Search

Background Parametric feature selection methods for machine learning and association studies based on genetic data are not robust with respect to outliers or influential observations. While rank-based, distribution-free statistics offer a robust alternative to parametric methods, their practical utility can be limited, as they demand significant computational resources when analyzing high-dimensional data. For genetic studies that seek to identify variants, the hypothesis is constrained, since it is typically assumed that the effect of the genotype on the phenotype is monotone (e.g., an additive genetic effect). Similarly, predictors for machine learning applications may have natural ordering constraints. Cross-validation for feature selection in these high-dimensional contexts necessitates highly efficient computational algorithms for the robust evaluation of many features. Results We have developed an R extension package, fastJT, for conducting genome-wide association studies and feature selection for machine learning using the Jonckheere-Terpstra statistic for constrained hypotheses. The kernel of the package features an efficient algorithm for calculating the statistics, replacing the pairwise comparison and counting processes with a data sorting and searching procedure, reducing computational complexity from O(n2) to O(n log(n)). The computational efficiency is demonstrated through extensive benchmarking, and example applications to real data are presented. Conclusions fastJT is an open-source R extension package, applying the Jonckheere-Terpstra statistic for robust feature selection for machine learning and association studies. The package implements an efficient algorithm which leverages internal information among the samples to avoid unnecessary computations, and incorporates shared-memory parallel programming to further boost performance on multi-core machines. Electronic supplementary material The online version of this article (10.1186/s12859-019-2869-3) contains supplementary material, which is available to authorized users.

The objective of this study was to evaluate effects of whole body radiation exposure early in life on influenza vaccination immune responses much later in life. A total of 292 volunteers recruited from the cohort members of ongoing Adult Health Study (AHS) of Japanese atomic bomb (A-bomb) survivors completed this observational study spanning two influenza seasons (2011-2012 and 2012-2013). Peripheral blood samples were collected prior to and three weeks after vaccination. Serum hemagglutination inhibition (HAI) antibody titers were measured as well as concentrations of 25 cytokines and chemokines in culture supernatant from peripheral blood mononuclear cells, with and without in vitro stimulation with influenza vaccine. We found that influenza vaccination modestly enhanced serum HAI titers in this unique cohort of elderly subjects, with seroprotection ranging from 18 to 48% for specific antigen/season combinations. Twelve percent of subjects were seroprotected against all three vaccine antigens post-vaccination. Males were generally more likely to be seroprotected for one or more antigens post-vaccination, with no differences in vaccine responses based on age at vaccination or radiation exposure in early life. These results show that early life exposure to ionizing radiation does not prevent responses of elderly A-bomb survivors to seasonal influenza vaccine.

Purpose Ionizing radiation causes acute damage to hematopoietic and immune cells, but the long-term immunologic consequences of irradiation are poorly understood. We therefore performed a prospective study of the delayed immune effects of radiation using a rhesus macaque model. Methods and Materials Ten macaques received 4 Gy high-energy x-ray total body irradiation (TBI) and 6 control animals received sham irradiation. TBI caused transient lymphopenia that resolved over several weeks. Once white blood cell counts recovered, flow cytometry was used to immunophenotype the circulating adaptive immune cell populations 4, 9, and 21 months after TBI. Data were fit using a mixed-effects model to determine age-dependent, radiation-dependent, and interacting effects. T cell receptor (TCR) sequencing and quantification of TCR Excision Circles were used to determine relative contributions of thymopoiesis and peripheral expansion to T cell repopulation. Two years after TBI, the cohort was vaccinated with a 23-valent pneumococcal polysaccharide vaccine and a tetravalent influenza hemagglutinin vaccine. Results Aging, but not TBI, led to significant changes in the frequencies of dendritic cells, CD4 and CD8 T cells, and B cells. However, irradiated animals exhibited increased frequencies of central memory T cells and decreased frequencies of naive T cells. These consequences of irradiation were time-dependent and more prolonged in the CD8 T cell population. Irradiation led to transient increases in CD8+ T cell TCR Excision Circles and had no significant effect on TCR sequence entropy, indicating T cell recovery was partially mediated by thymopoiesis. Animals that were irradiated and then vaccinated showed normal immunoglobulin G binding and influenza neutralization titers in response to the 4 protein antigens but weaker immunoglobulin G binding titers to 10 of the 23 polysaccharide antigens. Conclusions These findings indicate that TBI causes subtle but long-lasting immune defects that are evident years after recovery from lymphopenia.

