Lingfei Wang

Variations in blood protein levels have been linked to numerous complex diseases, including cardiovascular conditions. These associations highlight the intricate interplay between local and systemic factors in cardiovascular disease development, emphasizing the need for a comprehensive, systems-level understanding of its etiology. To address this, we develop a causal network inference framework using data from one of the largest serum proteomics studies to date, comprising measurements of 7523 serum proteins in the prospective, population-based Age, Gene/Environment Susceptibility-Reykjavik Study (AGES) cohort of 5376 older adults. Using cis -acting protein quantitative trait loci (pQTLs) as instrumental variables within a causal inference framework designed to mitigate hidden confounding, we identify 185 high-confidence causal serum protein subnetworks collectively interacting with 5611 targets. Several subnetworks, many forming hierarchical frameworks of directional relationships, are significantly associated with multiple cardiometabolic traits and with future risk of myocardial infarction and its long-term complication, heart failure.

Single-cell expression quantitative trait loci (sceQTL) mapping offers a powerful approach for understanding gene regulation and its heterogeneity across cell types and states. It has profound applications in genetics and genomics, particularly causal gene regulatory network (cGRN) inference to unravel the molecular circuits governing cell identity and function. However, computational scalability remains a critical bottleneck for sceQTL mapping, prohibiting thorough benchmarking and optimization of statistical accuracy. We present airqtl, an efficient method to overcome these challenges through algorithmic advances and efficient implementations of linear mixed models. Airqtl achieves superior time complexity and over 10 ⁸ times of acceleration, enabling objective method benchmarking and optimization. Airqtl offers de novo inference of robust, experimentally validated cell state-specific cGRNs that reflect perturbation outcomes. Our results dissect the drivers of cGRN heterogeneity and underscore the value of natural genetic variations in primary human cell types for biologically relevant single-cell cGRN inference.

Variations in blood protein levels have been associated with a broad spectrum of complex diseases, including atherosclerotic cardiovascular disease (ACVD). These associations highlight the intricate interplay between local (e.g., cardiovascular) and systemic (non-cardiovascular) factors for the development of ACVD, emphasizing the need for a comprehensive, systems-level understanding of its etiology. To accomplish this, we developed a causal network inference framework by analyzing one of the largest serum proteomics studies to date, the Age, Gene/Environment Susceptibility-Reykjavik Study (AGES), a prospective population-based study of 7,523 serum proteins measured in 5,376 older adults. To reconstruct a causal network of serum proteins, we used cis -acting protein quantitative trait loci (pQTLs) as instrumental variables to infer causal relationships between protein pairs, while accounting for potential unobserved confounding factors. We identified 185 causal protein subnetworks (FDR = 1%, n ≥ 10 members), which collectively interacted with 5,611 target proteins, offering valuable biological insights and an overview of systemic homeostasis. Several subnetworks, many of which interact to establish a hierarchy of directional relationships, were significantly associated with future myocardial infarction and/or its long-term complications like heart failure, as well as with key cardiometabolic traits that contribute to the onset of ACVD.

Single-cell expression quantitative trait loci (sceQTL) mapping offers a powerful approach for understanding gene regulation and its heterogeneity across cell types and states. It has profound applications in genetics and genomics, particularly causal gene regulatory network (cGRN) inference to unravel the molecular circuits governing cell identity and function. However, computational scalability remains a critical bottleneck for sceQTL mapping, prohibiting thorough benchmarking and optimization of statistical accuracy. We present airqtl, a novel method to overcome these challenges through algorithmic advances and efficient implementations of linear mixed models. Airqtl achieves superior time complexity and over 10 ⁸ times of acceleration, enabling objective method benchmarking and optimization. Airqtl offers de novo inference of robust, experimentally validated cell state-specific cGRNs that reflect perturbation outcomes. Our results dissect the drivers of cGRN heterogeneity and underscore the value of natural genetic variations in primary human cell types for biologically relevant single-cell cGRN inference.

Gene regulatory networks (GRNs) are key determinants of cell function and identity and are dynamically rewired during development and disease. Despite decades of advancement, challenges remain in GRN inference: dynamic rewiring, causal inference, feedback-loop modeling, and context specificity. To address them, we develop Dictys, a dynamic GRN inference and analysis method which leverages multi-omic single-cell assays of chromatin accessibility and gene expression, context specific transcription factor (TF) footprinting, stochastic process network, and efficient probabilistic modeling of scRNA-seq read counts. Dictys improves GRN reconstruction accuracy and reproducibility and enables the inference and comparative analysis of context specific and dynamic GRNs across developmental contexts. Dictys’ network analyses recover unique insights in human blood and mouse skin development with cell-type specific and dynamic GRNs. Its dynamic network visualizations enable time-resolved discovery and investigation of developmental driver TFs and their regulated targets. Dictys is available as a free, open source, and user-friendly Python package.

