[03056] Data-driven methods to discover within-host biological dynamics
Session Time & Room : 5D (Aug.25, 15:30-17:10) @E804
Type : Proposal of Minisymposium
Abstract : Our minisymposium brings together researchers whose goal is to develop and apply algorithms towards discovering and documenting within-host dynamics of the immune system and infectious disease. One of the common challenges of this field is the degree to which existing databases of prior knowledge should be incorporated to enhance the analysis of new bioinformatic data sets. Presenters will discuss their varied research projects in this space. Throughout the minisymposium, we aim to foster a dialogue among the presenters, as well as minisymposium participants, from which the nuances of this field can be discussed.
[03279] How to find a pertinent research question: the identification and exploration of known unknowns
Format : Online Talk on Zoom
Author(s) :
Mayla R Boguslav (Colorado State University)
Nourah M Salem (University of Colorado Anschutz Medical Campus)
Elizabeth White (University of Colorado Anschutz Medical Campus)
Katherine J Sullivan (University of Colorado Anschutz Medical Campus)
Michael Bada (University of Colorado Anschutz Medical Campus)
Teri L Hernandez (University of Colorado Anschutz Medical Campus)
Sonia M Leach (National Jewish Health)
Lawrence E Hunter (University of Colorado Anschutz Medical Campus)
Michael Kirby (Colorado State University)
Abstract : Scientific discovery progresses by exploring new and uncharted territory. More specifically, it advances by a process of transforming unknown unknowns first into known unknowns, and then into knowns. Over the last few decades, researchers have developed many knowledge bases to capture and connect the knowns, which has enabled topic exploration and contextualization of experimental results. But recognizing the unknowns is also critical for finding the most pertinent questions and their answers. Little work has focused on how scientists might use them to trace a given topic or experimental result in search of open questions and new avenues for exploration. We present methods and tools to help researchers automatically uncover these unknowns through the illumination of specific goals for scientific knowledge.
[05174] A summary of algorithms for sparse feature selection for Biological Data
Format : Online Talk on Zoom
Author(s) :
Michael Kirby (Colorado State University)
Abstract : Biological data sets such as transcriptomics, proteomics or metabolomics are characterised by their high dimension n and small sample size p. Realistically, it is a daunting challenge to deduce meaningful biological mechanisms when p<
[03409] Improving decoy detection for protein-protein interaction models
Format : Online Talk on Zoom
Author(s) :
Corey OHern (Yale University)
Abstract : Computational prediction and design of proteins is a difficult task that results in models with a wide variation in quality. Decoy detection algorithms seek to classify computational models as high-quality or low-quality without knowledge of the experimental structures. Recently, dramatic improvements have been made in decoy detection of models for single proteins, but decoy detection of protein-protein interface (PPI) models is still challenging. To assess the current state-of-the-art for PPI decoy detection, we scored computational models generated using rigid-body docking software, ZDOCK, from a dataset of 33 heterodimeric high-resolution x-ray crystal structures against a standard measure of similarity to the x-ray crystal structures. We found that for some targets there is a strong correlation between the docking and ground truth scores (i.e. easy targets), whereas for other targets there are only weak correlations (i.e. difficult targets). We show that a metric that characterizes the “flatness” of the target interfaces can distinguish easy from difficult targets, where flat targets possess only weak correlations between the docking and ground truth scores. In addition, most rigid docking software methods generate highly imbalanced datasets containing mostly low-quality computational models. Balanced datasets of PPI models reduce sampling bias, which makes it easier to identify the physical features that can classify PPI computational models.
[05206] Early Detection of Disease: An intersection between artificial intelligence and biomathematics
Format : Online Talk on Zoom
Author(s) :
Juan B Gutierrez (University of Texas at San Antonio)
Abstract : Many states of disease are progressive, having a silent phase in which pathogenesis advances without the manifestation of any symptoms. However, small perturbations can be detected by a complex signal such as electrocardiography. In this talk, I will present how to approach the detection of subtle physiological signals through the systematic design of the architecture of an artificial neural network capable of detecting liver-stage malaria with an accuracy of +90%.
