Cell Line Data Research Articles

Multi-task deep latent spaces for cancer survival and drug sensitivity prediction.

Cancer is a very heterogeneous disease that can be difficult to treat without addressing the specific mechanisms driving tumour progression in a given patient. High-throughput screening and sequencing data from cancer cell-lines has driven many developments in drug development, however, there are important aspects crucial to precision medicine that are often overlooked, namely the inherent differences between tumours in patients and the cell-lines used to model them in vitro. Recent developments in transfer learning methods for patient and cell-line data have shown progress in translating results from cell-lines to individual patients in silico. However, transfer learning can be forceful and there is a risk that clinically relevant patterns in the omics profiles of patients are lost in the process. We present MODAE, a novel deep learning algorithm to integrate omics profiles from cell-lines and patients for the purposes of exploring precision medicine opportunities. MODAE implements patient survival prediction as an additional task in a drug-sensitivity transfer learning schema and aims to balance autoencoding, domain adaptation, drug-sensitivity prediction, and survival prediction objectives in order to better preserve the heterogeneity between patients that is relevant to survival. While burdened with these additional tasks, MODAE performed on par with baseline survival models, but struggled in the drug-sensitivity prediction task. Nevertheless, these preliminary results were promising and show that MODAE provides a novel AI-based method for prioritizing drug treatments for high-risk patients. https://github.com/UEFBiomedicalInformaticsLab/MODAE.

MiR-6721-5p as a natural regulator of Meta-VCL is upregulated in the serum of patients with coronary artery disease

BackgroundCoronary artery disease (CAD), the leading cause of mortality globally, arises from atherosclerotic blockage of the coronary arteries. Meta-vinculin (meta-VCL), a large spliced isoform of VCL, co-localizes in muscular adhesive structures and plays significant roles in cardiac physiology and pathophysiology. This study aimed to identify microRNAs (miRNAs) regulating meta-VCL expression and investigate the expression alterations of the miRNAs of interest and meta-VCL as potential biomarkers in the serum of CAD patients. MethodsBioinformatics tools were employed to select miRNAs targeting meta-VCL. Cell-based ectopic expression analysis and a dual-luciferase assay were used to examine the interactions between miRNAs and meta-VCL. An ELISA assessed the concentrations of interleukin-6 (IL-6), IL-10, and tumor necrosis factor-α (TNF-α). MiRNA and meta-VCL expression patterns and biomarker suitability were evaluated in serum samples from CAD and non-CAD individuals using real-time PCR. A cardiac cell-line data set and CAD blood exosome samples were analyzed using bioinformatics and ROC curve analyses, respectively. ResultsmiR-6721-5p directly interacted with the putative target sites at the 3′-UTR of meta-VCL and regulated its expression. IL-10 and TNF-α concentrations, which may act as anti-inflammatory factors, decreased following miR-6721-5p upregulation and meta-VCL downregulation. Bioinformatics and experimental expression analyses confirmed downregulated meta-VCL expression and upregulated miR-6721-5p expression in CAD samples. ROC curve analysis yielded an AUC score of 0.705 (P = 0.018), indicating the potential suitability of miR-6721-5p as a biomarker for CAD. ConclusionsmiR-6721-5p plays a regulatory role in meta-VCL expression and may contribute to CAD development by reducing anti-inflammatory factors. These findings suggest that miR-6721-5p could serve as a novel biomarker in the pathogenesis of CAD.

MS Ana: Improving Sensitivity in Peptide Identification with Spectral Library Search

Spectral library search can enable more sensitive peptide identification in tandem mass spectrometry experiments. However, its drawbacks are the limited availability of high-quality libraries and the added difficulty of creating decoy spectra for result validation. We describe MS Ana, a new spectral library search engine that enables high sensitivity peptide identification using either curated or predicted spectral libraries as well as robust false discovery control through its own decoy library generation algorithm. MS Ana identifies on average 36% more spectrum matches and 4% more proteins than database search in a benchmark test on single-shot human cell-line data. Further, we demonstrate the quality of the result validation with tests on synthetic peptide pools and show the importance of library selection through a comparison of library search performance with different configurations of publicly available human spectral libraries.

