Bayesian Boolean Matrix Factorisation

Tumor-associated macrophages (TAMs) constitute up to 50% of tumor bulk in glioblastoma (GBM) and play an important role in tumor maintenance and progression. The recently discovered differences between invading tumour periphery and hypoxic tumor core implies that macrophage biology is also distinct by location. This may provide further insight into the observed treatment resistance to immune modulation. We hypothesize that macrophage activation occurs through processes that are distinct in tumor periphery versus core. We therefore investigated regional differences in TAM recruitment and evolution in GBM by combining open source single cell and bulk gene expression data. We used single cell gene expression data from 4 glioblastomas (total of 3589 cells) and 122 total bulk samples obtained from 10 different patients. Cell identity, ontogeny (bone-marrow derived macrophages-BMDM vs microglia), and macrophage activation state were inferred using verified gene expression signatures. We captured the spectrum of immune states using cell trajectory analysis with pseudotime ordering. In keeping with previous studies, TAMs carrying BMDM identity were more abundant in tumor bulk while microglia-derived TAMs dominated the tumor periphery across all macrophage activation states including pre-activation. We note that core TAMs evolve towards a pro-inflammatory state and identify a subpopulation of cells based on a gene program exhibiting strong, opposing correlation with Programmed cell Death-1 (PD-1) signaling, which may correlate to their response to PD-1 inhibition. By contrast, peripheral TAMs evolve towards anti-inflammatory phenotype and contains a population of cells strongly associated with NFkB signaling. Our preliminary analysis suggests important regional differences in TAMs with regard to recruitment and evolution. We identify regionally distinct and potentially actionable cell subpopulations and advocate the need for a multi-targeted approach to GBM therapeutics.

Single cell transcriptomics measures gene expression data of large number of genes, concurrently, from tens of thousands of cells present in a studied biological sample. It is difficult to obtain good classification results due to high data dimensionality and variability of biological states. We performed a preliminary study to assess the feasibility of using supervised machine learning methods to classify peripheral blood mononuclear cell (PBMC) types from single cell gene expression data. We analyzed a large PBMC data set $(\sim 120,000$ PBMC cells), selected 47 genes (from 30698 features) suitable as SML classification features, and performed classification using 20 machine learning algorithms. Data sets represented three sample processing strategies: PBMC separation (two data sets), and experimental cell sorting by (two data sets). The accuracy in 5-class classification among 20 methods was 91-97% (PBMC separation), 97-100% (magnetic-activated cell sorting), and 82-99% (fluorescence-activated cell sorting). Our results indicate the feasibility of supervised machine learning for classification of cells into major PBMC cell types using a small number of classification features from single cell gene expression data.

Nonnegative matrix factorization (NMF) has been widely exploited in many computational intelligence and pattern recognition problems. In particular, it can be used to extract latent features from data. However, previous NMF models often assume a fixed number of features, which are normally tuned and searched using a trial and error approach. Learning binary features is also difficult, since the binary matrix posits a more challenging optimization problem. In this paper, we propose a new Bayesian model, termed the infinite nonnegative binary matrix trifactorization (iNBMT) model. This can automatically learn both latent binary features and feature numbers, based on the Indian buffet process (IBP). It exploits a trifactorization process that decomposes the nonnegative matrix into a product of three components: two binary matrices and a nonnegative real matrix. In contrast to traditional bifactorization, trifactorization can better reveal latent structures among samples and features. Specifically, an IBP prior is imposed on two infinite binary matrices, while a truncated Gaussian distribution is assumed on the weight matrix. To optimize the model, we develop a modified variational-Bayesian algorithm, with iteration complexity one order lower than the recently proposed maximization-expectation-IBP model [1] and the correlated IBP-IBP model [2] . A series of simulation experiments are carried out, both qualitatively and quantitatively, using benchmark feature extraction, reconstruction, and clustering tasks. Comparative results show that our proposed iNBMT model significantly outperforms state-of-the-art algorithms on a range of synthetic and real-world data. The new Bayesian model can thus serve as a benchmark technique for the computational intelligence research community.

One of the major impediments in deployment of Autonomous Driving Systems (ADS) is their safety and reliability. The primary reason for the complexity of testing ADS is that it operates in an open world characterized by its non-deterministic, high-dimensional and non-stationary nature where the actions of other actors in the environment are uncontrollable from the ADS's perspective. This leads to a state space explosion problem and one way of mitigating this problem is by concretizing the scope for the system under test (SUT) by testing for a set of behavioral competencies which an ADS must demonstrate. A popular approach to testing ADS is scenario-based testing where the ADS is presented with driving scenarios from real world (and synthetically generated) data and expected to meet defined safety criteria while navigating through the scenario. We present SAFR-AV, an end-to-end ADS testing platform to enable scenario-based ADS testing. Our work addresses key real-world challenges of building an efficient large scale data ingestion pipeline and search capability to identify scenarios of interest from real world data, creating digital twins of the real-world scenarios to enable Software-in-the-Loop (SIL) testing in ADS simulators and, identifying key scenario parameter distributions to enable optimization of scenario coverage. These along with other modules of SAFR-AV would allow the platform to provide ADS pre-certifications.

