Dokyoon Kim

Integrating Multisource Target Evidence with Machine-Learning Potency Prediction for Alzheimer's Disease Drug Discovery

The paper develops a platform integrating multisource target evidence with machine learning for Alzheimer's drug discovery, improving potency prediction and model performance.

Decomposing scanner batch effects and their impact on pathology foundation models

This study decomposes scanner batch effects in pathology foundation models and shows that image corrections based on measured properties improve alignment but do not fully match model representations.

Limits of marker information transfer]{Limits of marker information transfer in cross-panel cytometry imputation

This study evaluates the limits of marker information transfer in cross-panel cytometry imputation, finding that imputation preserves broad lineage classification but not fine-phenotype details, and that patient matching improves cross-platform distribution recovery.

Absorption and Co-expression Modules Show Where Polygenic and Proteomic Risk Scores Diverge in Neurodegenerative Diseases

The paper quantifies the divergence between polygenic and proteomic risk scores in neurodegenerative diseases by analyzing absorption and co-expression modules across multiple cohorts.

CytoGate-Bench: an LLM benchmark for cross-panel cell gating in cytometry

CytoGate-Bench evaluates LLMs for cross-panel cell gating in cytometry by reformulating manual gating as a zero-shot task using expert-annotated data.

Deep learning representations of human Immune Health for precision immunology

The paper develops MAESTRO, a self-supervised deep learning framework to transform immune profiles into embeddings capturing immune health, enabling clinical prediction and precision immunotherapy.

Genetic liability to meniscus degeneration and its comorbidity patterns: a phenome-wide association study in the UK Biobank

This study identifies genetic risk factors for meniscus degeneration and shows they are linked to various systemic conditions through a phenome-wide association study.

Large-scale evaluation of proteomic and polygenic risk scores reveals complementary contributions to incident disease prediction

The paper evaluates proteomic and polygenic risk scores for disease prediction, finding proteomic scores outperform polygenic scores in most traits and show temporal responsiveness.

Genome-Wide Assessment of Pleiotropy Across >1000 Traits from Global Biobanks

The paper identifies pleiotropic loci across 1,167 traits using GWAS meta-analysis and shows they are enriched for functional annotations and improve polygenic score predictions.

Cytometry Masked Autoencoder: An Accurate and Interpretable Automated Immunophenotyper

The paper introduces cyMAE, a masked autoencoder for automated immunophenotyping, showing it improves accuracy and interpretability over manual methods in cytometry data analysis.

Uncovering genetic associations in the human diseasome using an endophenotype-augmented disease network

This study constructs an augmented disease network incorporating genetic correlations with endophenotypes to uncover genetic associations across diseases and biomarkers.

Quantifying factors that affect polygenic risk score performance across diverse ancestries and age groups for body mass index

This study evaluates how ancestry, GWAS sample size, and LD reference panels affect BMI PRS performance across different age groups and ancestries.

netCRS: Network-based comorbidity risk score for prediction of myocardial infarction using biobank-scaled PheWAS data

This paper introduces netCRS, a network-based comorbidity risk score that improves myocardial infarction prediction using PheWAS data, achieving better AUC than traditional PRS methods.

Genetic analysis of functional rare germline variants across 9 cancer types from the DiscovEHR study

The paper performs a genome-wide rare variant analysis across nine cancer types using germline WES data to identify cancer predisposition genes and pathways.

HiG2Vec: Hierarchical Representations of GeneOntology and Genes in the Poincare Ball

HiG2Vec improves gene ontology and gene representation by using Poincare embeddings to better capture hierarchical structures and predict gene interactions.

Multi-layered network-based pathway activity inference using directed random walks: application to predicting clinical outcomes in urologic cancer

The paper introduces iDRW, a multi-layered network-based pathway activity inference method, showing improved outcome prediction in urologic cancers compared to existing methods.

NETMAGE: a humaN-disEase phenoType MAp GEnerator for the Visualization of PheWAS

NETMAGE generates interactive phenotype networks from PheWAS data to visualize shared genetic associations and identify pleiotropy, aiding in drug target discovery.

Phenome-wide association study of a comprehensive health check-up database in a Korea population: Clinical application & trans-ethnic comparison

This paper conducts a PheWAS using a Korean health check-up database and compares results across ethnic groups to identify phenotype-genotype associations and their clinical relevance.

Min-Redundancy and Max-Relevance Multi-view Feature Selection for Predicting Ovarian Cancer Survival using Multi-omics Data

This study develops a multi-view feature selection method to predict ovarian cancer survival using multi-omics data and shows it outperforms single-view and baseline fusion approaches.

Collective feature selection to identify crucial epistatic variants

The paper introduces a collective feature selection approach to identify epistatic variants by combining multiple methods, showing improved performance in detecting true positive interactions in both simulated and real BMI data.