Deep transfer learning of cancer drug responses by integrating bulk and single-cell RNA-seq data

Drug screening data from massive bulk gene expression databases can be analyzed to determine the optimal clinical application of cancer drugs. The growing amount of single-cell RNA sequencing (scRNA-seq) data also provides insights into improving therapeutic effectiveness by helping to study the heterogeneity of drug responses for cancer cell subpopulations. Developing computational approaches to predict and interpret cancer drug response in single-cell data collected from clinical samples can be very useful. We propose scDEAL, a deep transfer learning framework for cancer drug response prediction at the single-cell level by integrating large-scale bulk cell-line data. The highlight in scDEAL involves harmonizing drug-related bulk RNA-seq data with scRNA-seq data and transferring the model trained on bulk RNA-seq data to predict drug responses in scRNA-seq. Another feature of scDEAL is the integrated gradient feature interpretation to infer the signature genes of drug resistance mechanisms. We benchmark scDEAL on six scRNA-seq datasets and demonstrate its model interpretability via three case studies focusing on drug response label prediction, gene signature identification, and pseudotime analysis. We believe that scDEAL could help study cell reprogramming, drug selection, and repurposing for improving therapeutic efficacy.

Predicting drug sensitivity of cancer cells based on DNA methylation levels.

Cancer cell lines, which are cell cultures derived from tumor samples, represent one of the least expensive and most studied preclinical models for drug development. Accurately predicting drug responses for a given cell line based on molecular features may help to optimize drug-development pipelines and explain mechanisms behind treatment responses. In this study, we focus on DNA methylation profiles as one type of molecular feature that is known to drive tumorigenesis and modulate treatment responses. Using genome-wide, DNA methylation profiles from 987 cell lines in the Genomics of Drug Sensitivity in Cancer database, we used machine-learning algorithms to evaluate the potential to predict cytotoxic responses for eight anti-cancer drugs. We compared the performance of five classification algorithms and four regression algorithms representing diverse methodologies, including tree-, probability-, kernel-, ensemble-, and distance-based approaches. We artificially subsampled the data to varying degrees, aiming to understand whether training based on relatively extreme outcomes would yield improved performance. When using classification or regression algorithms to predict discrete or continuous responses, respectively, we consistently observed excellent predictive performance when the training and test sets consisted of cell-line data. Classification algorithms performed best when we trained the models using cell lines with relatively extreme drug-response values, attaining area-under-the-receiver-operating-characteristic-curve values as high as 0.97. The regression algorithms performed best when we trained the models using the full range of drug-response values, although this depended on the performance metrics we used. Finally, we used patient data from The Cancer Genome Atlas to evaluate the feasibility of classifying clinical responses for human tumors based on models derived from cell lines. Generally, the algorithms were unable to identify patterns that predicted patient responses reliably; however, predictions by the Random Forests algorithm were significantly correlated with Temozolomide responses for low-grade gliomas.

NetControl4BioMed: a web-based platform for controllability analysis of protein-protein interaction networks.

MotivationThere is an increasing amount of data coming from genome-wide studies identifying disease-specific survivability-essential proteins and host factors critical to a cell becoming infected. Targeting such proteins has a strong potential for targeted, precision therapies. Typically however, too few of them are drug targetable. An alternative approach is to influence them through drug targetable proteins upstream of them. Structural target network controllability is a suitable solution to this problem. It aims to discover suitable source nodes (e.g. drug targetable proteins) in a directed interaction network that can control (through a suitable set of input functions) a desired set of targets.ResultsWe introduce NetControl4BioMed, a free open-source web-based application that allows users to generate or upload directed protein–protein interaction networks and to perform target structural network controllability analyses on them. The analyses can be customized to focus the search on drug targetable source nodes, thus providing drug therapeutic suggestions. The application integrates protein data from HGNC, Ensemble, UniProt, NCBI and InnateDB, directed interaction data from InnateDB, Omnipath and SIGNOR, cell-line data from COLT and DepMap, and drug–target data from DrugBank.Availabilityand implementationThe application and data are available online at https://netcontrol.combio.org/. The source code is available at https://github.com/Vilksar/NetControl4BioMed under an MIT license.

