Kenney Ng
Kenney Ng has a total of 275 co-authors. Here are the top co-authors by number of shared publications:
- Patrick Thomas Ellinor
Massachusetts General Hospital
12 shared publications - Puneet Batra
Data Sciences Platform, Broad Institute of MIT and Harvard, Cambridge, USA
10 shared publications - Steven A. Lubitz
Cardiovascular Research Center, Massachusetts General Hospital, Boston, USA
10 shared publications - Bum Chul Kwon
IBM Research, Cambridge, MA, USA
8 shared publications - Amit V. Khera
Program in Medical and Population Genetics, Broad Institute of MIT and Harvard, Cambridge, MA
7 shared publications - James P. Pirruccello
Division of Cardiology, University of California-San Francisco, San Francisco, California, USA
6 shared publications
Latest Publications
- Open Access
- 28 January 2026
- Published by Wiley in Advanced Science
- Vol. 13 (14) , e17840
Multi‐View Biomedical Foundation Models for Molecule‐Target and Property Prediction
Molecular foundation models hold promise to provide accurate predictions for a large and diverse set of downstream tasks in bio‐medical research. Quality molecular representations are key and foundation model development has typically focused on a single representation or molecular view, which may have strengths or weaknesses on a given task. We develop Multi‐view Molecular Embedding with Late Fusion (MMELON), an approach that integrates pre‐trained graph, image and text foundation models and may be readily extended to additional views and models. The multi‐view model performs robustly and is validated on over 120 tasks, including molecular solubility, ADME properties, and activity against G Protein‐Coupled receptors (GPCRs). The GPCR model array is leveraged to perform a virtual screen in search of ligands binding to Alzheimer's disease related GPCRs. We identify a number of such targets and employ the multi‐view model to select strong binders from a compound screen. Predictions are validated through structure‐based modeling and identification of key binding motifs.
- 28 January 2026
- Published by Wiley in Advanced Science
- Vol. 13 (14) , e17840
Open AccessMulti‐View Biomedical Foundation Models for Molecule‐Target and Property Prediction
Molecular foundation models hold promise to provide accurate predictions for a large and diverse set of downstream tasks in bio‐medical research. Quality molecular representations are key and foundation model development has typically focused on a single representation or molecular view, which may have strengths or weaknesses on a given task. We develop Multi‐view Molecular Embedding with Late Fusion (MMELON), an approach that integrates pre‐trained graph, image and text foundation models and may be readily extended to additional views and models. The multi‐view model performs robustly and is validated on over 120 tasks, including molecular solubility, ADME properties, and activity against G Protein‐Coupled receptors (GPCRs). The GPCR model array is leveraged to perform a virtual screen in search of ligands binding to Alzheimer's disease related GPCRs. We identify a number of such targets and employ the multi‐view model to select strong binders from a compound screen. Predictions are validated through structure‐based modeling and identification of key binding motifs.
- Open Access
- 14 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-11
PhysiOpt: Physics-Driven Shape Optimization for 3D Generative Models
- 14 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-11
Open AccessPhysiOpt: Physics-Driven Shape Optimization for 3D Generative Models
- Open Access
- 20 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-2
PhysiOpt: Physics-Driven Shape Optimization for 3D Generative Models
- 20 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Open AccessPhysiOpt: Physics-Driven Shape Optimization for 3D Generative Models
- Open Access
- 1 August 2024
- Published by International Joint Conferences on Artificial Intelligence
- Vol. 9, 8713-8716
SPARK: Harnessing Human-Centered Workflows with Biomedical Foundation Models for Drug Discovery
Biomedical foundation models, trained on diverse sources of small molecule data, hold great potential for accelerating drug discovery. However, their complex nature often presents a barrier for researchers seeking scientific insights and drug candidate generation. SPARK addresses this challenge by providing a user-friendly, web-based interface that empowers researchers to leverage these powerful models in their scientific workflows. Through SPARK, users can specify target proteins and desired molecule properties, adjust pre-trained models for tailored inferences, generate lists of potential drug candidates, analyze and compare molecules through interactive visualizations, and filter candidates based on key metrics (e.g., toxicity). By seamlessly integrating human knowledge and biomedical AI models' capabilities through an interactive web-based system, SPARK can improve the efficiency of collaboration between human experts and AI, thereby accelerating drug candidate discovery and ultimately leading to breakthroughs in finding cures for various diseases.
