivalent molecular construct and exhibits an avidity effect. With the synergy of oligo(dT) spacers and the avidity effect, detection of picomolar-range thrombin in the presence of either 10% unlabeled serum or a 10,000-fold excess of labeled serum total protein is achieved. It corresponds to a 100-1000-fold sensitivity enhancement as compared to using an individual aptamer without a spacer.

Design and fabrication of spotted long oligonucleotide microarrays for gene expression analysis.

DNA microarray technology has advanced rapidly since the first use of cDNA microarrays almost a decade ago. For gene expression studies on organisms, for which the genomes have been sequenced, cDNA microarrays are being gradually replaced by gene-specific oligonucleotide microarrays. Although, cDNA microarrays give higher signal intensity than oligonucleotide microarrays, they cannot be used for the measurement of gene-specific expression, whereas, oligonucleotide microarrays can. To obtain both a high signal intensity and specificity in gene expression measurements, gene-specific oligonucleotide probes as long as 150-mers, designed using sequence databases and algorithms to identify unique sequences of genes, are used as microarray probes. In order to achieve a high signal intensity, specificity, and accurate measurement of expression, in addition to the length and sequence of the probes, it is necessary to optimize other parameters such as the surface chemistry of the microarray slides, the addition of spacers and linkers to the probes, and the composition of the hybridization solution.

Suppression of growth and cancer-induced angiogenesis of aggressive human breast cancer cells (MDA-MB-231) on the chorioallantoic membrane of developing chicken embryos by E-peptide of pro-IGF-I.

E-peptide of the pro-Insulin-like growth factor-I (pro-IGF-I) is produced from pre-pro-IGF-I by proteolytic cleavage in the post-translational processing. Previous in vitro studies conducted in our laboratory showed that Ea4-peptide of rainbow trout (rt) pro-IGF-I or Eb-peptide of human (h) pro-IGF-I exhibited activities including induction of morphological differentiation, inhibition of anchorage-independent cell growth and suppression of invasion of several well established human cancer cell lines such as MDA-MB-231, HT-29, SK-N-F1, and HepG-2 (Chen et al. [2002] Gen Comp Endocrinol 126:342-351; Kuo and Chen [2002] Exp Cell Res 280:75-89). Seeding of aggressive human breast cancer cells, MDA-MB-231, on the chorioallantoic membrane (CAM) of 5 days old chicken embryos resulted in rapid growth and invasion of the cells and induction of blood vessel formation around the MDA-MB-231 cell mass in the chicken embryos. The invasion of MDA-MB-231 cells in the chicken embryos was further confirmed by immunocytochemistry. The rapid growth and invasion of MDA-MB-231 cells and the induction of blood vessel formation by MDA-MB-231 cells on chicken CAM are inhibited by treatment with a single or multiple doses of rtEa4- or hEb-peptide. Furthermore, a dose-dependent inhibition of angiogenesis by rtEa4- or hEb-peptide was also demonstrated by the chicken CAM assay. Results of microarray analysis of human gene chips (containing 9,500 unique cDNA clones) and confirmation by comparative real-time RT-PCR analysis showed that a group of genes related to cancer cell activities are up- or down-regulated in MDA-MB-231 cells transfected with a rtEa4-peptide gene. Together these results confirm the anti-tumor activity of rtEa4- and hEb-peptides, and further suggest that these peptides could be developed as therapeutics for treating human cancers.

Human kallikrein 8 protease confers a favorable clinical outcome in non-small cell lung cancer by suppressing tumor cell invasiveness.