Single-cell RNA sequencing (scRNA-seq) provides unprecedented technical and statistical potential to study gene regulation but is subject to technical variations and sparsity. Furthermore, statistical association testing remains difficult for scRNA-seq. Here we present Normalisr, a normalization and statistical association testing framework that unifies single-cell differential expression, co-expression, and CRISPR screen analyses with linear models. By systematically detecting and removing nonlinear confounders arising from library size at mean and variance levels, Normalisr achieves high sensitivity, specificity, speed, and generalizability across multiple scRNA-seq protocols and experimental conditions with unbiased p-value estimation. The superior scalability allows us to reconstruct robust gene regulatory networks from trans-effects of guide RNAs in large-scale single cell CRISPRi screens. On conventional scRNA-seq, Normalisr recovers gene-level co-expression networks that recapitulated known gene functions.

Single-cell RNA sequencing (scRNA-seq) provides unprecedented technical and statistical potential to study gene regulation but is subject to technical variations and sparsity. Here we present Normalisr, a linear-model-based normalization and statistical hypothesis testing framework that unifies single-cell differential expression, co-expression, and CRISPR scRNA-seq screen analyses. By systematically detecting and removing nonlinear confounding from library size, Normalisr achieves high sensitivity, specificity, speed, and generalizability across multiple scRNA-seq protocols and experimental conditions with unbiased P-value estimation. We use Normalisr to reconstruct robust gene regulatory networks from trans-effects of gRNAs in large-scale CRISPRi scRNA-seq screens and gene-level co-expression networks from conventional scRNA-seq.

Transcriptional control of proteostasis: Tissue homeostasis requires the coordinated activity of multiple cell types to initiate and then resolve inflammation. Intrinsic cellular stress-response pathways facilitate adaptation to stress and tissue restitution. Among these stress pathways, the unfolded protein response can elicit two divergent outcomes: adaptation to endoplasmic reticulum (ER) stress or termination by programmed cell death. You et al. identified QRICH1 as a transcriptional regulator controlling adaptation to ER stress at the level of protein translation and secretion. The authors further demonstrate the role of the QRICH1 program in inflammatory diseases of the colon and liver. Science , this issue p. 45

A genome wide meta-analysis by the CORtisol NETwork (CORNET) consortium(1) has identified genetic variants spanning the SERPINA6/SERPINA1 locus on chromosome 14, associated with morning plasma cortisol and predictive of cardiovascular disease (Crawford et al, Unpublished). SERPINA6 encodes Corticosteroid Binding Globulin (CBG), responsible for binding most cortisol in blood and putatively mediating delivery of cortisol to target tissues. We hypothesised that genetic variants in SERPINA6 influence CBG expression in liver and cortisol delivery to extra-hepatic tissues, influencing cortisol-regulated gene expression. The Stockholm Tartu Atherosclerosis Reverse Networks Engineering Task study (STARNET)(2) provides RNA sequencing data in 9 vascular and metabolic tissues from 600 genotyped individuals (mean age 65.8, 70.3% male) undergoing coronary artery bypass grafting. We used STARNET to identify SNPs associated with plasma cortisol at genome wide significance in CORNET as cis-eQTLs for SERPINA6 in liver and as trans-eQTLs for the expression of genes across STARNET tissues. Causal inference methodologies(3) were then employed for the network reconstruction of these trans-genes and their downstream targets. We identified 21 SNPs that both were associated with cortisol at genome wide significance in CORNET (p ≤ 5x10-8) and were cis-eQTLs for SERPINA6 expression in liver (q ≤ 0.05). Moreover, these SNPs were trans-eQTLs for sets of genes in liver, subcutaneous and visceral abdominal adipose tissue, with over-representation of known glucocorticoid-regulated genes in adipose. The highest confidence gene network identified was specific to subcutaneous adipose, with the interferon regulatory trans-gene, IRF2, controlling a putative glucocorticoid-regulated network. Targets in this network include LDB2 and LIPA, both associated with coronary artery disease. We conclude that variants in the SERPINA6/SERPINA1 locus mediate their effect on plasma cortisol through variation in SERPINA6 expression in liver, and in turn affect gene expression in extra-hepatic tissues through modulating cortisol delivery. This supports a dynamic role for CBG in modulating cortisol delivery to tissues. The cortisol-responsive gene networks identified here represent candidate pathways to mediate cardiovascular risk attributable to elevated cortisol. (1) Bolton, et al. (2014) PLOS Genet. 10:e1004474., (2) Franzén et al. (2016). Science 353:827., (3) Wang and Michoel. (2017). PLOS Comput. Biol. 13:e1005703.