- 1 August 2024
- Published by International Joint Conferences on Artificial Intelligence
- Vol. 9, 8713-8716
Open AccessSPARK: Harnessing Human-Centered Workflows with Biomedical Foundation Models for Drug Discovery
Biomedical foundation models, trained on diverse sources of small molecule data, hold great potential for accelerating drug discovery. However, their complex nature often presents a barrier for researchers seeking scientific insights and drug candidate generation. SPARK addresses this challenge by providing a user-friendly, web-based interface that empowers researchers to leverage these powerful models in their scientific workflows. Through SPARK, users can specify target proteins and desired molecule properties, adjust pre-trained models for tailored inferences, generate lists of potential drug candidates, analyze and compare molecules through interactive visualizations, and filter candidates based on key metrics (e.g., toxicity). By seamlessly integrating human knowledge and biomedical AI models' capabilities through an interactive web-based system, SPARK can improve the efficiency of collaboration between human experts and AI, thereby accelerating drug candidate discovery and ultimately leading to breakthroughs in finding cures for various diseases.
- Open Access
- 11 June 2024
- Published by American Diabetes Association in Diabetes Care
- Vol. 47 (8) , 1424-1431
Data-Driven Phenotyping of Presymptomatic Type 1 Diabetes Using Longitudinal Autoantibody Profiles
OBJECTIVE: To characterize distinct islet autoantibody profiles preceding stage 3 type 1 diabetes RESEARCH DESIGN AND METHODS: The T1DI (Type 1 Diabetes Intelligence) study combined data from 1,845 genetically susceptible prospectively observed children who were positive for at least one islet autoantibody: insulin autoantibody (IAA), GAD antibody (GADA), or islet antigen 2 antibody (IA-2A). Using a novel similarity algorithm that considers an individual’s temporal autoantibody profile, age at autoantibody appearance, and variation in the positivity of autoantibody types, we performed an unsupervised hierarchical clustering analysis. Progression rates to diabetes were analyzed via survival analysis. RESULTS: We identified five main clusters of individuals with distinct autoantibody profiles characterized by seroconversion age and sequence of appearance of the three autoantibodies. The highest 5-year risk from first positive autoantibody to type 1 diabetes (69.9%; 95% CI 60.0–79.2) was observed in children who first developed IAA in early life (median age 1.6 years) followed by GADA (1.9 years) and then IA-2A (2.1 years). Their 10-year risk was 89.9% (95% CI 81.9–95.4). A high 5-year risk was also found in children with persistent IAA and GADA (39.1%) and children with persistent GADA and IA-2A (30.9%). A lower 5-year risk (10.5%) was observed in children with a late appearance of persistent GADA (6.1 years). The lowest 5-year diabetes risk (1.6%) was associated with positivity for a single, often reverting, autoantibody. CONCLUSIONS: The novel clustering algorithm identified children with distinct islet autoantibody profiles and progression rates to diabetes. These results are useful for prediction, selection of individuals for prevention trials, and studies investigating various pathways to type 1 diabetes.
- 11 June 2024
- Published by American Diabetes Association in Diabetes Care
- Vol. 47 (8) , 1424-1431
Open AccessData-Driven Phenotyping of Presymptomatic Type 1 Diabetes Using Longitudinal Autoantibody Profiles
OBJECTIVE: To characterize distinct islet autoantibody profiles preceding stage 3 type 1 diabetes RESEARCH DESIGN AND METHODS: The T1DI (Type 1 Diabetes Intelligence) study combined data from 1,845 genetically susceptible prospectively observed children who were positive for at least one islet autoantibody: insulin autoantibody (IAA), GAD antibody (GADA), or islet antigen 2 antibody (IA-2A). Using a novel similarity algorithm that considers an individual’s temporal autoantibody profile, age at autoantibody appearance, and variation in the positivity of autoantibody types, we performed an unsupervised hierarchical clustering analysis. Progression rates to diabetes were analyzed via survival analysis. RESULTS: We identified five main clusters of individuals with distinct autoantibody profiles characterized by seroconversion age and sequence of appearance of the three autoantibodies. The highest 5-year risk from first positive autoantibody to type 1 diabetes (69.9%; 95% CI 60.0–79.2) was observed in children who first developed IAA in early life (median age 1.6 years) followed by GADA (1.9 years) and then IA-2A (2.1 years). Their 10-year risk was 89.9% (95% CI 81.9–95.4). A high 5-year risk was also found in children with persistent IAA and GADA (39.1%) and children with persistent GADA and IA-2A (30.9%). A lower 5-year risk (10.5%) was observed in children with a late appearance of persistent GADA (6.1 years). The lowest 5-year diabetes risk (1.6%) was associated with positivity for a single, often reverting, autoantibody. CONCLUSIONS: The novel clustering algorithm identified children with distinct islet autoantibody profiles and progression rates to diabetes. These results are useful for prediction, selection of individuals for prevention trials, and studies investigating various pathways to type 1 diabetes.
