Bret Hanlon

Gene Expression Profile Alone Is Inadequate In Predicting Complete Response In Multiple Myeloma

With advent of several treatment options in multiple myeloma (MM), a selection of effective regimen has become an important issue. Use of gene expression profile (GEP) is considered an important tool in predicting outcome; however, it is unclear whether such genomic analysis alone can adequately predict therapeutic response. We evaluated the ability of GEP to predict complete response (CR) in MM. GEP from pretreatment MM cells from 136 uniformly treated MM patients with response data on an IFM, France led study were analyzed. To evaluate variability in predictive power due to microarray platform or treatment types, additional data sets from three different studies (n=511) were analyzed using same methods. We used several machine learning methods to derive a prediction model using training and test subsets of the original four data sets. Among all methods employed for GEP-based CR predictive capability, we got accuracy range of 56–78% in test data sets and no significant difference with regard to GEP platforms, treatment regimens or in newly diagnosed or relapsed patients. Importantly, permuted P-value showed no statistically significant CR predictive information in GEP data. This analysis suggests that GEP-based signature has limited power to predict CR in MM, highlighting the need to develop comprehensive predictive model using integrated genomics approach.

Prognostic factors modifying the treatment-free interval in recurrent ovarian cancer

OBJECTIVES While primary treatment for high-grade serous ovarian cancer tends to be uniform - maximal debulking and platinum/taxane adjuvant chemotherapy - there is little standardization of treatment in the recurrent setting beyond the exhaustive use of platinum therapies. Using secondary data from multiple centers participating in the Cancer Genome Atlas study (TCGA), we seek to characterize clinical features, timing and serial response data to provide empirical evidence for treatment expectations in the recurrent setting. METHODS We conducted a retrospective survival analysis of TCGA study primary and secondary patient chemotherapy regimens by characterizing the dynamics of 1119 lines of therapy comprising the treatment of 461 high-grade serous ovarian cancer patients. All patients with post-surgical drug therapy information from the TCGA database were included in this study. RESULTS A complete response to adjuvant therapy led to longer overall survival, but did not affect treatment free intervals (TFIs) after relapse of disease. A strong predictor of the TFI on the next treatment regimen was the previous TFI with a decaying effect. The number of previous treatments, of platinum treatments, and the length of time from surgery all have an exponentially decreasing effect on TFI. Re-treatment times appear to cluster at predictable times following surgery. CONCLUSIONS While patients experience a consistent reduction in TFI with increasing re-treatment, the initial adjuvant interval is unrelated to later interval lengths. Platinum re-treatment remained an effective option in patients typically thought to be platinum resistant and the timing of monitoring visits may drive overall re-treatment patterns.

Inference for Quantitation Parameters in Polymerase Chain Reactions via Branching Processes With Random Effects

The quantitative polymerase chain reaction (qPCR) is a widely used tool for gene quantitation and has been applied extensively in several scientific areas. The current methods used for analyzing qPCR data fail to account for multiple sources of variability present in the PCR dynamics, leading to biased estimates and incorrect inference. In this article, we introduce a branching process model with random effects to account for within-reaction and between-reaction variability in PCR experiments. We describe, in terms of the observed fluorescence data, new statistical methodology for gene quantitation. Using simulations, PCR experiments, and asymptotic theory we demonstrate the improvements achieved by our methodology compared to existing methods. This article has supplemental materials online.

Discrete mixture modeling to address genetic heterogeneity in time-to-event regression

Motivation: Time-to-event regression models are a critical tool for associating survival time outcomes with molecular data. Despite mounting evidence that genetic subgroups of the same clinical disease exist, little attention has been given to exploring how this heterogeneity affects time-to-event model building and how to accommodate it. Methods able to diagnose and model heterogeneity should be valuable additions to the biomarker discovery toolset. Results: We propose a mixture of survival functions that classifies subjects with similar relationships to a time-to-event response. This model incorporates multivariate regression and model selection and can be fit with an expectation maximization algorithm, we call Cox-assisted clustering. We illustrate a likely manifestation of genetic heterogeneity and demonstrate how it may affect survival models with little warning. An application to gene expression in ovarian cancer DNA repair pathways illustrates how the model may be used to learn new genetic subsets for risk stratification. We explore the implications of this model for censored observations and the effect on genomic predictors and diagnostic analysis. Availability and implementation: R implementation of CAC using standard packages is available at https://gist.github.com/programeng/8620b85146b14b6edf8f Data used in the analysis are publicly available. Contact: kevin.eng@roswellpark.org Supplementary information: Supplementary data are available at Bioinformatics online.