A Personalized Recommender System from Probabilistic Relational Model and Users’ Preferences

Endocytic alterations are one of the earliest changes to occur in Alzheimer's disease (AD), and are hypothesized to be involved in the selective vulnerability of specific neuronal populations during the progression of AD. Previous microarray and real-time quantitative PCR experiments revealed an upregulation of the early endosomal effector rab5 and the late endosome constituent rab7 in the hippocampus of people with mild cognitive impairment (MCI) and AD. To assess whether these select rab GTPase gene expression changes are reflected in protein levels within selectively vulnerable brain regions (basal forebrain, frontal cortex, and hippocampus) and relatively spared areas (cerebellum and striatum), we performed immunoblot analysis using antibodies directed against rab5 and rab7 on postmortem human brain tissue harvested from cases with a premortem clinical diagnosis of no cognitive impairment (NCI), MCI, and AD. Results indicate selective upregulation of both rab5 and rab7 levels within basal forebrain, frontal cortex, and hippocampus in MCI and AD, which also correlated with Braak staging. In contrast, no differences in protein levels were found in the less vulnerable cerebellum and striatum. These regional immunoblot assays are consistent with single cell gene expression data, and provide protein-based evidence for endosomal markers contributing to the vulnerability of cell types within selective brain regions during the progression of AD.

Most tumors are hierarchically organized and driven by a population of cells that display stem cell properties. These cancer stem cells (CSCs) have been characterized by in vitro assays such as sphere formation and in vivo by tumor initiation. In addition cell surface markers have been utilized to enrich for CSC populations. More recently a subtype of CSCs have been reported to display autofluorescence (AF), a property mediated by cellular uptake and concentration of riboflavin. In order to characterize the molecular heterogeneity of CSCs, we studied AF positive sorted single cells in different subtypes of breast cancer cell lines and also patient derived xenograft (PDX) breast tumor cells. Exposure of cell lines and PDXs to riboflavin significantly increased the percentage of AF+ cells 3-5 fold. These sorted AF+ cells were enriched for tumorsphere forming capacity in vitro. Flow cytometry analysis revealed partial overlap of AF+ cells with the previously characterized EMT-CSC phenotype CD44+/CD24- cells. We also utilized the microfluidic C1 and BioMark HD instruments to determine the expression patterns of 96 target genes using TaqMan assays in a multiplex RT-qPCR setting at single cell resolution. Single cell gene expression data showed a distinct signature in AF+ cells with high expression levels of MKI67, PCNA, ABCG2, BRCA1, Jag2, EZH2, and MMP9 genes. Since AF is mediated by specific cellular uptake and concentration of riboflavin, and the report that riboflavin is capable of generating cytotoxic free radicals upon light exposure, we determined whether this property could be exploited to selectively target CSCs. The AF+ cells were cultured in the presence and absence of riboflavin, exposed to visible light and analyzed by MTT cytotoxicity assay. Data revealed significant cytotoxicity of riboflavin on AF+ cells that were exposed to visible light. In conclusion, AF+ cells show distinct cancer stem cell features and are sensitive to light-activated riboflavin. These findings will help better understanding of tumor biology to identify new target therapies for treatment of cancer patients. Citation Format: Shamileh Fouladdel, Tahra Luther, Kaitlin Harvey, Rachel Martin-Trevino, Jill Granger, Ebrahim Azizi, Max S. Wicha. Riboflavin targets autofluorescent-positive breast cancer stem cells upon light exposure. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 3310.

Quantitative High-Resolution Cellular Map of the Organ of Corti.

Clustering high dimensional data is an emerging research field.Subspace clusteringorprojected clusteringgroup similar objects in subspaces, i.e. projections, of the full space. In the past decade, several clustering paradigms have been developed in parallel, without thorough evaluation and comparison between these paradigms on a common basis.Conclusive evaluation and comparison is challenged by three major issues. First, there is no ground truth that describes the "true" clusters in real world data. Second, a large variety of evaluation measures have been used that reflect different aspects of the clustering result. Finally, in typical publications authors have limited their analysis to their favored paradigm only, while paying other paradigms little or no attention.In this paper, we take a systematic approach to evaluate the major paradigms in a common framework. We study representative clustering algorithms to characterize the different aspects of each paradigm and give a detailed comparison of their properties. We provide a benchmark set of results on a large variety of real world and synthetic data sets. Using different evaluation measures, we broaden the scope of the experimental analysis and create a common baseline for future developments and comparable evaluations in the field. For repeatability, all implementations, data sets and evaluation measures are available on our website.

