A Network Flow-Based Method to Predict Anticancer Drug Sensitivity

Predicting anticancer drug sensitivity can enhance the ability to individualize patient treatment, thus making development of cancer therapies more effective and safe. In this paper, we present a new network flow-based method, which utilizes the topological structure of pathways, for predicting anticancer drug sensitivities. Mutations and copy number alterations of cancer-related genes are assumed to change the pathway activity, and pathway activity difference before and after drug treatment is used as a measure of drug response. In our model, Contributions from different genetic alterations are considered as free parameters, which are optimized by the drug response data from the Cancer Genome Project (CGP). 10-fold cross validation on CGP data set showed that our model achieved comparable prediction results with existing elastic net model using much less input features.     

DISIS: Prediction of Drug Response through an Iterative Sure Independence Screening

Prediction of drug response based on genomic alterations is an important task in the research of personalized medicine. Current elastic net model utilized a sure independence screening to select relevant genomic features with drug response, but it may neglect the combination effect of some marginally weak features. In this work, we applied an iterative sure independence screening scheme to select drug response relevant features from the Cancer Cell Line Encyclopedia (CCLE) dataset. For each drug in CCLE, we selected up to 40 features including gene expressions, mutation and copy number alterations of cancer-related genes, and some of them are significantly strong features but showing weak marginal correlation with drug response vector. Lasso regression based on the selected features showed that our prediction accuracies are higher than those by elastic net regression for most drugs.     

Recurrent BRAF kinase fusions in melanocytic tumors offer an opportunity for targeted therapy

BRAF is the most prevalent oncogene and an important therapeutic target in melanoma. In some cancers, BRAF is activated by rearrangements that fuse its kinase domain to 5′ partner genes. We examined 848 comparative genomic hybridization profiles of melanocytic tumors and found copy number transitions within BRAF in 10 tumors, of which six could be further characterized by sequencing. In all, the BRAF kinase domain was fused in‐frame to six N‐terminal partners. No other mutations were identified in melanoma oncogenes. One of the seven melanoma cell lines without known oncogenic mutations harbored a similar BRAF fusion, which constitutively activated the MAP kinase pathway. Sorafenib, but not vemurafenib, could block MAP kinase pathway activation and proliferation of the cell line at clinically relevant concentrations, whereas BRAF^V600E mutant melanoma cell lines were significantly more sensitive to vemurafenib. The patient from whom the cell line was derived showed a durable clinical response to sorafenib.     

ROCK1 is a potential combinatorial drug target for BRAF mutant melanoma

Treatment of BRAF mutant melanomas with specific BRAF inhibitors leads to tumor remission. However, most patients eventually relapse due to drug resistance. Therefore, we designed an integrated strategy using (phospho)proteomic and functional genomic platforms to identify drug targets whose inhibition sensitizes melanoma cells to BRAF inhibition. We found many proteins to be induced upon PLX4720 (BRAF inhibitor) treatment that are known to be involved in BRAF inhibitor resistance, including FOXD3 and ErbB3. Several proteins were down‐regulated, including Rnd3, a negative regulator of ROCK1 kinase. For our genomic approach, we performed two parallel shRNA screens using a kinome library to identify genes whose inhibition sensitizes to BRAF or ERK inhibitor treatment. By integrating our functional genomic and (phospho)proteomic data, we identified ROCK1 as a potential drug target for BRAF mutant melanoma. ROCK1 silencing increased melanoma cell elimination when combined with BRAF or ERK inhibitor treatment. Translating this to a preclinical setting, a ROCK inhibitor showed augmented melanoma cell death upon BRAF or ERK inhibition in vitro. These data merit exploration of ROCK1 as a target in combination with current BRAF mutant melanoma therapies.     

Strongly preserved modules between cancer tissue and cell line contribute to drug resistance analysis across multiple cancer types

The existence and emergence of drug resistance in tumor cells is the main burden of cancer treatment. Most cancer drug resistance analyses are based entirely on cell line data and ignore the discordance between human tumors and cell lines, leading to biased preclinical model transformation. Based on cancer tissue data in TCGA and cancer cell line data in CCLE, this study identified and excluded non-preserved module (NP module) between cancer tissue and cell lines. We used strongly preserved module (SP module) for clinically relevant drug resistance analysis and identified 2068 “cancer-drug-module” pairs of 7 cancer types and 212 drugs based on data in GDSC. Furthermore, we identified potentially ineffective combination therapy (PICT) from multiple perspectives. Finally, we found 1608 sets of predictors that can predict drug response. These results provide insights and clues for the clinical selection of effective chemotherapy drugs to overcome cancer resistance in a new perspective.     