Total body irradiation (TBI) damages hematopoietic cells in the bone marrow and thymus; however, the long-term effects of irradiation with aging remain unclear. In this study, we found that the impact of radiation on thymopoiesis in mice varied by sex and dose but, overall, thymopoiesis remained suppressed for ≥12 mo after a single exposure. Male and female mice showed a long-term dose-dependent reduction in thymic cKit+ lymphoid progenitors that was maintained throughout life. Damage to hematopoietic stem cells (HSCs) in the bone marrow was dose dependent, with as little as 0.5 Gy causing a significant long-term reduction. In addition, the potential for T lineage commitment was radiation sensitive with aging. Overall, the impact of irradiation on the hematopoietic lineage was more severe in females. In contrast, the rate of decline in thymic epithelial cell numbers with age was radiation-sensitive only in males, and other characteristics including Ccl25 transcription were unaffected. Taken together, these data suggest that long-term suppression of thymopoiesis after sublethal irradiation was primarily due to fewer progenitors in the BM combined with reduced potential for T lineage commitment. A single irradiation dose also caused synchronization of thymopoiesis, with a periodic thymocyte differentiation profile persisting for at least 12 mo postirradiation. This study suggests that the number and capability of HSCs for T cell production can be dramatically and permanently damaged after a single relatively low TBI dose, accelerating aging-associated thymic involution. Our findings may impact evaluation and therapeutic intervention of human TBI events.

Mast cells (MCs) are known to regulate innate and adaptive immunity. MC activators have recently been described as safe and effective vaccine adjuvants. However, many currently known MC activators are inadequate for in vivo applications, and research on identifying novel MC activators is limited. In this study, we identified novel MC activators by employing high-throughput screening (HTS) assays using approximately 55,000 small molecules. Data sets obtained by the primary HTS assays were statistically evaluated using quality control rules and the B-score calculation, and compounds with B-scores of > 3.0 were chosen as mast cell activators (hits). These hits were re-evaluated with secondary and tertiary HTS assays, followed by further statistical analysis. From these hits, we selected 15 compounds that caused degranulation in murine and human MCs with potential for flexible chemical modification for further study. Among these 15 compounds, ST101036, ST029248 and ST026567 exhibited higher degranulation potency than other hit compounds in both human and mouse MCs. Additionally, the 15 compounds identified promote de novo synthesis of cytokines and induce the release of eicosanoids from human and mouse MCs. High-throughput screening enabled us to identify small molecule MC activators with unique properties that may be useful as vaccine adjuvants.

Angiogenesis is essential in tumor biology and is regulated by vascular endothelial growth factor (VEGF) ligands and receptors. Here we aimed to discover genetic variants associated with levels of circulating angiogenic proteins in cancer patients. Plasma was collected at baseline in 216 pancreatic and 114 colorectal cancer patients. Thirty-one angiogenic proteins were measured by ELISA. 484,523 Single Nucleotide Polymorphisms (SNP) were tested for association with plasma levels for each protein in pancreatic cancer patients. Three top-ranked hits were then genotyped in colorectal cancer patients, where associations with the same proteins were measured. The results demonstrated rs2284284 and MCP1 (P-value = 6.7e–08), rs7504372 and VEGF-C (P-value = 9.8e–09), and rs7767396 and VEGF-A (P-value = 5.8e–09) were SNP-protein pairs identified in pancreatic cancer patients. In colorectal cancer patients, only rs7767396 (A > G) and VEGF-A was validated (P-value = 5.18e–05). The AA genotype of rs7767396 exhibited 2.04–2.3 and 2.7–3.4-fold higher VEGF-A levels than those with AG and GG genotypes. The G allele of rs7767396 reduces binding of the NF-AT1 transcription factor. In conclusion, a common genetic variant predicts the plasma levels of VEGF-A in cancer patients through altered binding of NF-AT1.

