Neurotrophin signaling via TrkB and TrkC receptors promotes the growth of brain tumor-initiating cells.

Neurotrophins and their receptors are frequently expressed in malignant gliomas, yet their functions are largely unknown. Previously, we have shown that p75 neurotrophin receptor is required for glioma invasion and proliferation. However, the role of Trk receptors has not been examined. In this study, we investigated the importance of TrkB and TrkC in survival of brain tumor-initiating cells (BTICs). Here, we show that human malignant glioma tissues and also tumor-initiating cells isolated from fresh human malignant gliomas express the neurotrophin receptors TrkB and TrkC, not TrkA, and they also express neurotrophins NGF, BDNF, and neurotrophin 3 (NT3). Specific activation of TrkB and TrkC receptors by ligands BDNF and NT3 enhances tumor-initiating cell viability through activation of ERK and Akt pathways. Conversely, TrkB and TrkC knockdown or pharmacologic inhibition of Trk signaling decreases neurotrophin-dependent ERK activation and BTIC growth. Further, pharmacological inhibition of both ERK and Akt pathways blocked BDNF, and NT3 stimulated BTIC survival. Importantly, attenuation of BTIC growth by EGFR inhibitors could be overcome by activation of neurotrophin signaling, and neurotrophin signaling is sufficient for long term BTIC growth as spheres in the absence of EGF and FGF. Our results highlight a novel role for neurotrophin signaling in brain tumor and suggest that Trks could be a target for combinatorial treatment of malignant glioma.

p75 neurotrophin receptor cleavage by α- and γ-secretases is required for neurotrophin-mediated proliferation of brain tumor-initiating cells.

Malignant gliomas are highly invasive, proliferative, and resistant to treatment. Previously, we have shown that p75 neurotrophin receptor (p75NTR) is a novel mediator of invasion of human glioma cells. However, the role of p75NTR in glioma proliferation is unknown. Here we used brain tumor-initiating cells (BTICs) and show that BTICs express neurotrophin receptors (p75NTR, TrkA, TrkB, and TrkC) and their ligands (NGF, brain-derived neurotrophic factor, and neurotrophin 3) and secrete NGF. Down-regulation of p75NTR significantly decreased proliferation of BTICs. Conversely, exogenouous NGF stimulated BTIC proliferation through α- and γ-secretase-mediated p75NTR cleavage and release of its intracellular domain (ICD). In contrast, overexpression of the p75NTR ICD induced proliferation. Interestingly, inhibition of Trk signaling blocked NGF-stimulated BTIC proliferation and p75NTR cleavage, indicating a role of Trk in p75NTR signaling. Further, blocking p75NTR cleavage attenuated Akt activation in BTICs, suggesting role of Akt in p75NTR-mediated proliferation. We also found that p75NTR, α-secretases, and the four subunits of the γ-secretase enzyme were elevated in glioblastoma multiformes patients. Importantly, the ICD of p75NTR was commonly found in malignant glioma patient specimens, suggesting that the receptor is activated and cleaved in patient tumors. These results suggest that p75NTR proteolysis is required for BTIC proliferation and is a novel potential clinical target.

A recurrent mutation in the BMP type I receptor ACVR1 causes inherited and sporadic fibrodysplasia ossificans progressiva.

Fibrodysplasia ossificans progressiva (FOP) is a rare autosomal dominant disorder of skeletal malformations and progressive extraskeletal ossification. We mapped FOP to chromosome 2q23-24 by linkage analysis and identified an identical heterozygous mutation (617G --> A; R206H) in the glycine-serine (GS) activation domain of ACVR1, a BMP type I receptor, in all affected individuals examined. Protein modeling predicts destabilization of the GS domain, consistent with constitutive activation of ACVR1 as the underlying cause of the ectopic chondrogenesis, osteogenesis and joint fusions seen in FOP.

Iterative rank-order normalization of gene expression microarray data.