- 28 May 2024
- Published by Springer Nature in Nature Medicine
- Vol. 30 (6) , 1749-1760
Noninvasive assessment of organ-specific and shared pathways in multi-organ fibrosis using T1 mapping
Fibrotic diseases affect multiple organs and are associated with morbidity and mortality. To examine organ-specific and shared biologic mechanisms that underlie fibrosis in different organs, we developed machine learning models to quantify T1 time, a marker of interstitial fibrosis, in the liver, pancreas, heart and kidney among 43,881 UK Biobank participants who underwent magnetic resonance imaging. In phenome-wide association analyses, we demonstrate the association of increased organ-specific T1 time, reflecting increased interstitial fibrosis, with prevalent diseases across multiple organ systems. In genome-wide association analyses, we identified 27, 18, 11 and 10 independent genetic loci associated with liver, pancreas, myocardial and renal cortex T1 time, respectively. There was a modest genetic correlation between the examined organs. Several loci overlapped across the examined organs implicating genes involved in a myriad of biologic pathways including metal ion transport (SLC39A8, HFE and TMPRSS6), glucose metabolism (PCK2), blood group antigens (ABO and FUT2), immune function (BANK1 and PPP3CA), inflammation (NFKB1) and mitosis (CENPE). Finally, we found that an increasing number of organs with T1 time falling in the top quintile was associated with increased mortality in the population. Individuals with a high burden of fibrosis in ≥3 organs had a 3-fold increase in mortality compared to those with a low burden of fibrosis across all examined organs in multivariable-adjusted analysis (hazard ratio = 3.31, 95% confidence interval 1.77–6.19; P = 1.78 × 10−4). By leveraging machine learning to quantify T1 time across multiple organs at scale, we uncovered new organ-specific and shared biologic pathways underlying fibrosis that may provide therapeutic targets.
- 28 May 2024
- Published by Springer Nature in Nature Medicine
- Vol. 30 (6) , 1749-1760
Noninvasive assessment of organ-specific and shared pathways in multi-organ fibrosis using T1 mapping
Fibrotic diseases affect multiple organs and are associated with morbidity and mortality. To examine organ-specific and shared biologic mechanisms that underlie fibrosis in different organs, we developed machine learning models to quantify T1 time, a marker of interstitial fibrosis, in the liver, pancreas, heart and kidney among 43,881 UK Biobank participants who underwent magnetic resonance imaging. In phenome-wide association analyses, we demonstrate the association of increased organ-specific T1 time, reflecting increased interstitial fibrosis, with prevalent diseases across multiple organ systems. In genome-wide association analyses, we identified 27, 18, 11 and 10 independent genetic loci associated with liver, pancreas, myocardial and renal cortex T1 time, respectively. There was a modest genetic correlation between the examined organs. Several loci overlapped across the examined organs implicating genes involved in a myriad of biologic pathways including metal ion transport (SLC39A8, HFE and TMPRSS6), glucose metabolism (PCK2), blood group antigens (ABO and FUT2), immune function (BANK1 and PPP3CA), inflammation (NFKB1) and mitosis (CENPE). Finally, we found that an increasing number of organs with T1 time falling in the top quintile was associated with increased mortality in the population. Individuals with a high burden of fibrosis in ≥3 organs had a 3-fold increase in mortality compared to those with a low burden of fibrosis across all examined organs in multivariable-adjusted analysis (hazard ratio = 3.31, 95% confidence interval 1.77–6.19; P = 1.78 × 10−4). By leveraging machine learning to quantify T1 time across multiple organs at scale, we uncovered new organ-specific and shared biologic pathways underlying fibrosis that may provide therapeutic targets.