An interlaboratory comparison of radiation dosimetry was conducted to determine the accuracy of doses being used experimentally for animal exposures within a large multi-institutional research project. The background and approach to this effort are described and discussed in terms of basic findings, problems and solutions.Dosimetry tests were carried out utilizing optically stimulated luminescence (OSL) dosimeters embedded midline into mouse carcasses and thermal luminescence dosimeters (TLD) embedded midline into acrylic phantoms.The effort demonstrated that the majority (4/7) of the laboratories was able to deliver sufficiently accurate exposures having maximum dosing errors of ≤5%. Comparable rates of 'dosimetric compliance' were noted between OSL- and TLD-based tests. Data analysis showed a highly linear relationship between 'measured' and 'target' doses, with errors falling largely between 0 and 20%. Outliers were most notable for OSL-based tests, while multiple tests by 'non-compliant' laboratories using orthovoltage X-rays contributed heavily to the wide variation in dosing errors.For the dosimetrically non-compliant laboratories, the relatively high rates of dosing errors were problematic, potentially compromising the quality of ongoing radiobiological research. This dosimetry effort proved to be instructive in establishing rigorous reviews of basic dosimetry protocols ensuring that dosing errors were minimized.

High-throughput screening of compounds (chemicals) is an essential part of drug discovery, involving thousands to millions of compounds, with the purpose of identifying candidate hits. Most statistical tools, including the industry standard B-score method, work on individual compound plates and do not exploit cross-plate correlation or statistical strength among plates. We present a new statistical framework for high-throughput screening of compounds based on Bayesian nonparametric modeling. The proposed approach is able to identify candidate hits from multiple plates simultaneously, sharing statistical strength among plates and providing more robust estimates of compound activity. It can flexibly accommodate arbitrary distributions of compound activities and is applicable to any plate geometry. The algorithm provides a principled statistical approach for hit identification and false discovery rate control. Experiments demonstrate significant improvements in hit identification sensitivity and specificity over the B-score and R-score methods, which are highly sensitive to threshold choice. These improvements are maintained at low hit rates. The framework is implemented as an efficient extension package and is suitable for large scale data sets.

Previous immunological studies in atomic bomb survivors have suggested that radiation exposure leads to long-lasting changes, similar to immunological aging observed in T-cell-adaptive immunity. However, to our knowledge, late effects of radiation on dendritic cells (DCs), the key coordinators for activation and differentiation of T cells, have not yet been investigated in humans. In the current study, we hypothesized that numerical and functional decreases would be observed in relationship to radiation dose in circulating conventional DCs (cDCs) and plasmacytoid DCs (pDCs) among 229 Japanese A-bomb survivors. Overall, the evidence did not support this hypothesis, with no overall changes in DCs or functional changes observed with radiation dose. Multivariable regression analysis for radiation dose, age and gender effects revealed that total DC counts as well as subpopulation counts decreased in relationship to increasing age. Further analyses revealed that in women, absolute numbers of pDCs showed significant decreases with radiation dose. A hierarchical clustering analysis of gene expression profiles in DCs after Toll-like receptor stimulation in vitro identified two clusters of participants that differed in age-associated expression levels of genes involved in antigen presentation and cytokine/chemokine production in cDCs. These results suggest that DC counts decrease and expression levels of gene clusters change with age. More than 60 years after radiation exposure, we also observed changes in pDC counts associated with radiation, but only among women.

Purpose: Formalin-fixed, paraffin-embedded tumor samples from CALGB 80203 were analyzed for expression of EGFR axis–related genes to identify prognostic or predictive biomarkers for cetuximab treatment. Patients and Methods: Patients (238 total) with first-line metastatic colorectal cancer (mCRC) were randomized to FOLFOX or FOLFIRI chemotherapy ± cetuximab. qRT-PCR analyses were conducted on tissues from 103 patients at baseline to measure gene expression levels of HER-related genes, including amphiregulin (AREG), betacellulin (BTC), NT5E (CD73), DUSP4, EGF, EGFR, epigen (EPGN), epiregulin (EREG), HBEGF, ERBB2 (HER2), ERBB3 (HER3), ERBB4 (HER4), PHLDA1, and TGFA. The interactions between expression levels and treatment with respect to progression-free survival (PFS) and overall survival (OS) were modeled using multiplicative Cox proportional hazards models. Results: High tumor mRNA levels of HER2 [hazard ratio (HR), 0.64; P = 0.002] and EREG (HR, 0.89; P = 0.016) were prognostic markers associated with longer PFS across all patients. HER3 and CD73 expression levels were identified as potential predictive markers of benefit from cetuximab. In KRAS wild-type (WT) tumors, low HER3 expression was associated with longer OS from cetuximab treatment, whereas high HER3 expression was associated with shorter OS from cetuximab treatment (chemo + cetuximab: HR, 1.15; chemo-only: HR, 0.48; Pinteraction = 0.029). High CD73 expression was associated with longer PFS from cetuximab treatment in patients with KRAS-WT (chemo + cetuximab: HR, 0.91; chemo-only: HR, 1.57; Pinteraction = 0.026) and KRAS-mutant (Mut) tumors (chemo + cetuximab: HR, 0.80; chemo-only: HR, 1.29; P = 0.025). Conclusions: Gene expression of HER3 and CD73 was identified as a potential predictive marker for cetuximab. These data implicate HER axis signaling and immune modulation as potential mechanisms of cetuximab action and sensitivity. Clin Cancer Res; 21(5); 1078–86. ©2014 AACR.