NCI-60 Whole Exome Sequencing and Pharmacological CellMiner Analyses

Exome sequencing provides unprecedented insights into cancer biology and pharmacological response. Here we assess these two parameters for the NCI-60, which is among the richest genomic and pharmacological publicly available cancer cell line databases. Homozygous genetic variants that putatively affect protein function were identified in 1,199 genes (approximately 6% of all genes). Variants that are either enriched or depleted compared to non-cancerous genomes, and thus may be influential in cancer progression and differential drug response were identified for 2,546 genes. Potential gene knockouts are made available. Assessment of cell line response to 19,940 compounds, including 110 FDA-approved drugs, reveals ≈80-fold range in resistance versus sensitivity response across cell lines. 103,422 gene variants were significantly correlated with at least one compound (at p<0.0002). These include genes of known pharmacological importance such as IGF1R, BRAF, RAD52, MTOR, STAT2 and TSC2 as well as a large number of candidate genes such as NOM1, TLL2, and XDH. We introduce two new web-based CellMiner applications that enable exploration of variant-to-compound relationships for a broad range of researchers, especially those without bioinformatics support. The first tool, “Genetic variant versus drug visualization”, provides a visualization of significant correlations between drug activity-gene variant combinations. Examples are given for the known vemurafenib-BRAF, and novel ifosfamide-RAD52 pairings. The second, “Genetic variant summation” allows an assessment of cumulative genetic variations for up to 150 combined genes together; and is designed to identify the variant burden for molecular pathways or functional grouping of genes. An example of its use is provided for the EGFR-ERBB2 pathway gene variant data and the identification of correlated EGFR, ERBB2, MTOR, BRAF, MEK and ERK inhibitors. The new tools are implemented as an updated web-based CellMiner version, for which the present publication serves as a compendium.     

JFCR39, a panel of 39 human cancer cell lines, and its application in the discovery and development of anticancer drugs

Over the past few decades, panels of human cancer cell lines have made a significant contribution to the discovery and development of anticancer drugs. The National Cancer Institute 60 (NCI60), which consists of 60 cell lines from various human cancer types, remains the most powerful human cancer cell line panel for high throughput screening of anticancer drugs. The development of JFCR39, comprising a panel of 39 human cancer cell lines coupled with a drug-activity database, was based on NCI60. Like NCI60, JFCR39 not only provides disease-oriented information but can also predict the action mechanism or molecular target of a given antitumor agent by utilizing the COMPARE algorithm. The molecular targets of ZSTK474 as well as several other antitumor agents have been identified by using JFCR39 and some of these compounds have since entered clinical trials. In this review, we will describe human cancer cell line panels particularly JFCR39 and its application in the discovery and/or development of anticancer drug candidates.     

Machine Learning Prediction of Cancer Cell Sensitivity to Drugs Based on Genomic and Chemical Properties

Predicting the response of a specific cancer to a therapy is a major goal in modern oncology that should ultimately lead to a personalised treatment. High-throughput screenings of potentially active compounds against a panel of genomically heterogeneous cancer cell lines have unveiled multiple relationships between genomic alterations and drug responses. Various computational approaches have been proposed to predict sensitivity based on genomic features, while others have used the chemical properties of the drugs to ascertain their effect. In an effort to integrate these complementary approaches, we developed machine learning models to predict the response of cancer cell lines to drug treatment, quantified through IC₅₀ values, based on both the genomic features of the cell lines and the chemical properties of the considered drugs. Models predicted IC₅₀ values in a 8-fold cross-validation and an independent blind test with coefficient of determination R² of 0.72 and 0.64 respectively. Furthermore, models were able to predict with comparable accuracy (R² of 0.61) IC50s of cell lines from a tissue not used in the training stage. Our in silico models can be used to optimise the experimental design of drug-cell screenings by estimating a large proportion of missing IC₅₀ values rather than experimentally measuring them. The implications of our results go beyond virtual drug screening design: potentially thousands of drugs could be probed in silico to systematically test their potential efficacy as anti-tumour agents based on their structure, thus providing a computational framework to identify new drug repositioning opportunities as well as ultimately be useful for personalized medicine by linking the genomic traits of patients to drug sensitivity.     

