Illustration about Artificial Intelligence in healthcare

Erin Craig, 6th Year PhD Student
(12:15pm-12:45pm)
Title: MMIL: A novel algorithm for disease associated cell type discovery
Abstract: Single-cell datasets often lack individual cell labels, making it challenging to identify cells associated with disease. To address this, we introduce Mixture Modeling for Multiple Instance Learning (MMIL), an expectation maximization method that enables the training and calibration of cell-level classifiers using patient-level labels. Our approach can be used to train e.g. lasso logistic regression models, gradient boosted trees, and neural networks.

 

Matt Aguirre, 6th Year PhD Student
(12:45pm-1:15pm)
Title: How can gene expression heritability be so diffuse across the genome?
Abstract:  A major surprise from human genetic studies of expression quantitative trait loci (eQTLs) has been that cis-acting eQTLs located near a given gene typically explain a small fraction of the genetic variance in its expression. Although the cis-fraction of expression heritability has been repeatedly estimated to have a genome-wide median of around 20%, it is unclear how this property arises. Crucially, distal trans-eQTLs tend to have small effects, and the sample sizes of gene expression studies and increased multiple testing burden of genome-wide analysis both place limits on statistical power. Yet this also begs a paradoxical question: how can small trans-acting effects distributed throughout the genome end up explaining so much heritability?

 

Zoom Link: https://stanford.zoom.us/j/97747187137?pwd=QJqAPX1yRpGOwTTGPsIIwEYp7W2Aaz.1&from=addon
Password: 180750