Cory Y. McLean

This paper introduces an EHR foundation model integrating Polygenic Risk Scores to improve disease prediction and personalized healthcare using All of Us data.

This study compares Element Avidity sequencing to Illumina sequencing, finding higher accuracy in mapping and variant calling, especially at lower coverages and in challenging genomic regions.

GrAdaBeam combines gradient guidance with evolutionary search to improve nucleic acid design, outperforming existing methods across diverse genomic tasks.

This paper introduces a deep learning framework, REGLE, that improves genomic discovery and risk prediction for respiratory and circulatory diseases by learning low-dimensional embeddings from high-dimensional clinical data.

This paper introduces M-REGLE, a multimodal deep learning method that improves genetic analysis of cardiovascular traits by integrating multiple clinical data modalities and outperforms unimodal approaches in locus discovery and risk prediction.

This paper develops a retinal aging clock from fundus images to predict chronological age and identifies genetic variants associated with accelerated aging.

This study evaluates ML-based phenotyping's robustness to label noise using VCDR data, showing improved genetic discovery and PRS predictive power with denoising.

This paper trains a deep neural network on noisy COPD labels to predict COPD status from spirograms, improving genetic discovery by identifying new loci and replicating existing ones.

DeepConsensus improves PacBio HiFi read accuracy and assembly quality using a gap-aware transformer model.

This study identifies genetic modifiers of LRRK2 mutations in Parkinson's disease, finding a significant variant in CORO1C and a polygenic risk score effect on penetrance.

DeepTrio improves variant calling in family trios using deep learning, achieving higher accuracy than DeepVariant, especially at lower coverages.

The paper introduces a scalable variant-calling method using DeepVariant and GLnexus, achieving improved quality and cost savings over existing practices.

A convolutional network trained on 16S reads from 13,838 species reaches within 2.0% of training-set Bayes-optimal species accuracy and usually beats k-mer, alignment and Kraken2 on held-out-species genera.

The paper develops a reproducible pipeline to integrate sequencing datasets for high-confidence SNP, indel, and reference calls in five human genomes, showing improved variant detection compared to prior methods.

A convolutional network on read-pileup images outperforms hand-built statistical variant callers, won an FDA challenge award for SNPs, learns new sequencing technologies and generalizes to other mammals.

The paper develops a machine learning system using convolutional neural networks to predict cell type-specific epigenetic and transcriptional profiles from DNA sequence, showing alignment with eQTL causal variants.

This paper introduces a method for improving phenotypic measurements in high-content imaging screens by using a pre-trained deep metric network to classify drugs based on molecular mechanisms without requiring additional labeled data.