Ker-Chau Li

MicroRNA Signature Predicts Survival and Relapse in Lung Cancer

We investigated whether microRNA expression profiles can predict clinical outcome of NSCLC patients. Using real-time RT-PCR, we obtained microRNA expressions in 112 NSCLC patients, which were divided into the training and testing sets. Using Cox regression and risk-score analysis, we identified a five-microRNA signature for the prediction of treatment outcome of NSCLC in the training set. This microRNA signature was validated by the testing set and an independent cohort. Patients with high-risk scores in their microRNA signatures had poor overall and disease-free survivals compared to the low-risk-score patients. This microRNA signature is an independent predictor of the cancer relapse and survival of NSCLC patients.

Pretreatment Epidermal Growth Factor Receptor (EGFR) T790M Mutation Predicts Shorter EGFR Tyrosine Kinase Inhibitor Response Duration in Patients With Non–Small-Cell Lung Cancer

PURPOSE Patients with non-small-cell lung cancer (NSCLC) with epidermal growth factor receptor (EGFR)-activating mutations have excellent response to EGFR tyrosine kinase inhibitors (TKIs), but T790M mutation accounts for most TKI drug resistance. This study used highly sensitive methods to detect T790M before and after TKI therapy and investigated the association of T790M and its mutation frequencies with clinical outcome. PATIENTS AND METHODS Direct sequencing, matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) and next-generation sequencing (NGS) were used to assess T790M in the following two cohorts of patients with NSCLC: TKI-naive patients (n = 107) and TKI-treated patients (n = 85). Results were correlated with TKI treatment response and survival. RESULTS MALDI-TOF MS was highly sensitive in detecting and quantifying the frequency of EGFR-activating mutations and T790M (detection limits, 0.4% to 2.2%). MALDI-TOF MS identified more T790M than direct sequencing in TKI-naive patients with NSCLC (27 of 107 patients, 25.2% v three of 107 patients, 2.8%, respectively; P < .001) and in TKI-treated patients (before TKI: 23 of 73 patients, 31.5% v two of 73 patients, 2.7%, respectively; P < .001; and after TKI: 10 of 12 patients, 83.3% v four of 12 patients, 33.3%, respectively; P = .0143). The EGFR mutations and their frequencies were confirmed by NGS. T790M was an independent predictor of decreased progression-free survival (PFS) in patients with NSCLC who received TKI treatment (P < .05, multivariate Cox regression). CONCLUSION T790M may not be a rare event before or after TKI therapy in patients with NSCLC with EGFR-activating mutations. The pretreatment T790M mutation was associated with shorter PFS with EGFR TKI therapy in patients with NSCLC.

Cancer-associated fibroblasts regulate the plasticity of lung cancer stemness via paracrine signalling

Cancer stem cells (CSCs) are a promising target for treating cancer, yet how CSC plasticity is maintained in vivo is unclear and is difficult to study in vitro. Here we establish a sustainable primary culture of Oct3/4(+)/Nanog(+) lung CSCs fed with CD90(+) cancer-associated fibroblasts (CAFs) to further advance our knowledge of preserving stem cells in the tumour microenvironment. Using transcriptomics we identify the paracrine network by which CAFs enrich CSCs through de-differentiation and reacquisition of stem cell-like properties. Specifically, we find that IGF1R signalling activation in cancer cells in the presence of CAFs expressing IGF-II can induce Nanog expression and promote stemness. Moreover, this paracrine signalling predicts overall and relapse-free survival in stage I non-small cell lung cancer (NSCLC) patients. IGF-II/IGF1R signalling blockade inhibits Nanog expression and attenuates cancer stem cell features. Our data demonstrate that CAFs constitute a supporting niche for cancer stemness, and targeting this paracrine signalling may present a new therapeutic strategy for NSCLC.

