Research — Ding Lab

The lab develops statistical methodology motivated by real scientific problems. We are drawn to questions where the data are complex — censored, high-dimensional, correlated, or multi-source — and where careful modeling can change what a study is able to conclude. The four areas below describe our current directions; each is a placeholder to be expanded, with room for a representative figure.

Semiparametric methods for complex time-to-event data

We develop semiparametric and copula-based models for survival data that arise under interval censoring, left truncation, and multivariate (for example, bilateral or paired) outcomes. This includes estimation theory, goodness-of-fit diagnostics, and openly available software.

Architecture of the tdCoxSNN time-dependent Cox survival neural network — The tdCoxSNN architecture — a time-dependent Cox survival neural network for continuous-time dynamic prediction (Zeng et al., *JRSS Series C*, 2025).

Precision medicine: subgroups and individualized effects

We build methods for identifying and drawing confident inference about subgroups with differential treatment efficacy in randomized trials, and for estimating individualized treatment effects. A recurring theme is logically coherent inference — conclusions about subgroups and their mixtures that remain consistent with one another.

Causal subgroup discovery for heterogeneous treatment effects — Causal subgroup discovery for interpretable heterogeneous treatment effects in survival outcomes (Bo & Ding, *Lifetime Data Analysis*, 2026).

Statistical & deep learning for genetic and multi-omics data

We design statistical and deep-learning methods for high-dimensional genetic and multi-omics data: prediction from genome-wide association data, model-based clustering of single-cell sequencing data, and the joint analysis of multiple data modalities.

Deep-learning model overview for predicting AMD progression — A deep-learning model for predicting late age-related macular degeneration progression from GWAS and imaging data (Yan et al., *Nature Machine Intelligence*, 2020).

Scientific collaboration

Our methods are grounded in long-standing collaborations — among them studies of age-related macular degeneration, Alzheimer's disease and related psychosis, schizophrenia, and childhood asthma. These partnerships supply the questions that motivate new methodology and the data on which it is tested.

Overview of a unified framework for doublet detection in single-cell multi-omics data — A unified framework for doublet and multiplet detection in single-cell multi-omics data (Hu et al., *Nature Communications*, 2024).

Funding

The lab's work is supported by federal and institutional grants. Recent and current awards include:

As principal investigator

2026–31Advancing deep survival methods for prediction, subgroup identification, and causal inference. NIH/NIGMS, R35GM164304.
2022–27New statistical methods and software for modeling complex multivariate survival data with large-scale covariates. NIH/NIGMS, R01GM141076.
2025–26Enhancing pulmonary function assessment and out-of-clinic care through smartphone-based ultrasonic technology (SURE). University of Pittsburgh (SSOE / SPH / CTSI).
2025Multiomic profiling of lipidomic and proteomic associations in patients prior to developing asparaginase-associated pancreatitis. Stanford University.

As co-investigator (selected)

2022–27SCH: New advanced machine learning framework for mining heterogeneous ocular data. NIH/NIBIB, R01EB034116.
2023–28Providing new insight into adolescent dendritic development. NIH/NIMH, R01MH132586.
2018–28Accelerating treatment development for psychosis in Alzheimer's disease. NIH/NIMH, R01MH116046.

Selected publications →