Purpose: We sought to show the relevance of a lymphoblastoid cell line (LCL) model in the discovery of clinically relevant genetic variants affecting chemotherapeutic response by comparing LCL genome-wide association study (GWAS) results to clinical GWAS results. Experimental Design: A GWAS of paclitaxel-induced cytotoxicity was conducted in 247 LCLs from the HapMap Project and compared with a GWAS of sensory peripheral neuropathy in patients with breast cancer (n = 855) treated with paclitaxel in the Cancer and Leukemia Group B (CALGB) 40101 trial. Significant enrichment was assessed by permutation resampling analysis. Results: We observed an enrichment of LCL cytotoxicity-associated single-nucleotide polymorphisms (SNP) in the sensory peripheral neuropathy-associated SNPs from the clinical trial with concordant allelic directions of effect (empirical P = 0.007). Of the 24 SNPs that overlap between the clinical trial (P < 0.05) and the preclinical cytotoxicity study (P < 0.001), 19 of them are expression quantitative trait loci (eQTL), which is a significant enrichment of this functional class (empirical P = 0.0447). One of these eQTLs is located in RFX2, which encodes a member of the DNA-binding regulatory factor X family. Decreased expression of this gene by siRNA resulted in increased sensitivity of Neuroscreen-1(NS-1; rat pheochromocytoma) cells to paclitaxel as measured by reduced neurite outgrowth and increased cytotoxicity, functionally validating the involvement of RFX2 in nerve cell response to paclitaxel. Conclusions: The enrichment results and functional example imply that cellular models of chemotherapeutic toxicity may capture components of the underlying polygenic architecture of related traits in patients. Clin Cancer Res; 19(2); 491–9. ©2012 AACR.

Purpose: Discovery of SNPs that predict a patient's risk of docetaxel-induced neuropathy would enable treatment individualization to maximize efficacy and avoid unnecessary toxicity. The objectives of this analysis were to discover SNPs associated with docetaxel-induced neuropathy and mechanistically validate these associations in preclinical models of drug-induced neuropathy. Experimental Design: A genome-wide association study was conducted in metastatic castrate-resistant prostate cancer patients treated with docetaxel, prednisone and randomized to bevacizumab or placebo on CALGB 90401. SNPs were genotyped on the Illumina HumanHap610-Quad platform followed by rigorous quality control. The inference was conducted on the cumulative dose at occurrence of grade 3+ sensory neuropathy using a cause-specific hazard model that accounted for early treatment discontinuation. Genes with SNPs significantly associated with neuropathy were knocked down in cellular and mouse models of drug-induced neuropathy. Results: A total of 498,081 SNPs were analyzed in 623 Caucasian patients, 50 (8%) of whom experienced grade 3+ neuropathy. The 1,000 SNPs most associated with neuropathy clustered in relevant pathways including neuropathic pain and axonal guidance. An SNP in VAC14 (rs875858) surpassed genome-wide significance (P = 2.12 × 10−8, adjusted P = 5.88 × 10−7). siRNA knockdown of VAC14 in stem cell–derived peripheral neuronal cells increased docetaxel sensitivity as measured by decreased neurite processes (P = 0.0015) and branches (P < 0.0001). Prior to docetaxel treatment, VAC14 heterozygous mice had greater nociceptive sensitivity than wild-type litter mate controls (P = 0.001). Conclusions: VAC14 should be prioritized for further validation of its potential role as a predictor of docetaxel-induced neuropathy and biomarker for treatment individualization. Clin Cancer Res; 22(19); 4890–900. ©2016 AACR.