Genomic signatures to guide the use of chemotherapeutics

Using in vitro drug sensitivity data coupled with Affymetrix microarray data, we developed gene expression signatures that predict sensitivity to individual chemotherapeutic drugs. Each signature was validated with response data from an independent set of cell line studies. We further show that many of these signatures can accurately predict clinical response in individuals treated with these drugs. Notably, signatures developed to predict response to individual agents, when combined, could also predict response to multidrug regimens. Finally, we integrated the chemotherapy response signatures with signatures of oncogenic pathway deregulation to identify new therapeutic strategies that make use of all available drugs. The development of gene expression profiles that can predict response to commonly used cytotoxic agents provides opportunities to better use these drugs, including using them in combination with existing targeted therapies.     

Network quantification of EGFR signaling unveils potential for targeted combination therapy

The epidermal growth factor receptor (EGFR) signaling network is activated in most solid tumors, and small‐molecule drugs targeting this network are increasingly available. However, often only specific combinations of inhibitors are effective. Therefore, the prediction of potent combinatorial treatments is a major challenge in targeted cancer therapy. In this study, we demonstrate how a model‐based evaluation of signaling data can assist in finding the most suitable treatment combination. We generated a perturbation data set by monitoring the response of RAS/PI3K signaling to combined stimulations and inhibitions in a panel of colorectal cancer cell lines, which we analyzed using mathematical models. We detected that a negative feedback involving EGFR mediates strong cross talk from ERK to AKT. Consequently, when inhibiting MAPK, AKT activity is increased in an EGFR‐dependent manner. Using the model, we predict that in contrast to single inhibition, combined inactivation of MEK and EGFR could inactivate both endpoints of RAS, ERK and AKT. We further could demonstrate that this combination blocked cell growth in BRAF‐ as well as KRAS‐mutated tumor cells, which we confirmed using a xenograft model.     

A pharmacogenomic method for individualized prediction of drug sensitivity

Identifying the best drug for each cancer patient requires an efficient individualized strategy. We present MATCH (M erging genomic and pharmacologic A nalyses for T herapy CH oice), an approach using public genomic resources and drug testing of fresh tumor samples to link drugs to patients. Valproic acid (VPA) is highlighted as a proof‐of‐principle. In order to predict specific tumor types with high probability of drug sensitivity, we create drug response signatures using publically available gene expression data and assess sensitivity in a data set of >40 cancer types. Next, we evaluate drug sensitivity in matched tumor and normal tissue and exclude cancer types that are no more sensitive than normal tissue. From these analyses, breast tumors are predicted to be sensitive to VPA. A meta‐analysis across breast cancer data sets shows that aggressive subtypes are most likely to be sensitive to VPA, but all subtypes have sensitive tumors. MATCH predictions correlate significantly with growth inhibition in cancer cell lines and three‐dimensional cultures of fresh tumor samples. MATCH accurately predicts reduction in tumor growth rate following VPA treatment in patient tumor xenografts. MATCH uses genomic analysis with in vitro testing of patient tumors to select optimal drug regimens before clinical trial initiation.     

Systematic Identification of Combinatorial Drivers and Targets in Cancer Cell Lines

There is an urgent need to elicit and validate highly efficacious targets for combinatorial intervention from large scale ongoing molecular characterization efforts of tumors. We established an in silico bioinformatic platform in concert with a high throughput screening platform evaluating 37 novel targeted agents in 669 extensively characterized cancer cell lines reflecting the genomic and tissue-type diversity of human cancers, to systematically identify combinatorial biomarkers of response and co-actionable targets in cancer. Genomic biomarkers discovered in a 141 cell line training set were validated in an independent 359 cell line test set. We identified co-occurring and mutually exclusive genomic events that represent potential drivers and combinatorial targets in cancer. We demonstrate multiple cooperating genomic events that predict sensitivity to drug intervention independent of tumor lineage. The coupling of scalable in silico and biologic high throughput cancer cell line platforms for the identification of co-events in cancer delivers rational combinatorial targets for synthetic lethal approaches with a high potential to pre-empt the emergence of resistance.     