- Open Access
- 21 May 2024
- Published by Springer Nature in Nature Communications
- Vol. 15 (1) , 1-17
Deep learning of left atrial structure and function provides link to atrial fibrillation risk
Increased left atrial volume and decreased left atrial function have long been associated with atrial fibrillation. The availability of large-scale cardiac magnetic resonance imaging data paired with genetic data provides a unique opportunity to assess the genetic contributions to left atrial structure and function, and understand their relationship with risk for atrial fibrillation. Here, we use deep learning and surface reconstruction models to measure left atrial minimum volume, maximum volume, stroke volume, and emptying fraction in 40,558 UK Biobank participants. In a genome-wide association study of 35,049 participants without pre-existing cardiovascular disease, we identify 20 common genetic loci associated with left atrial structure and function. We find that polygenic contributions to increased left atrial volume are associated with atrial fibrillation and its downstream consequences, including stroke. Through Mendelian randomization, we find evidence supporting a causal role for left atrial enlargement and dysfunction on atrial fibrillation risk.
- 21 May 2024
- Published by Springer Nature in Nature Communications
- Vol. 15 (1) , 1-17
Open AccessDeep learning of left atrial structure and function provides link to atrial fibrillation risk
Increased left atrial volume and decreased left atrial function have long been associated with atrial fibrillation. The availability of large-scale cardiac magnetic resonance imaging data paired with genetic data provides a unique opportunity to assess the genetic contributions to left atrial structure and function, and understand their relationship with risk for atrial fibrillation. Here, we use deep learning and surface reconstruction models to measure left atrial minimum volume, maximum volume, stroke volume, and emptying fraction in 40,558 UK Biobank participants. In a genome-wide association study of 35,049 participants without pre-existing cardiovascular disease, we identify 20 common genetic loci associated with left atrial structure and function. We find that polygenic contributions to increased left atrial volume are associated with atrial fibrillation and its downstream consequences, including stroke. Through Mendelian randomization, we find evidence supporting a causal role for left atrial enlargement and dysfunction on atrial fibrillation risk.
- Open Access
Latent Space Explorer: Visual Analytics for Multimodal Latent Space Exploration
Machine learning models built on training data with multiple modalities can reveal new insights that are not accessible through unimodal datasets. For example, cardiac magnetic resonance images (MRIs) and electrocardiograms (ECGs) are both known to capture useful information about subjects' cardiovascular health status. A multimodal machine learning model trained from large datasets can potentially predict the onset of heart-related diseases and provide novel medical insights about the cardiovascular system. Despite the potential benefits, it is difficult for medical experts to explore multimodal representation models without visual aids and to test the predictive performance of the models on various subpopulations. To address the challenges, we developed a visual analytics system called Latent Space Explorer. Latent Space Explorer provides interactive visualizations that enable users to explore the multimodal representation of subjects, define subgroups of interest, interactively decode data with different modalities with the selected subjects, and inspect the accuracy of the embedding in downstream prediction tasks. A user study was conducted with medical experts and their feedback provided useful insights into how Latent Space Explorer can help their analysis and possible new direction for further development in the medical domain.
Open AccessLatent Space Explorer: Visual Analytics for Multimodal Latent Space Exploration
Machine learning models built on training data with multiple modalities can reveal new insights that are not accessible through unimodal datasets. For example, cardiac magnetic resonance images (MRIs) and electrocardiograms (ECGs) are both known to capture useful information about subjects' cardiovascular health status. A multimodal machine learning model trained from large datasets can potentially predict the onset of heart-related diseases and provide novel medical insights about the cardiovascular system. Despite the potential benefits, it is difficult for medical experts to explore multimodal representation models without visual aids and to test the predictive performance of the models on various subpopulations. To address the challenges, we developed a visual analytics system called Latent Space Explorer. Latent Space Explorer provides interactive visualizations that enable users to explore the multimodal representation of subjects, define subgroups of interest, interactively decode data with different modalities with the selected subjects, and inspect the accuracy of the embedding in downstream prediction tasks. A user study was conducted with medical experts and their feedback provided useful insights into how Latent Space Explorer can help their analysis and possible new direction for further development in the medical domain.