BACKGROUND: Many gene expression normalization algorithms exist for Affymetrix GeneChip microarrays. The most popular of these is RMA, primarily due to the precision and low noise produced during the process. A significant strength of this and similar approaches is the use of the entire set of arrays during both normalization and model-based estimation of signal. However, this leads to differing estimates of expression based on the starting set of arrays, and estimates can change when a single, additional chip is added to the set. Additionally, outlier chips can impact the signals of other arrays, and can themselves be skewed by the majority of the population. RESULTS: We developed an approach, termed IRON, which uses the best-performing techniques from each of several popular processing methods while retaining the ability to incrementally renormalize data without altering previously normalized expression. This combination of approaches results in a method that performs comparably to existing approaches on artificial benchmark datasets (i.e. spike-in) and demonstrates promising improvements in segregating true signals within biologically complex experiments. CONCLUSIONS: By combining approaches from existing normalization techniques, the IRON method offers several advantages. First, IRON normalization occurs pair-wise, thereby avoiding the need for all chips to be normalized together, which can be important for large data analyses. Secondly, the technique does not require similarity in signal distribution across chips for normalization, which can be important for maintaining biologically relevant differences in a heterogeneous background. Lastly, IRON introduces fewer post-processing artifacts, particularly in data whose behavior violates common assumptions. Thus, the IRON method provides a practical solution to common needs of expression analysis. A software implementation of IRON is available at [http://gene.moffitt.org/libaffy/].

TNFRSF10B polymorphisms and haplotypes associated with increased risk of death in non-small cell lung cancer.

Presently, there are few validated biomarkers that can predict survival or treatment response for non-small cell lung cancer (NSCLC) and most are based on tumor markers. Biomarkers based on germ line DNA variations represent a valuable complementary strategy, which could have translational implications by subclassifying patients to tailored, patient-specific treatment. We analyzed single nucleotide polymorphisms (SNPs) in 53 inflammation-related genes among 651 NSCLC patients. Multivariable Cox proportional hazard models, adjusted for lung cancer prognostic factors, were used to assess the association of genotypes and haplotypes with overall survival. Four of the top 15 SNPs associated with survival were located in the TNF-receptor superfamily member 10b (TNFRSF10B) gene. The T-allele of the top ranked SNP (rs11785599) was associated with a 41% increased risk of death (95% confidence interval [CI] = 1.16-1.70) and the other three TNFRSF10B SNPs (rs1047275, rs4460370 and rs883429) exhibited a 35% (95% CI = 1.11-1.65), 29% (95% CI = 1.06-1.57) and 24% (95% CI = 0.99-1.54) increased risk of death, respectively. Haplotype analyses revealed that the most common risk haplotype (TCTT) was associated with a 78% (95% CI = 1.25-2.54) increased risk of death compared with the low-risk haplotype (CGCC). When the data were stratified by treatment, the risk haplotypes exhibited statistically significantly increased risk of death among patients who had surgery only and no statistically significant effects among patients who had surgery and adjuvant chemotherapy. These data suggest that possessing one or more risk alleles in TNFRSF10B is associated with an increased risk of death. Validated germ line biomarkers may have potential important clinical implications by optimizing patient-specific treatment.

Proteomic contributions to personalized cancer care.

Cancer impacts each patient and family differently. Our current understanding of the disease is primarily limited to clinical hallmarks of cancer, but many specific molecular mechanisms remain elusive. Genetic markers can be used to determine predisposition to tumor development, but molecularly targeted treatment strategies that improve patient prognosis are not widely available for most cancers. Individualized care plans, also described as personalized medicine, still must be developed by understanding and implementing basic science research into clinical treatment. Proteomics holds great promise in contributing to the prevention and cure of cancer because it provides unique tools for discovery of biomarkers and therapeutic targets. As such, proteomics can help translate basic science discoveries into the clinical practice of personalized medicine. Here we describe how biological mass spectrometry and proteome analysis interact with other major patient care and research initiatives and present vignettes illustrating efforts in discovery of diagnostic biomarkers for ovarian cancer, development of treatment strategies in lung cancer, and monitoring prognosis and relapse in multiple myeloma patients.