Activities of multiple cancer-related pathways are associated with BRAF mutation and predict the resistance to BRAF/MEK inhibitors in melanoma cells

Drug resistance is a major obstacle in the targeted therapy of melanoma using BRAF/MEK inhibitors. This study was to identify BRAF V600E-associated oncogenic pathways that predict resistance of BRAF-mutated melanoma to BRAF/MEK inhibitors. We took in silico approaches to analyze the activities of 24 cancer-related pathways in melanoma cells and identify those whose activation was associated with BRAF V600E and used the support vector machine (SVM) algorithm to predict the resistance of BRAF-mutated melanoma cells to BRAF/MEK inhibitors. We then experimentally confirmed the in silico findings. In a microarray gene expression dataset of 63 melanoma cell lines, we found that activation of multiple oncogenic pathways preferentially occurred in BRAF-mutated melanoma cells. This finding was reproduced in 5 additional independent melanoma datasets. Further analysis of 46 melanoma cell lines that harbored BRAF mutation showed that 7 pathways, including TNFα, EGFR, IFNα, hypoxia, IFNγ, STAT3, and MYC, were significantly differently expressed in AZD6244-resistant compared with responsive melanoma cells. A SVM classifier built on this 7-pathway activation pattern correctly predicted the response of 10 BRAF-mutated melanoma cell lines to the MEK inhibitor AZD6244 in our experiments. We experimentally showed that TNFα, EGFR, IFNα, and IFNγ pathway activities were also upregulated in melanoma cell A375 compared with its sub-line DRO, while DRO was much more sensitive to AZD6244 than A375. In conclusion, we have identified specific oncogenic pathways preferentially activated in BRAF-mutated melanoma cells and a pathway pattern that predicts resistance of BRAF-mutated melanoma to BRAF/MEK inhibitors, providing novel clinical implications for melanoma therapy.     

In vitro pharmacogenomic database and chemosensitivity predictive genes in gastric cancer

Gastric cancer is one of the most common cancers worldwide, and there are clinical caveats in predicting tumor response to chemotherapy. This study describes the construction of an in vitro pharmacogenomic database, and the selection of genes associated with chemosensitivity in gastric cancer cell lines. Gene expression and chemosensitivity databases were integrated using the Pearson correlation coefficient to give the GC-matrix. The 85 genes were selected that were commonly associated with chemosensitivity of the major anticancer drugs. We then focused on the genes that were highly correlated with each specific drug. Classification of cell lines based on the set of genes associated with each drug was consistent with the division into resistant or sensitive groups according to the chemosensitivity results. The GC-matrix of the gastric cancer cell line database was used to identify different sets of chemosensitivity-related genes for specific drugs or multiple drugs.     

In silico drug combination discovery for personalized cancer therapy

Background

Drug combination therapy, which is considered as an alternative to single drug therapy, can potentially reduce resistance and toxicity, and have synergistic efficacy. As drug combination therapies are widely used in the clinic for hypertension, asthma, and AIDS, they have also been proposed for the treatment of cancer. However, it is difficult to select and experimentally evaluate effective combinations because not only is the number of cancer drug combinations extremely large but also the effectiveness of drug combinations varies depending on the genetic variation of cancer patients. A computational approach that prioritizes the best drug combinations considering the genetic information of a cancer patient is necessary to reduce the search space.

Results

We propose an in-silico method for personalized drug combination therapy discovery. We predict the synergy between two drugs and a cell line using genomic information, targets of drugs, and pharmacological information. We calculate and predict the synergy scores of 583 drug combinations for 31 cancer cell lines. For feature dimension reduction, we select the mutations or expression levels of the genes in cancer-related pathways. We also used various machine learning models. Extremely Randomized Trees (ERT), a tree-based ensemble model, achieved the best performance in the synergy score prediction regression task. The correlation coefficient between the synergy scores predicted by ERT and the actual observations is 0.738. To compare with an existing drug combination synergy classification model, we reformulate the problem as a binary classification problem by thresholding the synergy scores. ERT achieved an F1 score of 0.954 when synergy scores of 20 and -20 were used as the threshold, which is 8.7% higher than that obtained by the state-of-the-art baseline model. Moreover, the model correctly predicts the most synergistic combination, from approximately 100 candidate drug combinations, as the top choice for 15 out of the 31 cell lines. For 28 out of the 31 cell lines, the model predicts the most synergistic combination in the top 10 of approximately 100 candidate drug combinations. Finally, we analyze the results, generate synergistic rules using the features, and validate the rules through the literature survey.

Conclusion

Using various types of genomic information of cancer cell lines, targets of drugs, and pharmacological information, a drug combination synergy prediction pipeline is proposed. The pipeline regresses the synergy level between two drugs and a cell line as well as classifies if there exists synergy or antagonism between them. Discovering new drug combinations by our pipeline may improve personalized cancer therapy.