SCNrank: spectral clustering for network-based ranking to reveal potential drug targets and its application in pancreatic ductal adenocarcinoma

BackgroundPancreatic ductal adenocarcinoma (PDAC) is the most common pancreatic malignancy. Due to its wide heterogeneity, PDAC acts aggressively and responds poorly to most chemotherapies, causing an urgent need for the development of new therapeutic strategies. Cell lines have been used as the foundation for drug development and disease modeling. CRISPR-Cas9 plays a key role in every step-in drug discovery: from target identification and validation to preclinical cancer cell testing. Using cell-line models and CRISPR-Cas9 technology together make drug target prediction feasible. However, there is still a large gap between predicted results and actionable targets in real tumors. Biological network models provide great modus to mimic genetic interactions in real biological systems, which can benefit gene perturbation studies and potential target identification for treating PDAC. Nevertheless, building a network model that takes cell-line data and CRISPR-Cas9 data as input to accurately predict potential targets that will respond well on real tissue remains unsolved.MethodsWe developed a novel algorithm ‘Spectral Clustering for Network-based target Ranking’ (SCNrank) that systematically integrates three types of data: expression profiles from tumor tissue, normal tissue and cell-line PDAC; protein-protein interaction network (PPI); and CRISPR-Cas9 data to prioritize potential drug targets for PDAC. The whole algorithm can be classified into three steps: 1. using STRING PPI network skeleton, SCNrank constructs tissue-specific networks with PDAC tumor and normal pancreas tissues from expression profiles; 2. With the same network skeleton, SCNrank constructs cell-line-specific networks using the cell-line PDAC expression profiles and CRISPR-Cas 9 data from pancreatic cancer cell-lines; 3. SCNrank applies a novel spectral clustering approach to reduce data dimension and generate gene clusters that carry common features from both networks. Finally, SCNrank applies a scoring scheme called ‘Target Influence score’ (TI), which estimates a given target’s influence towards the cluster it belongs to, for scoring and ranking each drug target.ResultsWe applied SCNrank to analyze 263 expression profiles, CRPSPR-Cas9 data from 22 different pancreatic cancer cell-lines and the STRING protein-protein interaction (PPI) network. With SCNrank, we successfully constructed an integrated tissue PDAC network and an integrated cell-line PDAC network, both of which contain 4414 selected genes that are overexpressed in tumor tissue samples. After clustering, 4414 genes are distributed into 198 clusters, which include 367 targets of FDA approved drugs. These drug targets are all scored and ranked by their TI scores, which we defined to measure their influence towards the network. We validated top-ranked targets in three aspects: Firstly, mapping them onto the existing clinical drug targets of PDAC to measure the concordance. Secondly, we performed enrichment analysis to these drug targets and the clusters there are within, to reveal functional associations between clusters and PDAC; Thirdly, we performed survival analysis for the top-ranked targets to connect targets with clinical outcomes. Survival analysis reveals that overexpression of three top-ranked genes, PGK1, HMMR and POLE2, significantly increases the risk of death in PDAC patients.ConclusionSCNrank is an unbiased algorithm that systematically integrates multiple types of omics data to do potential drug target selection and ranking. SCNrank shows great capability in predicting drug targets for PDAC. Pancreatic cancer-associated gene candidates predicted by our SCNrank approach have the potential to guide genetics-based anti-pancreatic drug discovery.

Abstract 1843: Computational biology model (CBM) predicts nutrient dependency of cancer patients based on Tumor Genomics: Implication of precision diet in cancer therapy