IPEP: an in silico tool to examine proteolytic peptides for mass spectrometry.

UNLABELLED: Peptide-based proteomics supports identification and quantification as well as localization of post-translational modifications (PTMs) within proteins extracted from biological samples. The 'bottom-up' approach involves the digestion of proteins into peptide fragments that can be detected and sequenced with liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). A web-based application, iPEP, was developed to compare the effectiveness of different proteolytic digests in detecting specific sequences. Furthermore, peptide populations can be examined to help optimize detection of certain groups of proteins relative to the proteome and the digested peptidome. The application reports proteolytic peptide sequences, theoretical molecular weights and functional annotations using Gene Ontology (GO) terms. The iPEP tool can assist with experimental design by maximizing the detection of proteins, consensus sites and modified residues of interest for individual proteins or as part of large-scale proteomic assays. AVAILABILITY: http://ipep.moffitt.org

Integrated analysis of independent gene expression microarray datasets improves the predictability of breast cancer outcome.

BACKGROUND: Gene expression profiles based on microarray data have been suggested by many studies as potential molecular prognostic indexes of breast cancer. However, due to the confounding effect of clinical background, independent studies often obtained inconsistent results. The current study investigated the possibility to improve the quality and generality of expression profiles by integrated analysis of multiple datasets. Profiles of recurrence outcome were derived from two independent datasets and validated by a third dataset. RESULTS: The clinical background of patients significantly influenced the content and performance of expression profiles when the training samples were unbalanced. The integrated profiling of two independent datasets lead to higher classification accuracy (71.11% vs. 70.59%) and larger ROC curve area (0.789 vs. 0.767) of the testing samples. Cell cycle, especially M phase mitosis, was significantly overrepresented by the 60-gene profile obtained from integrated analysis (p < 0.0001). This profiles significantly differentiated poor and good prognosis in a third patient cohort (p = 0.003). Simulation procedures demonstrated that the change of profile specificity had more instant influence on the performance of expression profiles than the change of profile sensitivity. CONCLUSION: The current study confirmed that the gene expression profile generated by integrated analysis of multiple datasets achieved better prediction of breast cancer recurrence. However, the content and performance of profiles was confounded by clinical background of training patients. In future studies, prognostic profile applicable to the general population should be derived from more diversified and balanced patient cohorts in larger scale.

Use of the Total Cancer Care System to Enrich Screening for CD30-Positive Solid Tumors for Patient Enrollment Into a Brentuximab Vedotin Clinical Trial: A Pilot Study to Evaluate Feasibility.

BACKGROUND: One approach to identify patients who meet specific eligibility criteria for target-based clinical trials is to use patient and tumor registries to prescreen patient populations. OBJECTIVE: Here we demonstrate that the Total Cancer Care (TCC) Protocol, an ongoing, observational study, may provide a solution for rapidly identifying patients with CD30-positive tumors eligible for CD30-targeted therapies such as brentuximab vedotin. METHODS: The TCC patient gene expression profiling database was retrospectively screened for CD30 gene expression determined using HuRSTA-2a520709 Affymetrix arrays (GPL15048). Banked tumor tissue samples were used to determine CD30 protein expression by semiquantitative immunohistochemistry. Statistical comparisons of Z- and H-scores were performed using R statistical software (The R Foundation), and the predictive value, accuracy, sensitivity, and specificity of CD30 gene expression versus protein expression was estimated. RESULTS: As of March 2015, 120,887 patients have consented to the institutional review board-approved TCC Protocol. A total of 39,157 fresh frozen tumor specimens have been collected, from which over 14,000 samples have gene expression data available. CD30 RNA was expressed in a number of solid tumors; the highest median CD30 RNA expression was observed in primary tumors from lymph node, soft tissue (many sarcomas), lung, skin, and esophagus (median Z-scores 1.011, 0.399, 0.202, 0.152, and 1.011, respectively). High level CD30 gene expression significantly enriches for CD30-positive protein expression in breast, lung, skin, and ovarian cancer; accuracy ranged from 72% to 79%, sensitivity from 75% to 100%, specificity from 70% to 76%, positive predictive value from 20% to 40%, and negative predictive value from 95% to 100%. CONCLUSIONS: The TCC gene expression profiling database guided tissue selection that enriched for CD30 protein expression in a number of solid tumor types. Such an approach may improve screening efficiency for enrolling patients into biomarker-based clinical trials.

