Dae-Soon Son

Gene expression profiling for the prediction of lymph node metastasis in patients with cervical cancer.

We investigated whether gene expression profiling of primary cervical tumor tissue could be used to predict lymph node (LN) metastasis and compared this with conventional magnetic resonance imaging. We obtained 43 primary cervical cancer samples (16 with LN metastasis and 27 without LN metastasis) for microarray analysis. A prediction model for LN metastasis from the training set was developed by support vector machine methods using a 10‐fold cross‐validation. The ‘LN prediction model’ derived from the signature of 156 distinctive genes (P < 0.01) had a prediction accuracy of 77%. Correlation between mRNA expressions measured by microarray and semiquantitative reverse transcription–polymerase chain reaction was ascertained in four (RBM8A, SDHB, SERPINB13, and γ‐interferon) out of 10 genes. Magnetic resonance imaging showed accuracy (69%) for the prediction of LN metastasis. These results suggest that gene expression profiling allows reliable prediction of LN metastasis in cervical cancer. (Cancer Sci 2008; 99: 31–38)

Synthetic lethal screening reveals FGFR as one of the combinatorial targets to overcome resistance to Met-targeted therapy

Met is a receptor tyrosine kinase that promotes cancer progression. In addition, Met has been implicated in resistance of tumors to various targeted therapies such as epidermal growth factor receptor inhibitors in lung cancers, and has been prioritized as a key molecular target for cancer therapy. However, the underlying mechanism of resistance to Met-targeting drugs is poorly understood. Here, we describe screening of 1310 genes to search for key regulators related to drug resistance to an anti-Met therapeutic antibody (SAIT301) by using a small interfering RNA-based synthetic lethal screening method. We found that knockdown of 69 genes in Met-amplified MKN45 cells sensitized the antitumor activity of SAIT301. Pathway analysis of these 69 genes implicated fibroblast growth factor receptor (FGFR) as a key regulator for antiproliferative effects of Met-targeting drugs. Inhibition of FGFR3 increased target cell apoptosis through the suppression of Bcl-xL expression, followed by reduced cancer cell growth in the presence of Met-targeting drugs. Treatment of cells with the FGFR inhibitors substantially restored the efficacy of SAIT301 in SAIT301-resistant cells and enhanced the efficacy in SAIT301-sensitive cells. In addition to FGFR3, integrin β3 is another potential target for combination treatment with SAIT301. Suppression of integrin β3 decreased AKT phosphorylation in SAIT301-resistant cells and restored SAIT301 responsiveness in HCC1954 cells, which are resistant to SAIT301. Gene expression analysis using CCLE database shows that cancer cells with high levels of FGFR and integrin β3 are resistant to crizotinib treatment, suggesting that FGFR and integrin β3 could be used as predictive markers for Met-targeted therapy and provide a potential therapeutic option to overcome acquired and innate resistance for the Met-targeting drugs.

The minimal amount of starting DNA for Agilent's hybrid capture-based targeted massively parallel sequencing.

Targeted capture massively parallel sequencing is increasingly being used in clinical settings, and as costs continue to decline, use of this technology may become routine in health care. However, a limited amount of tissue has often been a challenge in meeting quality requirements. To offer a practical guideline for the minimum amount of input DNA for targeted sequencing, we optimized and evaluated the performance of targeted sequencing depending on the input DNA amount. First, using various amounts of input DNA, we compared commercially available library construction kits and selected Agilent’s SureSelect-XT and KAPA Biosystems’ Hyper Prep kits as the kits most compatible with targeted deep sequencing using Agilent’s SureSelect custom capture. Then, we optimized the adapter ligation conditions of the Hyper Prep kit to improve library construction efficiency and adapted multiplexed hybrid selection to reduce the cost of sequencing. In this study, we systematically evaluated the performance of the optimized protocol depending on the amount of input DNA, ranging from 6.25 to 200 ng, suggesting the minimal input DNA amounts based on coverage depths required for specific applications.

