Reproducibility Challenges & Opportunities in Population Neuroimaging

For years brain imaging studies have been limited to 2-digit sample sizes, but projects like the UK Biobank and the Adolescent Brain Cognitive Development Study have made 5- & 6-digit sample sizes a reality. After providing some context for the UK Biobank core and imaging extension, I will review three areas of work motivated by large scale data and the ubiquitous challenges of reproducibility. First, even with 'just' 1000 subjects, p-values can lose their meaning when the entire brain can be declared significant. I will discuss methods focused on effect size, and in particular characterising the spatial uncertainty in clusters defined by a given effect size. Whether for percent BOLD change or Cohen's d, we have developed methods to produce spatial confidence sets, providing inner and outer confidence clusters. Next I will discuss two pieces of work on reproducibility, with a view to characterising the variation in results obtained from the same data different ways. Whether analysed by the same group with careful manipulation of three different software packages (AFNI, FSL, & SPM), or by 70 different research groups making their own analysis decisions, there is remarkable variation in the results obtained. I'll close with some discussions on how best to manage this variability.

Genomic Epidemiology in SARS-CoV-2: new tools and challenges

Scientists around the world have sequenced over 2 million SARS-CoV-2 genomes in an effort to monitor and understand the evolution and transmission of this virus. Virus sequences can help us understand the emergence of new variants with new phenotypes, track the virus' geographical movements and can help us to understand local transmission. However, there are mathematical and statistical challenges in making the most of this potentially rich source of information about viruses and how they spread. In this talk I will introduce the field of genomic epidemiology in general, and then describe recent research in our group. We have developed a method to use SARS-CoV-2 genomes to estimate serial intervals: the time between symptoms (or in some cases, sample collection) in infector-infectee pairs. Serial intervals are important because they underlie estimates of the reproductive number, Rt, which in turn is used to help understand the strength of transmission and the impact of different levels of vaccine coverage. I will describe the results of this method applied to data from Victoria, Australia. I will conclude by noting some broader challenges and opportunities for the genomic surveillance of SARS-CoV-2.

Hierarchical statistical models in wildlife ecology

Hierarchical statistical models (HSM) are multi-level models in which one level is conditional on another. These models are used in a myriad ways, for example, to account for nested sampling designs, to model processes on multiple scales (e.g., spatial or temporal), or to disentangle ecological from observation processes. The latter is particularly important in wildlife research, as animals are notoriously difficult to observe. As a result, ecological state variables of interest, such as species presence or abundance, are observed imperfectly, e.g., a species may be present but never detected by sampling; or a population may consist of more individuals than are counted. In my talk, I will provide an overview over HSM that address this form of observation bias, also called imperfect detection, by describing an observation model that is conditioned on the true, but latent, underlying ecological state. I will present occupancy models, in which repeated species-level detection/non-detection data collected across multiple sites (observation) are used to estimate species occurrence (ecological state) while accounting for imperfect species detection, and which are often fit to joint data from multiple species (i.e., community occupancy models). I will touch on count-based models to estimate abundance, with a focus on distance sampling, in which individual detection probability is described as a declining function of distance from the observer, allowing for estimation of abundance (ecological state) from counts (observations). Finally, I will discuss traditional and spatial capture-recapture, the gold standard of abundance estimation, based on repeated detection data of animals that can be identified individually, through natural or artificial marks. I illustrate all approaches with case studies that also showcase extensions to the basic models. Overall, HSM is a flexible framework that can be tailored to specific sampling circumstances to investigate spatio-temporal processes in wildlife ecology while addressing imperfect detection.

Integration of molecular measurements across omics technologies, time and space

Modern sequencing technologies and complex experimental designs generate vast amounts of molecular data that can identify core biological processes and shed light onto their molecular regulation. A key challenge to leverage this data is the integration of the molecular measurements across different technologies, biological layers and conditions. In this talk I will present computational strategies for the joint analysis of such multi-modal data sets, focusing on probabilistic factor models that can extract major sources of variation in an interpretable manner. To accommodate studies with temporal or spatial resolution, I will further discuss extensions of these concepts to identify temporal or spatial patterns of variation in the data and compare them across different conditions. Case studies on biomedical omics data sets will demonstrate how such methods can help us to highlight core biological processes in health and disease or along development and pinpoint underlying molecular drivers.

Stability of animal movement: mechanics and neural control

The question of how to stably guide and control mechanical systems is at the heart of understanding animal movement. But the apparent grace and stability with which we move hides the slowness of our neural systems. Neural feedback control involving the brain is almost as slow as bouncing signals to a controller in geostationary orbit and back. How does biological control manage such large sensory feedback delays? I will use two examples to show how the intrinsic mechanics of our body and its neural modulation through muscle contraction are central to stability. The first example is the control of finger stability during precisions grips, the style of gripping where we use the fingertips for the careful application of force. The second example is whole body stability when running on uneven terrain, which is motivated by questions of persistence hunting in the context of human evolution. I will end the talk with a possibly provocative suggestion that animals are capable of such robust motor behavior because of, and not despite, the slowness of neural feedback.