Ionizing radiation (IR) is a major source of cellular damage and the immediate cellular response to IR has been well characterized. But the long-term impact of IR on cell function and its relationship with aging are not known. Here, we examined the IR effects on telomere length and other biomarkers 50 to 68 years post-exposure (two time points per person) in survivors of the atomic bombing at Hiroshima during WWII. We found that telomere length of leukocytes was inversely correlated with the dose of IR (p=0.008), and this effect was primarily found in survivors who were exposed at younger ages; specifically those

The Ducreyi serum resistance A (DsrA) protein of Haemophilus ducreyi belongs to a large family of multifunctional outer membrane proteins termed trimeric autotransporter adhesins responsible for resistance to the bactericidal activity of human complement (serum resistance), agglutination and adhesion. The ability of DsrA to confer serum resistance and bind extracellular matrix proteins lies in its N-terminal passenger domain. We have previously reported that immunization with a recombinant form of the passenger domain of DsrA, rNT-DsrA, in complete/incomplete Freund's adjuvant, protects against a homologous challenge in swine. We present herein the results of an immunogenicity study in mice aimed at investigating the persistence, type of immune response, and the effect of immunization route and adjuvants on surrogates of protection. Our results indicate that a 20 μg dose of rNT-DsrA administered with alum elicited antisera with comparable bacterial surface reactivity to that obtained with complete/incomplete Freund's adjuvant. At that dose, high titers and bacterial surface reactivity persisted for 211 days after the first immunization. Administration of rNT-DsrA with CpG or imiquimod as adjuvants elicited a humoral response with similar quantity and quality of antibodies (Abs) as seen with Freund's adjuvant. Furthermore, intramuscular administration of rNT-DsrA elicited high-titer Abs with significantly higher reactivity to the bacterial surface than those obtained with subcutaneous immunization. All rNT-DsrA/adjuvant combinations tested, save CpG, elicited a Th2-type response. Taken together, these findings show that a 20 μg dose of rNT-DsrA administered with the adjuvants alum, CpG or imiquimod elicits high-quality Abs with reactivity to the bacterial surface that could protect against an H. ducreyi infection.

Most FDA-approved adjuvants for infectious agents boost humoral but not cellular immunity, and have poorly-understood mechanisms. Stimulator of interferon genes (STING, also known as MITA, MPYS, or ERIS) is an exciting adjuvant target due to its role in cyclic dinucleotide (CDN)-driven anti-viral immunity; however, a major hindrance is STING's cytosolic localization which requires intracellular delivery of its agonists. As a result, STING agonists administered in a soluble form have elicited suboptimal immune responses. Delivery of STING agonists via particle platforms has proven a more successful strategy, but the opportunity for improved formulations and bioactivity remains. In this study we evaluated the adjuvant activity of the potent STING agonist, CDN 3′3′-cGAMP (cGAMP), encapsulated in acid-sensitive acetalated dextran (Ace-DEX) polymeric microparticles (MPs) which passively target antigen-presenting cells for intracellular release. This formulation was superior to all particle delivery systems evaluated and maintained its bioactivity following a sterilizing dose of gamma irradiation. Compared to soluble cGAMP, the Ace-DEX cGAMP MPs enhanced type-I interferon responses nearly 1000-fold in vitro and 50-fold in vivo , caused up to a 10 4 -fold boost in antibody titers, increased Th1-associated responses, and expanded germinal center B cells and memory T cells. Furthermore, the encapsulated cGAMP elicited no observable toxicity in animals and achieved protective immunity against a lethal influenza challenge seven months post-immunization when using CDN adjuvant doses up to 100-fold lower than previous reports. For these reasons, Ace-DEX MP-encapsulated cGAMP represents a potent vaccine adjuvant of humoral and cellular immunity.