Spatio-Temporal in vivo Recording of dCREB2 Dynamics in Drosophila Long-Term Memory Processing

CREB (cAMP response element-binding protein) is an evolutionarily conserved transcription factor, playing key roles in synaptic plasticity, intrinsic excitability and long-term memory (LTM) formation. The Drosophila homologue of mammalian CREB, dCREB2, is also important for LTM. However, the spatio-temporal nature of dCREB2 activity during memory consolidation is poorly understood. Using an in vivo reporter system, we examined dCREB2 activity continuously in specific brain regions during LTM processing. Two brain regions that have been shown to be important for Drosophila LTM are the ellipsoid body (EB) and the mushroom body (MB). We found that dCREB2 reporter activity is persistently elevated in EB R2/R4m neurons, but not neighboring R3/R4d neurons, following LTM-inducing training. In multiple subsets of MB neurons, dCREB2 reporter activity is suppressed immediately following LTM-specific training, and elevated during late windows. In addition, we observed heterogeneous responses across different subsets of neurons in MB αβ lobe during LTM processing. All of these changes suggest that dCREB2 functions in both the EB and MB for LTM formation, and that this activity contributes to the process of systems consolidation.

Chemodynamical Clustering Applied to APOGEE Data: Rediscovering Globular Clusters

We have developed a novel technique based on a clustering algorithm which searches for kinematically-and-chemically-clustered stars in the APOGEE DR12 Cannon data. As compared to classical chemical tagging, the kinematic information included in our methodology allows us to identify stars that are members of known globular clusters with greater confidence. Our methodology reduces the dimensionality of clustering in the chemical space according to anti-correlations found in the optical spectra between the elements Al and Mg, Na and O, and C and N in globular clusters. Our algorithm identifies globular clusters without a priori knowledge of their locations in the sky. Thus, not only does this technique promise to discover new globular clusters, but it also allows us to search for unusual stars in the halo to identify new streams of kinematically--and--chemically--clustered stars.

Penalized Likelihood Phenotyping: Unifying Voxelwise Analyses and Multi-Voxel Pattern Analyses in Neuroimaging Penalized Likelihood Phenotyping

Neuroimage phenotyping for psychiatric and neurological disorders is performed using voxelwise analyses also known as voxel based analyses or morphometry (VBM). A typical voxelwise analysis treats measurements at each voxel (e.g. fractional anisotropy, gray matter probability) as outcome measures to study the effects of possible explanatory variables (e.g. age, group) in a linear regression setting. Furthermore, each voxel is treated independently until the stage of correction for multiple comparisons. Recently, multi-voxel pattern analyses (MVPA), such as classification, have arisen as an alternative to VBM. The main advantage of MVPA over VBM is that the former employ multivariate methods which can account for interactions among voxels in identifying significant patterns. They also provide ways for computer-aided diagnosis and prognosis at individual subject level. However, compared to VBM, the results of MVPA are often more difficult to interpret and prone to arbitrary conclusions. In this paper, first we use penalized likelihood modeling to provide a unified framework for understanding both VBM and MVPA. We then utilize statistical learning theory to provide practical methods for interpreting the results of MVPA beyond commonly used performance metrics, such as leave-one-out-cross validation accuracy and area under the receiver operating characteristic (ROC) curve. Additionally, we demonstrate that there are challenges in MVPA when trying to obtain image phenotyping information in the form of statistical parametric maps (SPMs), which are commonly obtained from VBM, and provide a bootstrap strategy as a potential solution for generating SPMs using MVPA. This technique also allows us to maximize the use of available training data. We illustrate the empirical performance of the proposed framework using two different neuroimaging studies that pose different levels of challenge for classification using MVPA.