Simulating the contribution of a biospecimen and clinical data repository in a phase II clinical trial: A value of information analysis.

The potential contributions of a centralized data warehouse or repository in clinical research include the expedited accrual of subjects for phase II trials. Understanding the contribution of data warehouses that integrate clinical, biospecimen, and molecular data for the conduct of clinical trials is essential to inform private and public decisions on resource allocation and investment. We conducted a value of information analysis using data from recent trials at the Moffitt Cancer Center and simulated the potential reductions in trial size due to possible alternative scenarios of expedited accrual. In this study, we compared alternative data sets using a single model to assess value of information. Our findings suggest that the reductions in trial size range from 0% to 43%, depending on the amount of censoring in overall survival. The ability to expedite the accrual of patients for clinical trial studies using large data repositories that store data on inclusion/exclusion criteria and response to standard of care therapies demonstrated significant improvement in reducing the number of subjects needed to achieve similar end-results, as evaluated using value of information analysis with a limited number of parameters and a parsimonious model of overall survival.

Endosialin and Associated Protein Expression in Soft Tissue Sarcomas: A Potential Target for Anti-Endosialin Therapeutic Strategies.

Endosialin (CD248, TEM-1) is expressed in pericytes, tumor vasculature, tumor fibroblasts, and some tumor cells, including sarcomas, with limited normal tissue expression, and appears to play a key role in tumor-stromal interactions, including angiogenesis. Monoclonal antibodies targeting endosialin have entered clinical trials, including soft tissue sarcomas. We evaluated a cohort of 94 soft tissue sarcoma samples to assess the correlation between gene expression and protein expression by immunohistochemistry for endosialin and PDGFR-ß, a reported interacting protein, across available diagnoses. Correlations between the expression of endosialin and 13 other genes of interest were also examined. Within cohorts of soft tissue diagnoses assembled by tissue type (liposarcoma, leiomyosarcoma, undifferentiated sarcoma, and other), endosialin expression was significantly correlated with a better outcome. Endosialin expression was highest in liposarcomas and lowest in leiomyosarcomas. A robust correlation between protein and gene expression data for both endosialin and PDGFR-ß was observed. Endosialin expression positively correlated with PDGFR-ß and heparin sulphate proteoglycan 2 and negatively correlated with carbonic anhydrase IX. Endosialin likely interacts with a network of extracellular and hypoxia activated proteins in sarcomas and other tumor types. Since expression does vary across histologic groups, endosialin may represent a selective target in soft tissue sarcomas.

Examination of Epigenetic and other Molecular Factors Associated with mda-9/Syntenin Dysregulation in Cancer Through Integrated Analyses of Public Genomic Datasets.