Predicting Recurrence Using the Clinical Factors of Patients with Non-small Cell Lung Cancer After Curative Resection

We present a recurrence prediction model using multiple clinical parameters in patients surgically treated for non-small cell lung cancer. Among 1,578 lung cancer patients who underwent complete resection, we compared the early-recurrence group with the 3-yr non-recurrence group for evaluating those factors that influence early recurrence within one year after surgery. Adenocarcinoma and squamous cell carcinoma were analyzed independently. We used multiple logistic regression analysis to identify the independent clinical predictors of recurrence and Coxs proportional hazard regression method to develop a clinical prediction model. We randomly divided our patients into the training and test subsets. The pathologic stages, tumor cell type, differentiation of tumor, neoadjuvant therapy and age were significant factors on the multivariable analysis. We constructed the model for the training set with adenocarcinoma (n=236) and squamous cell carcinoma (n=305), and we applied it to the test set with adenocarcinoma (n=110) and squamous cell carcinoma (n=154). It was predictive for the in adenocarcinoma (P<0.001) and the squamous cell carcinoma (P=0.037), respectively. Our results showed that our recurrence prediction model based on the clinical parameters could significantly predict the individual patients who were at high risk or low risk for recurrence.

Analysis of intrapatient heterogeneity uncovers the microevolution of Middle East respiratory syndrome coronavirus

Genome sequence analysis of Middle East respiratory syndrome coronavirus (MERS-CoV) variants from patient specimens has revealed the evolutionary dynamics and mechanisms of pathogenesis of the virus. However, most studies have analyzed the consensus sequences of MERS-CoVs, precluding an investigation of intrapatient heterogeneity. Here, we analyzed non–consensus sequences to characterize intrapatient heterogeneity in cases associated with the 2015 outbreak of MERS in South Korea. Deep-sequencing analysis of MERS-CoV genomes performed on specimens from eight patients revealed significant intrapatient variation; therefore, sequence heterogeneity was further analyzed using targeted deep sequencing. A total of 35 specimens from 24 patients (including a super-spreader) were sequenced to detect and analyze variants displaying intrapatient heterogeneity. Based on the analysis of non–consensus sequences, we demonstrated the intrapatient heterogeneity of MERS-CoVs, with the highest level in the super-spreader specimen. The heterogeneity could be transmitted in a close association with variation in the consensus sequences, suggesting the occurrence of multiple MERS-CoV infections. Analysis of intrapatient heterogeneity revealed a relationship between D510G and I529T mutations in the receptor-binding domain (RBD) of the viral spike glycoprotein. These two mutations have been reported to reduce the affinity of the RBD for human CD26. Notably, although the frequency of both D510G and I529T varied greatly among specimens, the combined frequency of the single mutants was consistently high (87.7% ± 1.9% on average). Concurrently, the frequency of occurrence of the wild type at the two positions was only 6.5% ± 1.7% on average, supporting the hypothesis that selection pressure exerted by the host immune response played a critical role in shaping genetic variants and their interaction in human MERS-CoVs during the outbreak.

Vertical Magnetic Separation of Circulating Tumor Cells for Somatic Genomic-Alteration Analysis in Lung Cancer Patients.

Efficient isolation and genetic analysis of circulating tumor cells (CTCs) from cancer patients’ blood is a critical step for clinical applications using CTCs. Here, we report a novel CTC-isolation method and subsequent genetic analysis. CTCs from the blood were complexed with magnetic beads coated with antibodies against the epithelial cell adhesion molecule (EpCAM) and separated vertically on a density-gradient medium in a modified well-plate. The recovery rate of model CTCs was reasonable and the cell purity was enhanced dramatically when compared to those parameters obtained using a conventional magnetic isolation method. CTCs were recovered from an increased number of patient samples using our magnetic system vs. the FDA-approved CellSearch system (100% vs. 33%, respectively). In 8 of 13 cases, targeted deep sequencing analysis of CTCs revealed private point mutations present in CTCs but not in matched tumor samples and white blood cells (WBCs), which was also validated by droplet digital PCR. Copy-number alterations in CTCs were also observed in the corresponding tumor tissues for some patients. In this report, we showed that CTCs isolated by the EpCAM-based method had complex and diverse genetic features that were similar to those of tumor samples in some, but not all, cases.

Prediction of a time-to-event trait using genome wide SNP data

BackgroundA popular objective of many high-throughput genome projects is to discover various genomic markers associated with traits and develop statistical models to predict traits of future patients based on marker values.ResultsIn this paper, we present a prediction method for time-to-event traits using genome-wide single-nucleotide polymorphisms (SNPs). We also propose a MaxTest associating between a time-to-event trait and a SNP accounting for its possible genetic models. The proposed MaxTest can help screen out nonprognostic SNPs and identify genetic models of prognostic SNPs. The performance of the proposed method is evaluated through simulations.ConclusionsIn conjunction with the MaxTest, the proposed method provides more parsimonious prediction models but includes more prognostic SNPs than some naive prediction methods. The proposed method is demonstrated with real GWAS data.