Autonomous vehicles have the potential to completely upend the way we transport today, however deploying them safely at scale is not an easy task. Any autonomous driving system relies on multiple layers of software to function safely. Among these layers, the Perception layer is the most data intensive and also the most complex layer to get right. Companies need to collect and annotate lots of data to properly train deep learning perception models. Simulation systems have come up as an alternative to the expensive task of data collection and annotation. However, whether simulated data can be used as a proxy for real-world data is an ongoing debate. In this work, we attempt to address the question of whether models trained on simulated data can generalize well to the real-world. We collect datasets based on two different simulators with varying levels of graphics fidelity and use the KITTI dataset as an example of real- world data. We train three separate deep learning based object detection models on each of these datasets, and compare their performance on test sets collected from the same sources. We also add the recently released Waymo Open Dataset as a challenging test set. Performance is evaluated based on the mean average precision (mAP) metric for object detection. We find that training on simulation in general does not translate to generalizability on real-world data and that diversity in the training set is much more important than visual graphics’ fidelity.

PCP65 - KEY CONSIDERATIONS FOR THE COLLECTION OF PATIENT REPORTED OUTCOME (PRO) DATA IN REAL WORLD (RW) STUDIES

We appreciate Hennessy and Leonard’s [1] comments on our paper and their strong support for the need to carefully characterize the performance of epidemiologic methods and analysis choices (which we collectively refer to as analyses). The work performed as part of the Observational Medical Outcomes Partnership (OMOP) is but a first step in this journey. We particularly appreciate the authors making the point that ‘‘Problematically implemented studies do not invalidate the underlying research designs, just those implementations’’. We fully agree with this assertion and the importance of beginning to systematically answer questions about what makes analyses problematic. We also share their belief that the empirical assessment of the performance of analyses applied to observational datasets is an essential prerequisite for understanding the reliability of any evidence developed from observational studies. Every measurement approach is limited in precision and accuracy, and the OMOP investigators have consistently acknowledged that the performance evaluation framework we used has limitations [2]. However, without any measure of performance, we are limited to making assumptions based on our theories. Science advances primarily by testing such theories and improving them based on empiric evidence. Ideally, we would measure the performance of analyses with a reference set of positive controls with known effect sizes and negative controls without an effect. Unfortunately, such a collection does not exist so we developed a reference set based on a systematic evaluation of multiple knowledge sources (including literature, product labeling, and a systematic review) that provide insight into others’ assessment of a causal relationship between selected exposures and outcomes [3]. The evidence supporting these controls falls short of the desired, but inherently unknowable, gold standard, and is further limited by the fact that most positive controls may have been known to clinicians before our data were generated [4]. It is important to note that we relied on positive controls only for measures of discrimination, while estimates of error and calibration are based only on negative controls which suffer less from these shortcomings. Simulated data, whether fully synthetic or created by injecting signal into real data, is another approach to creating a ‘gold standard’ [5]. It suffers from its own limitations, including the validity of the assumptions and the degree to which it reflects the relevant complexities of the real world. Experiments based on either simulated or realworld data alone are unlikely to be sufficient given their limitations, and offer complementary data on empirical performance. We measured the performance of analyses using fully synthetic data and obtained essentially the same results as when using the real data [6]. Despite these limitations, it is clear to us that empirically measuring performance of analyses for a particular question on a specific dataset (not generic performance as a one-size-fits-all solution) and using the empirical operating characteristics to calibrate the result is essential. Expert opinion and subjective arguments about theoretical beliefs around potential bias are not a sound foundation on which to develop evidence that is intended to be used to inform medical decisions, whether at the population or individual patient level. As is the case for controls, there is no gold standard for the definition of health outcomes (HOIs). J. M. Overhage (&) Siemens Medical Solutions, Malvern, PA, USA e-mail: marc.overhage@siemens.com

Matrix completion by deep matrix factorization

The advanced capacities of high throughput single-cell technologies have facilitated a great understanding of cell behaviors during the process of its growth and expansion. Yet, using these data to establish causal relationships between genes and transcription factors at single cells resolved level remain a major challenge. Here we present PySCNet ( https://github.com/MingBit/PySCNet ), a python toolkit that enables reconstructing and analyzing gene regulatory network (GRN) from single-cell gene expression data. We applied this tool to the simulated data set and found that the power to predict high confident networks increases in the Automated Machine Learning (Auto-ML) and random forest classification methods. As this project is still in-progress, more GRN methods and network-based algorithms will be updated in the near future.

MP20-02 Aquablation at 4-years: the largest, real world, single-center study and longest follow-up data for this innovative BPH treatment