Abstract Background: Tumor metabolism is the hallmark of cancer cells. Cancer cells utilize different nutrient sources to drive the metabolic pathways to sustain tumor growth. Glucose (Glu) and Glutamine (Gln) are the primary nutrient sources on which cancer cells thrive. Developing precision diet based on patient’s molecular characteristics can help treat the cancer with dietary modulations along with traditional approaches. Methods: Computational Biology Model (CBM) captures the signaling and metabolic pathways to predict cancer phenotypes and biomarkers. Genomic aberrations (Mutations / Copy Number Variations (CNV)) from a patient’s tumor are input into the CBM to create the patient disease model. CBM is used for diet prediction based on the molecular characteristics of the patient’s disease. CBM is validated using a data set of 54 cancer cell-lines across indications, by assessing nutrient dependency for Glu and Gln. Simulation based prediction of Glu and Gln dependency is based on the expression of transporters and rate limiting enzymes of cellular glucose (SLC2A1-4, HK2) and glutamine (SLC1A5, GLS) uptake. The enzymes regulating the de-novo synthesis of glucose (PCK1/2, FBP1) and glutamine (GLUL) are negative determinants. In the CBM, an index is defined to measure Glu and Gln dependency. Glu Dependency Index = (SLC2A1 + SLC2A2 + SLC2A3 + SLC2A4 + HK2) / (PCK1 + PCK2) Gln Dependency Index = (SLC1A5 + GLS) / (GLUL) Threshold values for Glu and Gln dependency was determined based on the simulation correlation with the cell-line data. The validated CBM was then used for diet prediction for patient genomics. Results: Validation of genomics-based diet prediction by CBM using 54 cancer cell lines had an accuracy, positive predictive value and negative predictive value of 85%, 97% and 44% for Glu dependency and 82%, 94% and 50% for Gln dependency respectively. Using this validated CBM, we present predictions of patient nutrient source dependency based on their tumor genomics: Case Study 1: A Glioblastoma multiforme (GBM) patient case with PTEN EGFRVIII and ALK mutation and high copy number of HIF1A and MIR-145. CBM predicted the patient to be Glu dependent and Gln independent. Case Study 2: A GBM patient case with CTNNB1 mutation and low copy number of PTEN, RB1 and NF1. This patient was predicted to be both Glu and Gln dependent. Case Study 3: A Triple Negative Breast Cancer (TNBC) patient carrying mutations for MYC, BRD4, EP300 and CREBBP. CBM predicted this patient to be Gln dependent and Glu independent. The rationales for the nutrient source dependency predictions based on disease pathway characteristics were determined. Conclusion: Using CBM we could successfully use patient genomic data to predict nutrient dependency of patient’s tumor. This analysis enables creating options for precision diet for a patient to be used as an adjuvant alongside traditional approaches. Citation Format: Shireen Vali, Taher Abbasi, Subrat Mohapatra, Vishwas Joseph, Ashish Kumar Agrawal, Anuj Tyagi, Neelesh Lunkad, Ashokraja Balla. Computational biology model (CBM) predicts nutrient dependency of cancer patients based on Tumor Genomics: Implication of precision diet in cancer therapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1843.

Abstract 1554: Development of a deconvolution algorithm for tissue-based gene expression data

Abstract Tissue data provide substantially more information than cell-line data, and offer new opportunities to study cancer biology and evolution in its actual microenvironment, when multiple tissue samples of the same cancer type are analyzed together. However, it is very challenging to do information discovery from tissue data because of their compositional complexity - each dataset represents a mixture of gene-expression data from multiple cell types. Hence, meaningful tissue-data analyses require to first sort out the detailed contributions to the observed tissue-level data by different cell types. However, the computational challenge in solving the tissue data deconvolution problem stems from the reality: each cell type has a very large number of complex relations among its expressed genes and pathways, which are preserved under different conditions. To make deconvolution results meaningful, the co-expressions among functionally closely related genes, must be captured and enforced in a deconvolution problem formulation. We have a fundamentally novel formulation of the tissue-data deconvolution problem, which is pathway instead of gene based. It preserves co-expressions of genes of the same pathways through capturing and using an expression signature among such genes, and allows differential expressions of pathways by giving them varying weights in different tissues. In addition, this pathway-based model substantially reduces the number of free variables compared to gene-based models, making our problem formulation efficiently solvable. The unique ideas of our approach are: (i) estimate the proportion of each cell type within a mixture based on expression patterns of cell-type specific genes; (ii) identify co-expressed gene clusters among genes encoding each pathway (as defined by REACTOME) in each cell type based on cell line data; (iii) derive a condition-invariant expression signature for each of the ~2,000 REACTOME pathways in each cell type; (iv) demonstrate that each set of cell line gene-expression data can be uniquely represented as a weighted sum of such signatures; and (v) formulate the deconvolution problem as to estimate the weight of each pathway in each cell type that minimizes the Frobenius norm of deviance matrix of estimated and observed gene expression matrices. We demonstrated its effectiveness on simulated data, i.e., in silico mixtures of gene-expression data from different cell types with varying proportions of each cell type. We anticipate that the successful development and deployment of the planned deconvolution method will enable and inspire a wide range of new ways to study tissue-based expression data and uncover the very rich information hidden in cancer tissues about the fundamental biology of cancer. Citation Format: Sha Cao, Chi Zhang, Ying Xu. Development of a deconvolution algorithm for tissue-based gene expression data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 1554. doi:10.1158/1538-7445.AM2017-1554

Animal-Free Chemical Safety Assessment.