mda-9/Syntenin (melanoma differentiation-associated gene 9) is a PDZ domain containing, cancer invasion-related protein. In this study, we employed multiple integrated bioinformatic approaches to identify the probable epigenetic factors, molecular pathways, and functionalities associated with mda-9 dysregulation during cancer progression. Analyses of publicly available genomic data (e.g., expression, copy number, methylation) from TCGA, GEO, ENCODE, and Human Protein Atlas projects led to the following observations: (a) mda-9 expression correlates with both copy number and methylation level of an intronic CpG site (cg1719774) located downstream of the CpG island, (b) cg1719774 methylation is a likely prognostic marker in glioma, (c) among 22 cancer types, melanoma exhibits the highest mda-9 level, and lowest level of methylation at cg1719774, (d) cg1719774 hypomethylation is also associated with histone modifications (at the mda-9 locus) indicative of more active transcription, (e) using Gene Set Enrichment Analysis (GSEA), and the Virtual Gene Overexpression or Repression (VIGOR) analytical scheme, we were able to predict mda-9's association with extracellular matrix organization (e.g., MMPs, collagen, integrins), IGFBP2 and NF-κB signaling pathways, phospholipid metabolism, cytokines (e.g., interleukins), CTLA-4, and components of complement cascade pathways. Indeed, previous publications have shown that many of the aforementioned genes and pathways are associated with mda-9's functionality.

Clonal architectures and driver mutations in metastatic melanomas.

To reveal the clonal architecture of melanoma and associated driver mutations, whole genome sequencing (WGS) and targeted extension sequencing were used to characterize 124 melanoma cases. Significantly mutated gene analysis using 13 WGS cases and 15 additional paired extension cases identified known melanoma genes such as BRAF, NRAS, and CDKN2A, as well as a novel gene EPHA3, previously implicated in other cancer types. Extension studies using tumors from another 96 patients discovered a large number of truncation mutations in tumor suppressors (TP53 and RB1), protein phosphatases (e.g., PTEN, PTPRB, PTPRD, and PTPRT), as well as chromatin remodeling genes (e.g., ASXL3, MLL2, and ARID2). Deep sequencing of mutations revealed subclones in the majority of metastatic tumors from 13 WGS cases. Validated mutations from 12 out of 13 WGS patients exhibited a predominant UV signature characterized by a high frequency of C->T transitions occurring at the 3' base of dipyrimidine sequences while one patient (MEL9) with a hypermutator phenotype lacked this signature. Strikingly, a subclonal mutation signature analysis revealed that the founding clone in MEL9 exhibited UV signature but the secondary clone did not, suggesting different mutational mechanisms for two clonal populations from the same tumor. Further analysis of four metastases from different geographic locations in 2 melanoma cases revealed phylogenetic relationships and highlighted the genetic alterations responsible for differential drug resistance among metastatic tumors. Our study suggests that clonal evaluation is crucial for understanding tumor etiology and drug resistance in melanoma.

JAK1 truncating mutations in gynecologic cancer define new role of cancer-associated protein tyrosine kinase aberrations.

Cancer-associated protein tyrosine kinase (PTK) mutations usually are gain-of-function (GOF) mutations that drive tumor growth and metastasis. We have found 50 JAK1 truncating mutations in 36 of 635 gynecologic tumors in the Total Cancer Care® (TCC®) tumor bank. Among cancer cell lines containing JAK1 truncating mutations in the Cancer Cell Line Encyclopedia databank, 68% are gynecologic cancer cells. Within JAK1 the K142, P430, and K860 frame-shift mutations were identified as hot spot mutation sites. Sanger sequencing of cancer cell lines, primary tumors, and matched normal tissues confirmed the JAK1 mutations and showed that these mutations are somatic. JAK1 mediates interferon (IFN)-γ-regulated tumor immune surveillance. Functional assays show that JAK1 deficient cancer cells are defective in IFN-γ-induced LMP2 and TAP1 expression, loss of which inhibits presentation of tumor antigens. These findings identify recurrent JAK1 truncating mutations that could contribute to tumor immune evasion in gynecologic cancers, especially in endometrial cancer.

Development and application of a novel metric to assess effectiveness of biomedical data.

