^{Biostatistics}

Li Group: Qian Li

Developing robust methods and tools to analyze bulk, single cell, and spatial multi-omics data, including longitudinal omics data.

See Details

About the Li Group

As scientists collect increasingly complex datasets on the molecular profiles of humans, our methods to analyze and model these datasets should advance as well. To address this need, we develop models and computational tools to analyze and integrate different types of molecular data from clinical research. Our work strives to make the modeling and analysis of high-dimensional molecular data more accessible and reliable to researchers.

The Team

The Qian Li Group supports the analysis of omics data from different diseases in basic science, clinical trials, and translational studies.

Meet the Team

Li Group research summary

The modeling and analysis of high-dimensional molecular data—known as omics data—is an important aspect of biomedical studies. Our group focuses on statistical learning and deep learning methods and collaborations in multi-omics to support basic science or clinical research.

Method development and collaborative research

Our method development strives to build models and tools to analyze omics and/or multi-omics data collected from patients in various diseases, including pediatric CNS tumors, pediatric leukemia, neurodegenerative disease, type 1 diabetes. Our group supports advanced analysis of molecular data and clinical data from studies in pediatric cancers at St. Jude, such as retrospective studies or clinical trials in pediatric neuro-oncology. Our work in methodology development and collaborations seek to advance the analysis and interpretation of multi-omics data with clinical utility.

Featured software

MOADE

For more information, see “MOADE: A multimodal autoencoder for dissociating bulk multi-omics data” in the list below.

ISLET

MOADE

For more information, see "ISLET: Individual-Specific ceLl typE referencing Tool" in the list below.

Software packages

GMSimpute: Generalized Mass Spectrum Missing Peaks Abundance Imputation

In the GMSimpute R package, two-Step Lasso (TS-Lasso) and compound minimum methods are used to recover the abundance of missing peaks in mass spectrum analysis. TS-Lasso is an imputation method that handles various types of missing peaks simultaneously. This package provides the procedure to generate missing peaks (or data) for simulation study, as well as a tool to estimate and visualize the proportion of missing at random.
IncDIFF: Long Non-Coding RNA Differential Expression Analysis

In the IncDIFF R package, we developed an approach to detect differential expression features in long non-coding RNA low counts, using generalized linear model with zero-inflated exponential quasi likelihood ratio test. Methods implemented in this package are described in Li (2019).
ISLET: Individual-Specific ceLl typE referencing Tool

ISLET is a method to conduct signal deconvolution for general -omics data. It can estimate the individual-specific and cell-type-specific reference panels, when there are multiple samples observed from each subject. It takes the input of the observed mixture data (feature by sample matrix), the cell type mixture proportions (sample by cell type matrix), and the sample-to-subject information. ISLET can solve for the reference panel on the individual-basis and conduct tests to identify cell-type-specific differential expression (csDE) genes.
MICSQTL: Multi-omic deconvolution, Integration and Cell-type-specific Quantitative Trait Loci

Our pipeline, MICSQTL, utilizes scRNA-seq reference and bulk transcriptomes to estimate cellular composition in the matched bulk proteomes. The expression of genes and proteins at either bulk level or cell type level can be integrated by the Angle-based Joint and Individual Variation Explained (AJIVE) framework. Meanwhile, MICSQTL can perform cell-type-specific quantitative trait loci (QTL) mapping to proteins or transcripts based on the input of bulk expression data and the estimated cellular composition per molecule type, without the need for single cell sequencing. We use matched transcriptome-proteome from human brain frontal cortex tissue samples to demonstrate the input and output of our tool.
MOADE: A multimodal autoencoder for dissociating bulk multi-omics data

MOADE, a multimodal autoencoder pipeline linking multi-dimensional features to jointly predict personalized multi-omic profiles and cellular compositions, using pseudo-bulk data constructed by internal non-transcriptomic reference and external scRNA-seq data. MOADE is evaluated through rigorous simulation experiments and real multi-omic data from multiple tissue types, outperforming nine deconvolution pipelines with superior generalizability and fidelity.