The exponential growth of the Internet of Things and the global popularity and remarkable decline in cost of the mobile phone is driving the digital transformation of medical practice. The rapidly maturing digital, non-medical world of mobile (wireless) devices, cloud computing and social networking is coalescing with the emerging digital medical world of omics data, biosensors and advanced imaging which offers the increasingly realistic prospect of personalized medicine. Described as a potential “seismic” shift from the current “healthcare” model to a “wellness” paradigm that is predictive, preventative, personalized and participatory, this change is based on the development of increasingly sophisticated biosensors which can track and measure key biochemical variables in people. Additional key drivers in this shift are metabolomic and proteomic signatures, which are increasingly being reported as pre-symptomatic, diagnostic and prognostic of toxicity and disease. These advancements also have profound implications for toxicological evaluation and safety assessment of pharmaceuticals and environmental chemicals. An approach based primarily on human in vivo and high-throughput in vitro human cell-line data is a distinct possibility. This would transform current chemical safety assessment practice which operates in a human “data poor” to a human “data rich” environment. This could also lead to a seismic shift from the current animal-based to an animal-free chemical safety assessment paradigm.

Abstract P4-04-02: RNA-seq reveals functional lncRNAs associated with estrogen-receptor status in breast cancer

Abstract Background: Long non-coding RNAs (lncRNAs) are pervasively transcribed in the genome yet their role in human disease is not well understood. LncRNAs can have regulatory effects on coding mRNAs through a number of mechanisms, including repressing their sense-strand protein-coding partners. There is also emerging evidence that estrogen signaling affects the expression of a wide variety of non-coding functional RNA transcripts in addition to protein-coding transcripts. We performed an RNA-seq study to characterize changes in lncRNA expression and evaluate their association to estrogen receptor (ER) status and estrogen signaling. Methods: We sequenced transcriptomes of core biopsy RNA from 45 breast tumors obtained from neoadjuvant clinical trials BrUOG 211A/211B. RNA was derived from biopsy samples obtained before exposure to run-in monotherapy with either nab-paclitaxel, bevacizumab or trastuzumab. Paired-end sequencing was performed using amplified total RNA with 74bp read length, yielding genome-wide transcriptomic data. Transcriptomic abundance and differential expression were estimated assuming Poisson-distributed read-counts. Paired-end sequence data was aligned to a lncRNA database containing 14,572 unique lncRNAs. Changes in relative abundance of lncRNA transcripts were tested for association with estrogen receptor status using the Wilcoxon rank-sum test. Expression levels of the ER-associated lncRNAs were investigated in RNA-seq data from ER-positive MCF7 cells in response to treatment with E2 and tamoxifen (3hr, 12 hr and a non-treated control). ER-responsive binding sites on or near the ER-associated lncRNAs were investigated in a ChIP-seq study in MCF7 cells following estrogen treatment. MCF7 cell-line RNA-seq and ChIP-seq data were obtained from publicly available Short Read Archive (ERX030990 and SRX113365) at NCBI. Results: On average, in each patient 5000 lncRNAs were detectable. Expression of 22 lncRNAs were associated ER status (p<0.05) in breast tumors. 17 of the 22 ER-associated lncRNAs were detectable in ER-positive MCF7 breast cancer cells and were all responsive to E2 (estradiol) treatment. Interestingly, all 17 lncRNAs showed significant downregulation (> = 2 fold) upon tamoxifen treatment. To further evaluate the regulatory relationship between ER and the 22 lncRNAs, we identified the binding sites of ER in estrogen-treated MCF7 cells using ChIP-seq. We found that >50% of the ER-associated lncRNAs had ER binding site either overlapping or neighboring the lncRNA. Discussion: We have shown that lncRNA expression levels are associated with ER status in breast tumors. We have further established that they are estrogen-responsive with a majority being direct targets of ER using cell-line data. Further functional validation studies are ongoing. We are also exploring lncRNA-mRNA expression for coding partners of lncRNAs to identify coding/noncoding gene regulatory networks important in estrogen response. Understanding the regulatory effects of lncRNA expression opens up new opportunities for stratification and management of breast cancers. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P4-04-02.