OBJECTIVE: Design a metric to assess the comparative effectiveness of biomedical data elements within a study that incorporates their statistical relatedness to a given outcome variable as well as a measurement of the quality of their underlying data. MATERIALS AND METHODS: The cohort consisted of 874 patients with adenocarcinoma of the lung, each with 47 clinical data elements. The p value for each element was calculated using the Cox proportional hazard univariable regression model with overall survival as the endpoint. An attribute or A-score was calculated by quantification of an element's four quality attributes; Completeness, Comprehensiveness, Consistency and Overall-cost. An effectiveness or E-score was obtained by calculating the conditional probabilities of the p-value and A-score within the given data set with their product equaling the effectiveness score (E-score). RESULTS: The E-score metric provided information about the utility of an element beyond an outcome-related p value ranking. E-scores for elements age-at-diagnosis, gender and tobacco-use showed utility above what their respective p values alone would indicate due to their relative ease of acquisition, that is, higher A-scores. Conversely, elements surgery-site, histologic-type and pathological-TNM stage were down-ranked in comparison to their p values based on lower A-scores caused by significantly higher acquisition costs. CONCLUSIONS: A novel metric termed E-score was developed which incorporates standard statistics with data quality metrics and was tested on elements from a large lung cohort. Results show that an element's underlying data quality is an important consideration in addition to p value correlation to outcome when determining the element's clinical or research utility in a study.

Comparative effectiveness research in cancer genomics and precision medicine: current landscape and future prospects.

A major promise of genomic research is information that can transform health care and public health through earlier diagnosis, more effective prevention and treatment of disease, and avoidance of drug side effects. Although there is interest in the early adoption of emerging genomic applications in cancer prevention and treatment, there are substantial evidence gaps that are further compounded by the difficulties of designing adequately powered studies to generate this evidence, thus limiting the uptake of these tools into clinical practice. Comparative effectiveness research (CER) is intended to generate evidence on the "real-world" effectiveness compared with existing standards of care so informed decisions can be made to improve health care. Capitalizing on funding opportunities from the American Recovery and Reinvestment Act of 2009, the National Cancer Institute funded seven research teams to conduct CER in genomic and precision medicine and sponsored a workshop on CER on May 30, 2012, in Bethesda, Maryland. This report highlights research findings from those research teams, challenges to conducting CER, the barriers to implementation in clinical practice, and research priorities and opportunities in CER in genomic and precision medicine. Workshop participants strongly emphasized the need for conducting CER for promising molecularly targeted therapies, developing and supporting an integrated clinical network for open-access resources, supporting bioinformatics and computer science research, providing training and education programs in CER, and conducting research in economic and decision modeling.

12-Chemokine gene signature identifies lymph node-like structures in melanoma: potential for patient selection for immunotherapy?

We have interrogated a 12-chemokine gene expression signature (GES) on genomic arrays of 14,492 distinct solid tumors and show broad distribution across different histologies. We hypothesized that this 12-chemokine GES might accurately predict a unique intratumoral immune reaction in stage IV (non-locoregional) melanoma metastases. The 12-chemokine GES predicted the presence of unique, lymph node-like structures, containing CD20⁺ B cell follicles with prominent areas of CD3⁺ T cells (both CD4⁺ and CD8⁺ subsets). CD86⁺, but not FoxP3⁺, cells were present within these unique structures as well. The direct correlation between the 12-chemokine GES score and the presence of unique, lymph nodal structures was also associated with better overall survival of the subset of melanoma patients. The use of this novel 12-chemokine GES may reveal basic information on in situ mechanisms of the anti-tumor immune response, potentially leading to improvements in the identification and selection of melanoma patients most suitable for immunotherapy.

Developing a prototype system for integrating pharmacogenomics findings into clinical practice.