Background Many analyses of microarray association studies involve permutation, bootstrap resampling and cross-validation, that are ideally formulated as embarrassingly parallel computing problems. Given that these analyses are computationally intensive, scalable approaches that can take advantage of multi-core processor systems need to be developed. Results We have developed a CUDA based implementation, , that employs graphics processing units in microarray association studies. We illustrate the performance and applicability of within the context of permutation resampling for a number of test statistics. An extensive simulation study demonstrates a dramatic increase in performance when using on an NVIDIA GTX 280 card compared to an optimized C/C++ solution running on a conventional Linux server. Conclusions is available as an open-source stand-alone application and as an extension package for the statistical environment. It provides a dramatic increase in performance for permutation resampling analysis in the context of microarray association studies. The current version offers six test statistics for carrying out permutation resampling analyses for binary, quantitative and censored time-to-event traits.

The kernel stick-breaking process (KSBP) is employed to segment general imagery, imposing the condition that patches (small blocks of pixels) that are spatially proximate are more likely to be associated with the same cluster (segment). The number of clusters is not set a priori and is inferred from the hierarchical Bayesian model. Further, KSBP is integrated with a shared Dirichlet process prior to simultaneously model multiple images, inferring their inter-relationships. This latter application may be useful for sorting and learning relationships between multiple images. The Bayesian inference algorithm is based on a hybrid of variational Bayesian analysis and local sampling. In addition to providing details on the model and associated inference framework, example results are presented for several image-analysis problems.

This paper presents a scheme for estimating two-band amplitude scale attack within a quantization-based watermarking context. Quantization-based watermarking schemes comprise a class of watermarking schemes that achieves the channel capacity in terms of additive noise attacks 1 . Unfortunately, Quantization-based watermarking schemes are not robust against Linear Time Invariant (LTI) filtering attacks. We concentrate on a multi-band amplitude scaling attack that modifies the spectrum of the signal using an analysis/synthesis filter bank. First we derive the probability density function (PDF) of the attacked data. Second, using a simplified approximation of the PDF model, we derive a Maximum Likelihood (ML) procedure for estimating two-band amplitude scaling factor. Finally, experiments are performed with synthetic and real audio signals showing the good performance of the proposed estimation technique under realistic

Quantization-based watermarking schemes comprise a class of watermarking schemes that achieves the channel capacity in terms of additive noise attacks. The existence of good high dimensional lattices that can be efficiently implemented and incorporated into watermarking structures, made quantization-based watermarking schemes of practical interest. Because of the structure of the lattices, watermarking schemes making use of them are vulnerable to non-additive operations, like amplitude scaling in combination with additive noise. In this paper, we propose a secure Maximum Likelihood (ML) estimation technique for amplitude scaling factors using subtractive dither. The dither has mainly security purposes and is assumed to be known to the watermark encoder and decoder. We derive the probability density function (PDF) models of the watermarked and attacked data in the presence of subtractive dither. The derivation of these models follows the lines of reference 5, where we derived the PDF models in the absence of dither. We derive conditions for the dither sequence statistics such that a given security level is achieved using the error probability of the watermarking system as objective function. Based on these conditions we are able to make approximations to the PDF models that are used in the ML estimation procedure. Finally, experiments are performed with real audio and speech signals showing the good performance of the proposed estimation technique under realistic conditions.

Quantization-based watermarking schemes comprise a class of watermarking schemes that achieves channel capacity in terms of additive noise attacks. Unfortunately, quantization-based watermarking schemes are not robust against linear time invariant (LTI) filtering attacks. In this paper we concentrate on amplitude scaling of a multi-band system. First we derive the probability density function (PDF) of the attacked data. Second, using a simplified approximation of the PDF model, we derive a maximum likelihood (ML) procedure for estimating two-band amplitude scaling factor. Finally, experiments are performed with real audio signals showing the good performance of the proposed estimation technique under realistic conditions.