A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data

Motivation: The Illumina Infinium 450 k DNA Methylation Beadchip is a prime candidate technology for Epigenome-Wide Association Studies (EWAS). However, a difficulty associated with these beadarrays is that probes come in two different designs, characterized by widely different DNA methylation distributions and dynamic range, which may bias downstream analyses. A key statistical issue is therefore how best to adjust for the two different probe designs.Results: Here we propose a novel model-based intra-array normalization strategy for 450 k data, called BMIQ (Beta MIxture Quantile dilation), to adjust the beta-values of type2 design probes into a statistical distribution characteristic of type1 probes. The strategy involves application of a three-state beta-mixture model to assign probes to methylation states, subsequent transformation of probabilities into quantiles and finally a methylation-dependent dilation transformation to preserve the monotonicity and continuity of the data. We validate our method on cell-line data, fresh frozen and paraffin-embedded tumour tissue samples and demonstrate that BMIQ compares favourably with two competing methods. Specifically, we show that BMIQ improves the robustness of the normalization procedure, reduces the technical variation and bias of type2 probe values and successfully eliminates the type1 enrichment bias caused by the lower dynamic range of type2 probes. BMIQ will be useful as a preprocessing step for any study using the Illumina Infinium 450 k platform.Availability: BMIQ is freely available from http://code.google.com/p/bmiq/.Contact: a.teschendorff@ucl.ac.ukSupplementary information: Supplementary data are available at Bioinformatics online

Analysis of bypass signaling in EGFR pathway and profiling of bypass genes for predicting response to anticancer EGFR tyrosine kinase inhibitors

Some drugs, such as anticancer EGFR tyrosine kinase inhibitors, elicit markedly different clinical response rates due to differences in drug bypass signaling as well as genetic variations of drug target and downstream drug-resistant genes. The profiles of these bypass signaling are expected to be useful for improved drug response prediction, which have not been systematically explored previously. In this work, we searched and analyzed 16 literature-reported EGFR tyrosine kinase inhibitor bypass signaling routes in the EGFR pathway, which include 5 compensatory routes of EGFR transactivation by another receptor, and 11 alternative routes activated by another receptor. These 16 routes are reportedly regulated by 11 bypass genes. Their expression profiles together with the mutational, amplification and expression profiles of EGFR and 4 downstream drug-resistant genes, were used as new sets of biomarkers for identifying 53 NSCLC cell-lines sensitive or resistant to EGFR tyrosine kinase inhibitors gefitinib, erlotinib and lapatinib. The collective profiles of all 16 genes distinguish sensitive and resistant cell-lines are better than those of individual genes and the combined EGFR and downstream drug resistant genes, and their derived cell-line response rates are consistent with the reported clinical response rates of the three drugs. The usefulness of cell-line data for drug response studies was further analyzed by comparing the expression profiles of EGFR and bypass genes in NSCLC cell-lines and patient samples, and by using a machine learning feature selection method for selecting drug response biomarkers. Our study suggested that the profiles of drug bypass signaling are highly useful for improved drug response prediction.

Abstract 147: Identification of miRNA associated with chemoresponse using combined clinical, population, and cellular data in lung cancer

Abstract Introduction: MiRNA are potentially important moderators of chemoresponse in human cancer due to their widespread regulatory function. Identification of a set of clinically validated and biologically relevant miRNA could serve as potential markers for prognostication, chemoresponse prediction, and novel therapeutic targets. The objective of this study was to utilize population, clinical, and cell-line data as an integrative approach to identifying miRNA showing a consistent association with chemoresponse. Materials and Methods: Patients diagnosed with squamous cell carcinomas of the lung between 2004 and 2007 were retrieved from the linked SEER-Medicare database with data on tumor TNM markers, grade, survival, and treatment information (N=6,215). Medicare HCPCS and Revenue codes were used to identify patients receiving cisplatin (N=142), carboplatin (N=1,125), etoposide, (N=37), and paclitaxel (N=779). The AJCC TNM markers and grade were used to partition patients into disjoint groups indicative of tumor progression, with the average survival being calculated for each group. A second set of clinical data containing miRNA expression (N=57, GSE 16025) was partitioned in the same manner, with the average miRNA expression on 328 probes being calculated for each group. Linear regression was used to estimate the association between miRNA expression in the clinical cohort and survival in the population sample given similar tumor characteristics. This analysis was also performed on two subsets of patients receiving surgery and chemotherapy both with and without preoperative radiation. A second analysis utilizing the NCI-60 anti-cancer screen estimated the association between miRNA expression and the amount of each chemotherapeutic agent needed for 50% growth-inhibition in an experimental setting. MiRNA which showed a significant association in at least one of the analyses and concordance in all others were considered for further analysis using IPA software. Results: Multiple miRNA showed consistent associations with chemoresponse across analytic approaches as well as having documented interactions with molecular species known to play a role in human cancers. Notably, over-expression of miR-25, 27a, 141, and 181a showed a strong association with chemosensitivity in three of the four agents considered and formed an interaction network including transcriptional regulation, HDAC, and EGF activity. Over-expression of miR-133b was associated with chemoresistance in three of four agents. Numerous other miRNA formed networks specific to a single agent. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 147. doi:10.1158/1538-7445.AM2011-147