Findings from pharmacogenomics (PGx) studies have the potential to be applied to individualize drug therapy to improve efficacy and reduce adverse drug events. Researchers have identified factors influencing uptake of genomics in medicine, but little is known about the specific technical barriers to incorporating PGx into existing clinical frameworks. We present the design and development of a prototype PGx clinical decision support (CDS) system that builds on existing clinical infrastructure and incorporates semi-active and active CDS. Informing this work, we updated previous evaluations of PGx knowledge characteristics, and of how the CDS capabilities of three local clinical systems align with data and functional requirements for PGx CDS. We summarize characteristics of PGx knowledge and technical needs for implementing PGx CDS within existing clinical frameworks. PGx decision support rules derived from FDA drug labels primarily involve drug metabolizing genes, vary in maturity, and the majority support the post-analytic phase of genetic testing. Computerized provider order entry capabilities are key functional requirements for PGx CDS and were best supported by one of the three systems we evaluated. We identified two technical needs when building on this system, the need for (1) new or existing standards for data exchange to connect clinical data to PGx knowledge, and (2) a method for implementing semi-active CDS. Our analyses enhance our understanding of principles for designing and implementing CDS for drug therapy individualization and our current understanding of PGx characteristics in a clinical context. Characteristics of PGx knowledge and capabilities of current clinical systems can help govern decisions about CDS implementation, and can help guide decisions made by groups that develop and maintain knowledge resources such that delivery of content for clinical care is supported.

Informatics methods to enable sharing of quantitative imaging research data.

INTRODUCTION: The National Cancer Institute Quantitative Research Network (QIN) is a collaborative research network whose goal is to share data, algorithms and research tools to accelerate quantitative imaging research. A challenge is the variability in tools and analysis platforms used in quantitative imaging. Our goal was to understand the extent of this variation and to develop an approach to enable sharing data and to promote reuse of quantitative imaging data in the community. METHODS: We performed a survey of the current tools in use by the QIN member sites for representation and storage of their QIN research data including images, image meta-data and clinical data. We identified existing systems and standards for data sharing and their gaps for the QIN use case. We then proposed a system architecture to enable data sharing and collaborative experimentation within the QIN. RESULTS: There are a variety of tools currently used by each QIN institution. We developed a general information system architecture to support the QIN goals. We also describe the remaining architecture gaps we are developing to enable members to share research images and image meta-data across the network. CONCLUSIONS: As a research network, the QIN will stimulate quantitative imaging research by pooling data, algorithms and research tools. However, there are gaps in current functional requirements that will need to be met by future informatics development. Special attention must be given to the technical requirements needed to translate these methods into the clinical research workflow to enable validation and qualification of these novel imaging biomarkers.

Radiomics: the process and the challenges.

"Radiomics" refers to the extraction and analysis of large amounts of advanced quantitative imaging features with high throughput from medical images obtained with computed tomography, positron emission tomography or magnetic resonance imaging. Importantly, these data are designed to be extracted from standard-of-care images, leading to a very large potential subject pool. Radiomics data are in a mineable form that can be used to build descriptive and predictive models relating image features to phenotypes or gene-protein signatures. The core hypothesis of radiomics is that these models, which can include biological or medical data, can provide valuable diagnostic, prognostic or predictive information. The radiomics enterprise can be divided into distinct processes, each with its own challenges that need to be overcome: (a) image acquisition and reconstruction, (b) image segmentation and rendering, (c) feature extraction and feature qualification and (d) databases and data sharing for eventual (e) ad hoc informatics analyses. Each of these individual processes poses unique challenges. For example, optimum protocols for image acquisition and reconstruction have to be identified and harmonized. Also, segmentations have to be robust and involve minimal operator input. Features have to be generated that robustly reflect the complexity of the individual volumes, but cannot be overly complex or redundant. Furthermore, informatics databases that allow incorporation of image features and image annotations, along with medical and genetic data, have to be generated. Finally, the statistical approaches to analyze these data have to be optimized, as radiomics is not a mature field of study. Each of these processes will be discussed in turn, as well as some of their unique challenges and proposed approaches to solve them. The focus of this article will be on images of non-small-cell lung cancer.