21 research outputs found
Cancer Genomics Dataset
No full textWe provide a pre-processed version of data provided by the curatedBreastData R package (Planey (2020) that was used in the paper "Multi-Task Learning for Sparsity Pattern Heterogeneity: A Discrete Optimization Approach," Loewinger et al., 2022. "breastCancer_data.zip" includes a reduced file with a subset of the covariates after an initial screening.
PLANEY, K. (2020). curatedBreastData: Curated breast cancer gene expression data with survival and treatment information R package version 2.18.0
Cancer Genomics Dataset
No full textWe provide a pre-processed version of data provided by the curatedBreastData R package (Planey (2020) that was used in the paper "Multi-Task Learning for Sparsity Pattern Heterogeneity: A Discrete Optimization Approach," Loewinger et al., 2022. "breastCancer_data.zip" includes a reduced file with a subset of the covariates after an initial screening.
PLANEY, K. (2020). curatedBreastData: Curated breast cancer gene expression data with survival and treatment information R package version 2.18.0
Recommended from our members
Statistical Learning Methods for Multi-Dataset Prediction
No full textIt has become increasingly common in the biomedical sciences to encounter settings where multiple datasets are available to train statistical learning models. These opportunities arise when fitting prediction models using datasets from, for example, repositories that aggregate studies from different labs or study populations. By training models on datasets that combine multiple sources or studies, it is tempting to assume the resulting prediction algorithm will be more robust to the problem of dataset shift, in which discrepancies in the distribution of training and test data can reduce out-of-sample prediction performance. However, common approaches such as pooling datasets before model fitting can perform poorly when datasets are highly heterogeneous. As such, the development of statistical methods that can explicitly account for heterogeneity across data sources is critical to training models that are replicable across populations. Here we propose statistical learning methods that leverage multiple datasets during model training to improve prediction performance.
In chapter 1, we introduce methods for domain generalization, in which we train a model on each of several datasets and create an aggregate ensemble prediction rule that is constructed to predict well on an unseen dataset, or “domain.” Specifically, we propose the “study strap ensemble,” which generalizes bagging for multi-dataset settings, using a hierarchical resampling procedure. By pairing the method with covariate similarity-based ensemble weighting schemes, we extend the method to multi-source domain adaptation problems, in which a sample of observations of the covariates from the target population is available at the time of model training. We prove existing domain generalization ensembling strategies, as well as standard bagging procedures, are special cases of the study strap ensemble. We demonstrate the effectiveness of our method in a human neuroscience application and in simulations.
In chapter 2, we propose methods for multi-source transfer learning, a setting that arises when an analyst has limited data collected from a distribution of interest, and they wish to leverage multiple auxillary training datasets to improve prediction performance on new observations from the target distribution. We build on “multi-study ensembling,” a multi-dataset procedure that uses a two-stage “stacking” strategy that first fits dataset-specific models and then aggregates ensemble models through a weighted average. Stacking estimates ensemble weights and model parameter weights separately, however, and therefore ignores the ensemble properties at the model-fitting stage, potentially resulting in a loss of efficiency. We therefore propose “optimal ensemble construction,” an “all-in-one” approach to multi-study stacking whereby we jointly estimate ensemble weights as well as parameters associated with each dataset-specific model via a unified optimization formulation. We establish that limiting cases of our approach yield existing methods such as multi-study stacking and pooling datasets before model fitting. We compare our approach to standard methods by applying it to a multi-country COVID-19 dataset for baseline mortality prediction.
In chapter 3, we propose a multi-task learning method, in which we jointly train a collection of sparse linear models, each fit on a separate dataset, to improve performance of each model on its respective domain or “task.” Specifically, we propose methods to extend the best subset selection problem, by placing a separate sparsity constraint on the regression parameters from each task, allowing the supports of the regression coefficients to differ across tasks. We propose a “support heterogeneity regularization” penalty that shrinks together the supports of the model coefficients across tasks, thereby encouraging models to share information during variable selection. We propose approaches based on first-order optimization and local combinatorial search in order to scale the method to high dimensional settings. We showcase the effectiveness of our method on neuroscience and cancer genomics applications
A statistical framework for analysis of trial-level temporal dynamics in fiber photometry experiments
No full textFiber photometry has become a popular technique to measure neural activity in vivo, but common analysis strategies can reduce the detection of effects because they condense within-trial signals into summary measures, and discard trial-level information by averaging across-trials. We propose a novel photometry statistical framework based on functional linear mixed modeling, which enables hypothesis testing of variable effects at every trial time-point, and uses trial-level signals without averaging. This makes it possible to compare the timing and magnitude of signals across conditions while accounting for between-animal differences. Our framework produces a series of plots that illustrate covariate effect estimates and statistical significance at each trial time-point. By exploiting signal autocorrelation, our methodology yields joint 95% confidence intervals that account for inspecting effects across the entire trial and improve the detection of event-related signal changes over common multiple comparisons correction strategies. We reanalyze data from a recent study proposing a theory for the role of mesolimbic dopamine in reward learning, and show the capability of our framework to reveal significant effects obscured by standard analysis approaches. For example, our method identifies two dopamine components with distinct temporal dynamics in response to reward delivery. In simulation experiments, our methodology yields improved statistical power over common analysis approaches. Finally, we provide an open-source package and analysis guide for applying our framework
Phasic Mesolimbic Dopamine Release Tracks Reward Seeking During Expression of Pavlovian-to-Instrumental Transfer
No full textBackgroundRecent theories addressing mesolimbic dopamine's role in reward processing emphasize two apparently distinct functions, one in reinforcement learning (i.e., prediction error) and another in incentive motivation (i.e., the invigoration of reward seeking elicited by reward-paired cues). Here, we evaluate the latter.MethodsUsing fast-scan cyclic voltammetry, we monitored, in real time, dopamine release in the nucleus accumbens core of rats (n = 9) during a Pavlovian-to-instrumental transfer task in which the effects of a reward-predictive cue on an independently trained instrumental action were assessed. Voltammetric data were parsed into slow and phasic components to determine whether these forms of dopamine signaling were differentially related to task performance.ResultsWe found that a reward-paired cue, which increased reward-seeking actions, induced an increase in phasic mesolimbic dopamine release and produced slower elevations in extracellular dopamine. Interestingly, phasic dopamine release was temporally related to and positively correlated with lever-press activity generally, while slow dopamine changes were not significantly related to such activity. Importantly, the propensity of the reward-paired cue to increase lever pressing was predicted by the amplitude of phasic dopamine release events, indicating a possible mechanism through which cues initiate reward-seeking actions.ConclusionsThese data suggest that those phasic mesolimbic dopamine release events thought to signal reward prediction error may also be related to the incentive motivational impact of reward-paired cues on reward-seeking actions
Predictability of phenotype information from functional connectivity in large imaging datasets - OHBM 2024
No full textSubmission to OHBM 202
Operant self-stimulation of thalamic terminals in the dorsomedial striatum is constrained by metabotropic glutamate receptor 2
No full text