Circulating tumor DNA shows variable clonal response of breast cancer during neoadjuvant chemotherapy

Circulating tumor DNA (ctDNA) correlates with tumor burden and provides early detection of treatment response and tumor genetic alterations in breast cancer (BC). In this study, we aimed to identify genetic alterations during the process of tumor clonal evolution and examine if ctDNA level well indicated clinical response to neoadjuvant chemotherapy (NAC) and BC recurrence. We performed targeted ultra-deep sequencing of plasma DNAs, matched germline DNAs and tumor DNAs from locally advanced BC patients. Serial plasma DNAs were collected at diagnosis, after the 1st cycle of NAC and after curative surgery. For the target enrichment, we designed RNA baits covering a total of ∼202kb regions of the human genome including a total of 82 cancer-related genes. For ctDNA, 15 serial samples were collected and 87% of plasma SNVs were detected in 13 BC samples that had somatic alterations in tumor tissues. The TP53 mutation was most commonly detected in primary tumor tissues and plasma followed by BRCA1 and BRCA2. At BC diagnosis, the amount of plasma SNVs did not correlate with clinical stage at diagnosis. With respect to the therapeutic effects of NAC, we found two samples in which ctDNA disappeared after the 1st NAC cycle achieved a pathologic complete response (pCR). In addition, the amount of ctDNA correlated with residual cancer volume detected by breast MRI. This targeted ultra-deep sequencing for ctDNA analysis would be useful for monitoring tumor burden and drug resistance. Most of all, we suggest that ctDNA could be the earliest predictor of NAC response.

Nonlinear tumor evolution from dysplastic nodules to hepatocellular carcinoma

Dysplastic nodules are premalignant neoplastic nodules found in explanted livers with cirrhosis. Genetic signatures of premalignant dysplastic nodules (DNs) with concurrent hepatocellular carcinoma (HCC) may provide an insight in the molecular evolution of hepatocellular carcinogenesis. We analyzed four patients with multifocal nodular lesions and cirrhotic background by whole-exome sequencing (WES). The genomic profiles of somatic single nucleotide variations (SNV) and copy number variations (CNV) in DNs were compared to those of HCCs. The number and variant allele frequency of somatic SNVs of DNs and HCCs in each patient was identical along the progression of pathological grade. The somatic SNVs in DNs showed little conservation in HCC. Additionally, CNVs showed no conservation. Phylogenetic analysis based on SNVs and copy number profiles indicated a nonlinear segregation pattern, implying independent development of DNs and HCC in each patient. Thus, somatic mutations in DNs may be developed separately from other malignant nodules in the same liver, suggesting a nonlinear model for hepatocarcinogenesis from DNs to HCC.

Practical approach to determine sample size for building logistic prediction models using high-throughput data

An empirical method of sample size determination for building prediction models was proposed recently. Permutation method which is used in this procedure is a commonly used method to address the problem of overfitting during cross-validation while evaluating the performance of prediction models constructed from microarray data. But major drawback of such methods which include bootstrapping and full permutations is prohibitively high cost of computation required for calculating the sample size. In this paper, we propose that a single representative null distribution can be used instead of a full permutation by using both simulated and real data sets. During simulation, we have used a dataset with zero effect size and confirmed that the empirical type I error approaches to 0.05. Hence this method can be confidently applied to reduce overfitting problem during cross-validation. We have observed that pilot data set generated by random sampling from real data could be successfully used for sample size determination. We present our results using an experiment that was repeated for 300 times while producing results comparable to that of full permutation method. Since we eliminate full permutation, sample size estimation time is not a function of pilot data size. In our experiment we have observed that this process takes around 30min. With the increasing number of clinical studies, developing efficient sample size determination methods for building prediction models is critical. But empirical methods using bootstrap and permutation usually involve high computing costs. In this study, we propose a method that can reduce required computing time drastically by using representative null distribution of permutations. We use data from pilot experiments to apply this method for designing clinical studies efficiently for high throughput data.