Statistical techniques to construct assays for identifying likely responders to a treatment under evaluation from cell line genomic data

BackgroundDeveloping the right drugs for the right patients has become a mantra of drug development. In practice, it is very difficult to identify subsets of patients who will respond to a drug under evaluation. Most of the time, no single diagnostic will be available, and more complex decision rules will be required to define a sensitive population, using, for instance, mRNA expression, protein expression or DNA copy number. Moreover, diagnostic development will often begin with in-vitro cell-line data and a high-dimensional exploratory platform, only later to be transferred to a diagnostic assay for use with patient samples. In this manuscript, we present a novel approach to developing robust genomic predictors that are not only capable of generalizing from in-vitro to patient, but are also amenable to clinically validated assays such as qRT-PCR.MethodsUsing our approach, we constructed a predictor of sensitivity to dacetuzumab, an investigational drug for CD40-expressing malignancies such as lymphoma using genomic measurements of cell lines treated with dacetuzumab. Additionally, we evaluated several state-of-the-art prediction methods by independently pairing the feature selection and classification components of the predictor. In this way, we constructed several predictors that we validated on an independent DLBCL patient dataset. Similar analyses were performed on genomic measurements of breast cancer cell lines and patients to construct a predictor of estrogen receptor (ER) status.ResultsThe best dacetuzumab sensitivity predictors involved ten or fewer genes and accurately classified lymphoma patients by their survival and known prognostic subtypes. The best ER status classifiers involved one or two genes and led to accurate ER status predictions more than 85% of the time. The novel method we proposed performed as well or better than other methods evaluated.ConclusionsWe demonstrated the feasibility of combining feature selection techniques with classification methods to develop assays using cell line genomic measurements that performed well in patient data. In both case studies, we constructed parsimonious models that generalized well from cell lines to patients.

Hybridization modeling of oligonucleotide SNP arrays for accurate DNA copy number estimation

Affymetrix SNP arrays have been widely used for single-nucleotide polymorphism (SNP) genotype calling and DNA copy number variation inference. Although numerous methods have achieved high accuracy in these fields, most studies have paid little attention to the modeling of hybridization of probes to off-target allele sequences, which can affect the accuracy greatly. In this study, we address this issue and demonstrate that hybridization with mismatch nucleotides (HWMMN) occurs in all SNP probe-sets and has a critical effect on the estimation of allelic concentrations (ACs). We study sequence binding through binding free energy and then binding affinity, and develop a probe intensity composite representation (PICR) model. The PICR model allows the estimation of ACs at a given SNP through statistical regression. Furthermore, we demonstrate with cell-line data of known true copy numbers that the PICR model can achieve reasonable accuracy in copy number estimation at a single SNP locus, by using the ratio of the estimated AC of each sample to that of the reference sample, and can reveal subtle genotype structure of SNPs at abnormal loci. We also demonstrate with HapMap data that the PICR model yields accurate SNP genotype calls consistently across samples, laboratories and even across array platforms.

In vitro activity of bortezomib in cultures of patient tumour cells—potential utility in haematological malignancies

Bortezomib represents a new class of anti-cancer drugs, the proteasome inhibitors. We evaluated the in vitro activity of bortezomib with regard to tumour-type specificity and possible mechanisms of drug resistance in 115 samples of tumour cells from patients and in a cell-line panel, using the short-term fluorometric microculture cytotoxicity assay. Bortezomib generally showed dose-response curves with a steep slope. In patient cells, bortezomib was more active in haematological than in solid tumour samples. Myeloma and chronic myeloid leukaemia were the most sensitive tumour types although with great variability in drug response between the individual samples. Colorectal and kidney cancer samples were the least sensitive. In the cell-line panel, only small differences in response were seen between the different cell lines, and the proteasome inhibitors, lactacystin and MG 262, showed an activity pattern similar to that of bortezomib. The cell-line data suggest that resistance to bortezomib was not mediated by MRP-, PgP, GSH-; tubulin and topo II-associated MDR. Combination experiments indicated synergy between bortezomib and arsenic trioxide or irinotecan. The data support the current use of bortezomib but also points to its potential utility in other tumour types and in combination with cytotoxic drugs.