     

一种预测个体肿瘤的抗癌药物反应分类计算模型及其应用

目的 不同患者对同一抗癌药物的反应可能不同,了解患者之间对抗癌药物的反应差异对癌症精准医疗具有重大参考价值。方法 高通量测序数据为构建抗癌药物反应分类预测模型提供了强大的数据支撑。针对两大经典数据集癌症细胞百科全书(CCLE)和癌症药物敏感性基因组学数据集(GDSC),本文提出了基于最大相关最小冗余(mRMR)算法和支持向量机(SVM)的计算模型mRMR-SVM。利用基因表达数据,通过方差排序和mRMR算法提取特征基因,借助SVM实现抗癌药物对细胞系的“敏感-抑制”二分类预测。结果 对于CCLE中的22种药物,mRMR-SVM的平均准确率为0.904;对于GDSC中的11种药物,平均准确率为0.851。结论 mRMR-SVM不仅在预测性能方面优于传统的支持向量机、随机森林、深度反应森林、深度神经网络和细胞系-药物复杂网络模型,而且具有良好的泛化能力,对于三类特定组织的抗癌药物反应分类预测也取得了令人满意的结果。此外,mRMR-SVM可以识别与癌症发生发展密切相关的生物标志物。     

Predictive Modeling of Drug Response in Non-Hodgkin’s Lymphoma

We combine mathematical modeling with experiments in living mice to quantify the relative roles of intrinsic cellular vs. tissue-scale physiological contributors to chemotherapy drug resistance, which are difficult to understand solely through experimentation. Experiments in cell culture and in mice with drug-sensitive (Eµ-myc/Arf-/-) and drug-resistant (Eµ-myc/p53-/-) lymphoma cell lines were conducted to calibrate and validate a mechanistic mathematical model. Inputs to inform the model include tumor drug transport characteristics, such as blood volume fraction, average geometric mean blood vessel radius, drug diffusion penetration distance, and drug response in cell culture. Model results show that the drug response in mice, represented by the fraction of dead tumor volume, can be reliably predicted from these inputs. Hence, a proof-of-principle for predictive quantification of lymphoma drug therapy was established based on both cellular and tissue-scale physiological contributions. We further demonstrate that, if the in vitro cytotoxic response of a specific cancer cell line under chemotherapy is known, the model is then able to predict the treatment efficacy in vivo. Lastly, tissue blood volume fraction was determined to be the most sensitive model parameter and a primary contributor to drug resistance.     

Impact of RNA degradation on gene expression profiling

Background

Cancer is a heterogeneous disease caused by genomic aberrations and characterized by significant variability in clinical outcomes and response to therapies. Several subtypes of common cancers have been identified based on alterations of individual cancer genes, such as HER2, EGFR, and others. However, cancer is a complex disease driven by the interaction of multiple genes, so the copy number status of individual genes is not sufficient to define cancer subtypes and predict responses to treatments. A classification based on genome-wide copy number patterns would be better suited for this purpose.

Method

To develop a more comprehensive cancer taxonomy based on genome-wide patterns of copy number abnormalities, we designed an unsupervised classification algorithm that identifies genomic subgroups of tumors. This algorithm is based on a modified genomic Non-negative Matrix Factorization (gNMF) algorithm and includes several additional components, namely a pilot hierarchical clustering procedure to determine the number of clusters, a multiple random initiation scheme, a new stop criterion for the core gNMF, as well as a 10-fold cross-validation stability test for quality assessment.

Result

We applied our algorithm to identify genomic subgroups of three major cancer types: non-small cell lung carcinoma (NSCLC), colorectal cancer (CRC), and malignant melanoma. High-density SNP array datasets for patient tumors and established cell lines were used to define genomic subclasses of the diseases and identify cell lines representative of each genomic subtype. The algorithm was compared with several traditional clustering methods and showed improved performance. To validate our genomic taxonomy of NSCLC, we correlated the genomic classification with disease outcomes. Overall survival time and time to recurrence were shown to differ significantly between the genomic subtypes.

Conclusions

We developed an algorithm for cancer classification based on genome-wide patterns of copy number aberrations and demonstrated its superiority to existing clustering methods. The algorithm was applied to define genomic subgroups of three cancer types and identify cell lines representative of these subgroups. Our data enabled the assembly of representative cell line panels for testing drug candidates.