The human kallikrein 8 (KLK8) gene, a member of the human tissue kallikrein gene family, encodes a serine protease. The KLK8 protein (hK8) is known to be a favorable prognostic marker in ovarian cancer, but the biological basis of this is not understood. We found that overexpressing the KLK8 gene in highly invasive lung cancer cell lines suppresses their invasiveness. This role in invasiveness was further confirmed by the fact that inhibition of endogenous KLK8 expression with a specific short hairpin RNA reduced cancer cell invasiveness. In situ degradation and cell adhesion assays showed that proteins produced from KLK8 splice variants modify the extracellular microenvironment by cleaving fibronectin. DNA microarray experiments and staining of cells for actin filaments revealed that the degradation of fibronectin by hK8 suppresses integrin signaling and retards cancer cell motility by inhibiting actin polymerization. In addition, studies in a mouse model coupled with the detection of circulating tumor cells by quantitative PCR for the human Alu sequence showed that KLK8 suppresses tumor growth and invasion in vivo. Finally, studies of clinical specimens from patients with non-small cell lung cancer showed that the time to postoperative recurrence was longer for early-stage patients (stages I and II) with high KLK8 expression (mean, 49.9 months) than for patients with low KLK8 expression (mean, 22.9 months). Collectively, these findings show that KLK8 expression confers a favorable clinical outcome in non-small cell lung cancer by suppressing tumor cell invasiveness.

Cartridge-based high-throughput purification of oligonucleotides for reliable oligonucleotide arrays.

A novel, cartridge-based procedure for the efficient and irreversible detritylation of oligonucleotides is reported. This method, combined with a process for the elimination of depurinated fragments produces, in a highly parallel fashion, oligonucleotides with better purity than those traditionally obtained using reversed-phase high-performance liquid chromotography purification. Our combined detritylation and purification methodology compares favorably with commercial cartridge-based purification systems. The benefits of working with pure oligonucleotides, with regard to higher signal and better signal linearity, are shown in array-based hybridization experiments.

Design of microarray probes for virus identification and detection of emerging viruses at the genus level.

BACKGROUND: Most virus detection methods are geared towards the detection of specific single viruses or just a few known targets, and lack the capability to uncover the novel viruses that cause emerging viral infections. To address this issue, we developed a computational method that identifies the conserved viral sequences at the genus level for all viral genomes available in GenBank, and established a virus probe library. The virus probes are used not only to identify known viruses but also for discerning the genera of emerging or uncharacterized ones. RESULTS: Using the microarray approach, the identity of the virus in a test sample is determined by the signals of both genus and species-specific probes. The genera of emerging and uncharacterized viruses are determined based on hybridization of the viral sequences to the conserved probes for the existing viral genera. A detection and classification procedure to determine the identity of a virus directly from detection signals results in the rapid identification of the virus. CONCLUSION: We have demonstrated the validity and feasibility of the above strategy with a small number of viral samples. The probe design algorithm can be applied to any publicly available viral sequence database. The strategy of using separate genus and species probe sets enables the use of a straightforward virus identity calculation directly based on the hybridization signals. Our virus identification strategy has great potential in the diagnosis of viral infections. The virus genus and specific probe database and the associated summary tables are available at http://genestamp.sinica.edu.tw/virus/index.htm.

A multivariate prediction model for microarray cross-hybridization.

BACKGROUND: Expression microarray analysis is one of the most popular molecular diagnostic techniques in the post-genomic era. However, this technique faces the fundamental problem of potential cross-hybridization. This is a pervasive problem for both oligonucleotide and cDNA microarrays; it is considered particularly problematic for the latter. No comprehensive multivariate predictive modeling has been performed to understand how multiple variables contribute to (cross-) hybridization. RESULTS: We propose a systematic search strategy using multiple multivariate models [multiple linear regressions, regression trees, and artificial neural network analyses (ANNs)] to select an effective set of predictors for hybridization. We validate this approach on a set of DNA microarrays with cytochrome p450 family genes. The performance of our multiple multivariate models is compared with that of a recently proposed third-order polynomial regression method that uses percent identity as the sole predictor. All multivariate models agree that the 'most contiguous base pairs between probe and target sequences,' rather than percent identity, is the best univariate predictor. The predictive power is improved by inclusion of additional nonlinear effects, in particular target GC content, when regression trees or ANNs are used. CONCLUSION: A systematic multivariate approach is provided to assess the importance of multiple sequence features for hybridization and of relationships among these features. This approach can easily be applied to larger datasets. This will allow future developments of generalized hybridization models that will be able to correct for false-positive cross-hybridization signals in expression experiments.