Dimension Reduction for Multivariate Response Data

This article concerns the analysis of multivariate response data with multivariate regressors. Methods for reducing the dimensionality of response variables are developed, with the goal of preserving as much regression information as possible. We note parallels between this goal and the goal of sliced inverse regression, which intends to reduce the regressor dimension in a univariate regression while preserving as much regression information as possible. A detailed discussion is given for the case where the response is a curve measured at fixed points. The problem in this setting is to select basis functions for fitting an aggregate of curves. We propose that instead of focusing on goodness of fit, attention should be shifted to the problem of explaining the variation of the curves in terms of the regressor variables. A data-adaptive basis searching method based on dimension reduction theory is proposed. Simulation results and an application to a climatology problem are given.

Inference of transcriptional regulatory network by two-stage constrained space factor analysis

MOTIVATION Microarray gene expression and cross-linking chromatin immunoprecipitation data contain voluminous information that can help the identification of transcriptional regulatory networks at the full genome scale. Such high-throughput data are noisy however. In contrast, from the biomedical literature, we can find many evidenced transcription factor (TF)-target gene binding relationships that have been elucidated at the molecular level. But such sporadically generated knowledge only offers glimpses on limited patches of the network. How to incorporate this valuable knowledge resource to build more reliable network models remains a question. RESULTS We present a modified factor analysis approach. Our algorithm starts with the evidenced TF-gene linkages. It iterates between the network configuration estimation step and the connection strength estimation step, using the high-throughput data, till convergence. We report two comprehensive regulatory networks obtained for Saccharomyces cerevisiae, one under the normal growth condition and the other under the environmental stress condition. SUPPLEMENTARY INFORMATION http://kiefer.stat.ucla.edu/lap2/download/bti656_supplement.pdf.

A comparison of isolated circulating tumor cells and tissue biopsies using whole-genome sequencing in prostate cancer

Previous studies have demonstrated focal but limited molecular similarities between circulating tumor cells (CTCs) and biopsies using isolated genetic assays. We hypothesized that molecular similarity between CTCs and tissue exists at the single cell level when characterized by whole genome sequencing (WGS). By combining the NanoVelcro CTC Chip with laser capture microdissection (LCM), we developed a platform for single-CTC WGS. We performed this procedure on CTCs and tissue samples from a patient with advanced prostate cancer who had serial biopsies over the course of his clinical history. We achieved 30X depth and ≥ 95% coverage. Twenty-nine percent of the somatic single nucleotide variations (SSNVs) identified were founder mutations that were also identified in CTCs. In addition, 86% of the clonal mutations identified in CTCs could be traced back to either the primary or metastatic tumors. In this patient, we identified structural variations (SVs) including an intrachromosomal rearrangement in chr3 and an interchromosomal rearrangement between chr13 and chr15. These rearrangements were shared between tumor tissues and CTCs. At the same time, highly heterogeneous short structural variants were discovered in PTEN, RB1, and BRCA2 in all tumor and CTC samples. Using high-quality WGS on single-CTCs, we identified the shared genomic alterations between CTCs and tumor tissues. This approach yielded insight into the heterogeneity of the mutational landscape of SSNVs and SVs. It may be possible to use this approach to study heterogeneity and characterize the biological evolution of a cancer during the course of its natural history.

A Four-Gene Signature from NCI-60 Cell Line for Survival Prediction in Non–Small Cell Lung Cancer

Purpose: Metastasis is the main cause of mortality in non–small cell lung cancer (NSCLC) patients. Genes that can discriminate the invasion ability of cancer cells may become useful candidates for clinical outcome prediction. We identify invasion-associated genes through computational and laboratorial approach that supported this idea in NSCLC. Experimental Design: We first conducted invasion assay to characterize the invasion abilities of NCI-60 lung cancer cell lines. We then systematically exploited NCI-60 microarray databases to identify invasion-associated genes that showed differential expression between the high and the low invasion cell line groups. Furthermore, using the microarray data of Duke lung cancer cohort (GSE 3141), invasion-associated genes with good survival prediction potentials were obtained. Finally, we validated the findings by conducting quantitative PCR assay on an in-house collected patient group (n = 69) and by using microarray data from two public western cohorts (n = 257 and 186). Results: The invasion-associated four-gene signature (ANKRD49, LPHN1, RABAC1, and EGLN2) had significant prediction in three validation cohorts (P = 0.0184, 0.002, and 0.017, log-rank test). Moreover, we showed that four-gene signature was an independent prognostic factor (hazard ratio, 2.354, 1.480, and 1.670; P = 0.028, 0.014, and 0.033), independent of other clinical covariates, such as age, gender, and stage. Conclusion: The invasion-associated four-gene signature derived from NCI-60 lung cancer cell lines had good survival prediction power for NSCLC patients. (Clin Cancer Res 2009;15(23):7309–15)