- Open Access
Efficient estimation of weighted cumulative treatment effects by double/debiased machine learning
In empirical studies with time-to-event outcomes, investigators often leverage observational data to conduct causal inference on the effect of exposure when randomized controlled trial data is unavailable. Model misspecification and lack of overlap are common issues in observational studies, and they often lead to inconsistent and inefficient estimators of the average treatment effect. Estimators targeting overlap weighted effects have been proposed to address the challenge of poor overlap, and methods enabling flexible machine learning for nuisance models address model misspecification. However, the approaches that allow machine learning for nuisance models have not been extended to the setting of weighted average treatment effects for time-to-event outcomes when there is poor overlap. In this work, we propose a class of one-step cross-fitted double/debiased machine learning estimators for the weighted cumulative causal effect as a function of restriction time. We prove that the proposed estimators are consistent, asymptotically linear, and reach semiparametric efficiency bounds under regularity conditions. Our simulations show that the proposed estimators using nonparametric machine learning nuisance models perform as well as established methods that require correctly-specified parametric nuisance models, illustrating that our estimators mitigate the need for oracle parametric nuisance models. We apply the proposed methods to real-world observational data from a UK primary care database to compare the effects of anti-diabetic drugs on cancer clinical outcomes.
Open AccessEfficient estimation of weighted cumulative treatment effects by double/debiased machine learning
In empirical studies with time-to-event outcomes, investigators often leverage observational data to conduct causal inference on the effect of exposure when randomized controlled trial data is unavailable. Model misspecification and lack of overlap are common issues in observational studies, and they often lead to inconsistent and inefficient estimators of the average treatment effect. Estimators targeting overlap weighted effects have been proposed to address the challenge of poor overlap, and methods enabling flexible machine learning for nuisance models address model misspecification. However, the approaches that allow machine learning for nuisance models have not been extended to the setting of weighted average treatment effects for time-to-event outcomes when there is poor overlap. In this work, we propose a class of one-step cross-fitted double/debiased machine learning estimators for the weighted cumulative causal effect as a function of restriction time. We prove that the proposed estimators are consistent, asymptotically linear, and reach semiparametric efficiency bounds under regularity conditions. Our simulations show that the proposed estimators using nonparametric machine learning nuisance models perform as well as established methods that require correctly-specified parametric nuisance models, illustrating that our estimators mitigate the need for oracle parametric nuisance models. We apply the proposed methods to real-world observational data from a UK primary care database to compare the effects of anti-diabetic drugs on cancer clinical outcomes.
- Top Cited PapersOpen Access
- 28 April 2023
- Published by Springer Nature in Nature Communications
- Vol. 14 (1) , 1-12
Cross-modal autoencoder framework learns holistic representations of cardiovascular state
A fundamental challenge in diagnostics is integrating multiple modalities to develop a joint characterization of physiological state. Using the heart as a model system, we develop a cross-modal autoencoder framework for integrating distinct data modalities and constructing a holistic representation of cardiovascular state. In particular, we use our framework to construct such cross-modal representations from cardiac magnetic resonance images (MRIs), containing structural information, and electrocardiograms (ECGs), containing myoelectric information. We leverage the learned cross-modal representation to (1) improve phenotype prediction from a single, accessible phenotype such as ECGs; (2) enable imputation of hard-to-acquire cardiac MRIs from easy-to-acquire ECGs; and (3) develop a framework for performing genome-wide association studies in an unsupervised manner. Our results systematically integrate distinct diagnostic modalities into a common representation that better characterizes physiologic state.
- 28 April 2023
- Published by Springer Nature in Nature Communications
- Vol. 14 (1) , 1-12
Top Cited PapersOpen AccessCross-modal autoencoder framework learns holistic representations of cardiovascular state
A fundamental challenge in diagnostics is integrating multiple modalities to develop a joint characterization of physiological state. Using the heart as a model system, we develop a cross-modal autoencoder framework for integrating distinct data modalities and constructing a holistic representation of cardiovascular state. In particular, we use our framework to construct such cross-modal representations from cardiac magnetic resonance images (MRIs), containing structural information, and electrocardiograms (ECGs), containing myoelectric information. We leverage the learned cross-modal representation to (1) improve phenotype prediction from a single, accessible phenotype such as ECGs; (2) enable imputation of hard-to-acquire cardiac MRIs from easy-to-acquire ECGs; and (3) develop a framework for performing genome-wide association studies in an unsupervised manner. Our results systematically integrate distinct diagnostic modalities into a common representation that better characterizes physiologic state.