11053 Background: There are currently no effective methods for predicting, preventing, or treating chemotherapy-induced peripheral neuropathy. We performed a genome-wide association study in a clinical trial of castration-resistant prostate cancer (CRPC) to discover variants that may be useful for identifying patients at high risk of neuropathy during docetaxel treatment. Methods: Treatment and toxicity data were collected prospectively on the Cancer and Leukemia Group B (CALGB) 90401 trial of chemotherapy naïve CRPC patients treated with docetaxel and prednisone ± bevacizumab. Genotyping was performed by the RIKEN Institute using the Illumina HumanHap610-Quad platform. Genetically defined European subjects were included in the discovery analysis of all single nucleotide polymorphisms (SNPs) that passed quality control. The primary endpoint was the cumulative dose level triggering a grade 3+ sensory neuropathy. The inference was conducted within the framework of a competing risk model accounting for early treatment termination induced by death or progression, or other toxicities. SNPs that were highly associated with neuropathy were assessed for a broader taxane effect in a cohort of paclitaxel-treated patients from a breast cancer clinical trial, CALGB 40101. Results: 623 Caucasian patients and 498,022 SNPs were included in the discovery analysis. The incidence of grade 3 neuropathy was 8%. One intergenic SNP (rs11017056) was associated with increased risk of neuropathy (HR=2.83, p=4.7x10-7). This association surpassed the genome-wide significance threshold after covariate adjustment (p=7.2x10-8). However, none of the 7 SNPs selected for replication were associated with neuropathy in the paclitaxel-treated breast cancer cohort. Conclusions: Using a prospectively enrolled prostate cancer patient cohort we identified multiple SNPs that may identify risk of docetaxel-induced peripheral neuropathy, but not paclitaxel-induced neuropathy. However, since it is unknown whether the genetic factors that affect taxane neuropathy are drug-specific, further replication studies in docetaxel-treated cohorts are of great interest.

11011 Background: Beyond KRAS status, there are no validated markers for anti-EGFR therapy in mCRC. While expression of genes within the EGF signaling axis has been reported to correlate with benefit, most reports used data from non-randomized trials which cannot distinguish between general prognostic markers and markers that predict for benefit from C. CALGB 80203 was a 238-patient (pt), 4-arm, randomized phase II study of FOLFOX or FOLFIRI +/- C in mCRC pts. Formalin-fixed, paraffin-embedded (FFPE) tumor samples from CALGB 80203 were analyzed for gene expression of HER family ligands, receptors and regulators. Methods: FFPE tumor samples from consenting pts were analyzed for AREG, BTC, CD73, DUSP4, EGF, EGFR, EPGN, EREG, HB-EGF, HER2, HER3, HER4, PHLDA1 and TGFa by RT-qPCR. Interaction between baseline gene expression levels and treatment (C) with respect to progression-free survival (PFS) and overall survival (OS) was modeled using a multiplicative Cox proportional hazards model. Results: Baseline tumor tissue was available from 103 pts; 55% of tumors were KRAS wild type (WT). Two prognostic markers were identified; higher tumor mRNA levels of HER2 (HR=0.64, p=0.002) and EREG (HR=0.89, p=0.016) were associated with longer PFS across all pts. Potential predictive markers of benefit for C were identified. In KRAS WT tumors, lower HER3 expression was associated with longer OS from C treatment while higher HER3 expression was associated with shorter OS from C treatment (chemo + C: HR=1.15; chemo only: HR=0.48, interaction p=0.029). No association was observed for HER3 and outcome in KRAS mutant (MT) tumors. Interestingly, higher CD73 expression was associated with longer PFS in C-treated pts in both KRASWT (chemo + C: HR=0.91; chemo only: HR=1.57, interaction p=0.026) and MT tumors (chemo + C: HR=0.80; chemo only: HR=1.29, p=0.025). Conclusions: In one of the first reports using data from a randomized study, gene expression of HER3 and CD73 were identified as potential predictive markers for C. These data implicate not only HER axis signaling but also immune modulation as potential mechanisms of C action and sensitivity, and warrant confirmation in other large randomized trials.

4520 Background: CALGB 90206 was a phase III trial of 732 pts with RCC comparing B+I versus I alone demonstrating no difference in OS. To date, there are no validated predictive biomarkers for B in RCC. For this reason, baseline plasma samples from CALGB 90206 pts were analyzed to identify and test predictive markers for B+I in RCC pts. Methods: Baseline EDTA plasma samples from 424 consenting pts were analyzed using an optimized multiplex ELISA platform for 32 candidate factors related to tumor growth, angiogenesis, and inflammation. The data were randomly split into training (n=286) and validation (n=138) sets. The proportional hazards model was used to test for treatment-marker interactions of OS. The estimated coefficients from the training set were used to compute a risk score (RS) for each pt in the validation set. The RS classified pts by risk in the validation set. The model was assessed for its predictive accuracy using area under the curve (AUC). Results: A statistically significant 3-way interaction between interleukin-6 (IL-6), hepatocyte growth factor (HGF) and treatment was observed in the training set (p