Genetic inhibition of autophagy in a transgenic mouse model of anal cancer

Background: The dynamic role of autophagy in cancer development is a topic of considerable research and debate. Previously published studies have shown that anal cancer development can be promoted or prevented with the pharmacologic inhibition or induction, respectively, of autophagy in a human papillomavirus (HPV) mouse model. However, these results are confounded by the fact that the drugs utilized are known to affect other pathways besides autophagy. It has also been shown that autophagic inhibition occurs in the setting of HPV16 oncoprotein expression (E6 and E7) and correlates with increased susceptibility to anal carcinogenesis. Materials and Methods: In this study, we employed a conditional, genetic, autophagic (Atg7) knockout mouse model to determine conclusively that autophagy has a role in anal cancer development, in the absence or presence of E6 and E7. Results: In mice lacking both HPV16 oncogenes, knockout of autophagy followed by exposure to a carcinogen resulted in a tumor incidence of 40%, compared to 0% in mice treated with a carcinogen alone with an intact autophagic pathway (P = 0.007). In mice expressing either one or both HPV16 oncoproteins, the addition of genetic knockout of autophagy to carcinogen treatment did not lead to a significant difference in tumor incidence compared to carcinogen treatment alone, consistent with the ability of HPV oncogenes to inhibit autophagy in themselves. Conclusions: These results provide the first conclusive evidence for the distinct role of autophagy in anal carcinogenesis, and suggest that autophagy is a plausible target for therapies aimed at reducing anal dysplasia and anal cancer development.

Watching the clock for 25 years in FlyClockbase: Variability in circadian clocks of Drosophila melanogaster as uncovered by biological model curation

High-quality model curation provides insights by organizing biological knowledge-fragments. We aim to integrate published results about circadian clocks in Drosophila melanogaster while exploring economies of scale in model curation. Clocks govern rhythms of gene-expression that impact fitness, health, cancer, memory, mental functions, and more. Human clock insights have been pioneered in flies. Flies simplify investigating complex gene regulatory networks, which express proteins cyclically using environmentally entrained interlocking feedback loops. Simulations could simplify research further, but currently few models test their quality directly against experimentally observed time series scattered across publications. We designed FlyClockbase for robust efficient access to such scattered data for biologists and modelers, prioritizing simplicity and openness to encourage experimentalists to preserve more annotations and raw-data. Such details could multiply long-term value for modelers interested in meta-analyses, parameter estimates, and hypothesis testing. Currently FlyClockbase contains over 400 wildtype time series of core circadian components systematically curated from 86 studies published between 1990 and 2015. Using FlyClockbase, we show that PERIOD protein amount peak time variance unexpectedly exceeds that of TIMELESS. We hypothesize, PERIOD’s exceedingly more complex phosphorylation rules are responsible. Human error analysis improved data quality and revealed significance-degrading outliers, possibly violating presumed absence of wildtype heterogeneity or lab evolution. We found PCR-measured peak time variances exceed those from other methods, pointing to initial count stochasticity. Our trans-disciplinary analyses demonstrate how compilers with more biology-friendly logic could simplify, guide, and naturally distribute biological model curation. Resulting quality increases and cost reductions benefit curation-dependent grand challenges like personalizing medicine. General Article Summary Circadian clocks impact health and fitness by controlling daily rhythms of gene-expression through complex gene-regulatory networks. Deciphering how they work requires experimentally tracking changes in amounts of clock components. We designed FlyClockbase to simplify data-access for biologists and modelers and curated over 400 time series observed in wildtype fruit flies from 25 years of research on clocks. We found differences in peak time variance of the clock-proteins ‘PERIOD’ and ‘TIMELESS’, which probably stem from differences in phosphorylation-network complexity. Combining in-depth circadian-biology, model-curation, and compiler logic, our trans-disciplinary research shows how biology-friendly compilers could simplify model curation enough to democratize it. Statement of data availability and stabilizing versioning number: QQv1r1 For review purposes: QQv1r1 zip-archive in Supplemental Material or upon request. Before final publication: FlyClockbase will be at https://github.com/FlyClockbase Abbreviations Table 1 Core clock components Table 2 Concepts in FlyClockbase

Estimation and testing problems in auditory neuroscience via clustering.

The processing of auditory information in neurons is an important area in neuroscience. We consider statistical analysis for an electrophysiological experiment related to this area. The recorded synaptic current responses from the experiment are observed as clusters, where the number of clusters is related to an important characteristic of the auditory system. This number is difficult to estimate visually because the clusters are blurred by biological variability. Using singular value decomposition and a Gaussian mixture model, we develop an estimator for the number of clusters. Additionally, we provide a method for hypothesis testing and sample size determination in the two-sample problem. We illustrate our approach with both simulated and experimental data.