Order-restricted inference for ordered gene expression (ORIOGEN) data under heteroscedastic variances.

This article extends the order restricted inference approach for time-course or dose-response gene expression microarray data, introduced by Peddada and colleagues (2003) for the case when gene expression is heteroscedastic over time or dose. The new methodology uses an iterative algorithm to estimate mean expression at various times/doses when mean expression is subject to pre-defined patterns or profiles, known as order-restrictions. Simulation studies reveal that the resulting bootstrap-based methodology for gene selection maintains the false positive rate at the nominal level while competing well with ORIOGEN in terms of power. The proposed methodology is illustrated using a breast cancer cell-line data analyzed by Peddada and colleagues (2003).

New therapeutic strategies for soft tissue sarcomas.

The treatment of patients with metastatic soft tissue sarcomas (STS) is complex. There are limited agents available and many are associated with significant toxicity. When evaluating a patient with metastatic disease, physicians should ask themselves whether there is a role for surgery to render the patient free of disease. Combination chemotherapy in patients who have not received chemotherapy in the adjuvant setting is one option, particularly in a young patient with a good performance status. Sequential single-agent therapy for patients who are more elderly or debilitated by their disease may be more appropriate. Gemcitabine appears to be an agent with activity, particularly in patients with leiomyosarcomas. The data regarding prolonged gemcitabine infusions suggest improved activity that was predicted based on prolonged intracellular gemcitabine levels. Because of these data, the prolonged infusion schedule should be used. In addition, because of the paucity of effective agents, consideration of clinical trial participation for patients with newly diagnosed metastatic disease is appropriate, particularly in chemotherapy-insensitive histologies. The role of the newer agents (eg, ecteinascidin-743, epothilones, and mammalian target of rapamycin) is undefined. Ecteinascidin-743 has been the most extensively tested agent, and its ability to slow growth kinetics of a tumor and stabilize it clinically is intriguing. Data regarding the response to BMS-247550 will be published shortly and will help define the further role of epothilones in this disease. There is a preclinical rationale that makes the mammalian target of rapamycin inhibitors attractive for the treatment of muscle-derived neoplasms. In addition, there are cell-line data suggesting activity in rhabdomyosarcoma. These agents are being tested in adult STS and will likely be tested in pediatric histologies when there are more safety data available in that population. SU11248 will continue to be tested in patients refractory to imatinib mesylate and may well prove to be another active agent for patients with gastrointestinal stromal tumors. As depicted by the analysis of gemcitabine efficacy, agents with activity in a subgroup of STS may be overlooked by the "come one come all" approach to clinical trials in STS. Identifying key targets in specific STS will be helpful in the testing of newer molecularly targeted agents. Biologic differences will support histology-specific trials to better understand the activity of an agent in a specific disease site or specifically target a biologic pathway with relevance to the malignant potential of the disease. For future clinical trials in STS to achieve the goal of histology-specific trials, cooperative group and multi-institutional trials will be required to obtain the appropriate patients with these rare histologies. It will also be increasingly important to be committed to obtaining tumor tissue in these patients to validate hypotheses regarding tumor biology and the effectiveness of therapeutic agents.

Toxic-waste management in neurons

Insoluble cellular aggregates are associated with a number of neurodegenerative disorders such as Alzheimer and Parkinson disease as well as diseases caused by expansion of a polyglutamine tract, such as Huntington disease and the spinocerebellar ataxias. How do polyglutamine-containing proteins cause neurodegeneration? The mutant forms of these proteins tend to aggregate in ubiquitin-positive inclusions in the nucleus. On page 148, Christopher Cummings and colleagues now report that ataxin-1 aggregates also stain positive for proteasome components and the HDJ-2/HSDJ chaperones. These findings raise the possibility that the cell may actively attempt to degrade the polyglutamine-containing ataxin-1 but is unable to finish the job. Keeping the cellular waste-removal machinery tied up in this futile attempt might limit the degradation and removal of other proteins — and ultimately contribute to cellular degeneration. Cummings et al. go on to show that overexpression of the HDJ-2/HSDJ chaperone in HeLa cells expressing mutant ataxin-1 decreases the frequency of aggregate formation. While it is too early to conclude that limited chaperone capacity is involved in disease pathology, the cell-line data raise the possibility that increased levels of available chaperone molecules directly suppress aggregate formation through refolding or removal of misfolded polyglutamine proteins. As the authors discuss, the next step will be to test whether similar results can be obtained in the intact animal.