Measurement Error Regression with Unknown Link: Dimension Reduction and Data Visualization

Abstract A general nonlinear regression problem is considered with measurement error in the predictors. We assume that the response is related to an unknown linear combination of a multidimensional predictor through an unknown link function. Instead of observing the predictor, we instead observe a surrogate with the property that its expectation is linearly related to the true predictor with constant variance. We identify an important transformation of the surrogate variable. Using this transformed variable, we show that if one proceeds with the usual analysis ignoring measurement error, then both ordinary least squares and sliced inverse regression yield estimates which consistently estimate the true regression parameter, up to a constant of proportionality. We derive the asymptotic distribution of the estimates. A simulation study is conducted applying sliced inverse regression in this context.

Detection of eQTL modules mediated by activity levels of transcription factors

MOTIVATION Studies of gene expression quantitative trait loci (eQTL) in different organisms have shown the existence of eQTL hot spots: each being a small segment of DNA sequence that harbors the eQTL of a large number of genes. Two questions of great interest about eQTL hot spots arise: (1) which gene within the hot spot is responsible for the linkages, i.e. which gene is the quantitative trait gene (QTG)? (2) How does a QTG affect the expression levels of many genes linked to it? Answers to the first question can be offered by available biological evidence or by statistical methods. The second question is harder to address. One simple situation is that the QTG encodes a transcription factor (TF), which regulates the expression of genes linked to it. However, previous results have shown that TFs are not overrepresented in the eQTL hot spots. In this article, we consider the scenario that the propagation of genetic perturbation from a QTG to other linked genes is mediated by the TF activity. We develop a procedure to detect the eQTL modules (eQTL hot spots together with linked genes) that are compatible with this scenario. RESULTS We first detect 27 eQTL modules from a yeast eQTL data, and estimate TF activity profiles using the method of Yu and Li (2005). Then likelihood ratio tests (LRTs) are conducted to find 760 relationships supporting the scenario of TF activity mediation: (DNA polymorphism --> cis-linked gene --> TF activity --> downstream linked gene). They are organized into 4 eQTL modules: an amino acid synthesis module featuring a cis-linked gene LEU2 and the mediating TF Leu3; a pheromone response module featuring a cis-linked gene GPA1 and the mediating TF Ste12; an energy-source control module featuring two cis-linked genes, GSY2 and HAP1, and the mediating TF Hap1; a mitotic exit module featuring four cis-linked genes, AMN1, CSH1, DEM1 and TOS1, and the mediating TF complex Ace2/Swi5. Gene Ontology is utilized to reveal interesting functional groups of the downstream genes in each module. AVAILABILITY Our methods are implemented in an R package: eqtl.TF, which includes source codes and relevant data. It can be freely downloaded at http://www.stat.ucla.edu/~sunwei/software.htm. SUPPLEMENTARY INFORMATION http://www.stat.ucla.edu/~sunwei/yeast_eQTL_TF/supplementary.pdf.

Interactive Tree-Structured Regression via Principal Hessian Directions

Abstract An interactive approach to tree-structured regression is introduced. Unlike other procedures driven by cost optimization, this approach focuses on the exploration of geometric information in the data. The procedure begins with finding a direction along which the regression surface bends the most. This direction is used for splitting the data into two regions. Within each region, another direction is found, and the partition is made in the same manner. The process continues recursively until the entire regressor domain is decomposed into regions wherein the surface no longer bends significantly and linear regression fit becomes appropriate. For implementing the direction search, the method of principal Hessian directions is applied. Several simulation and empirical results are reported. Comparison with three methods—CART, SUPPORT, and MARS—is made. The benefit of using geometric information is highlighted.