The Bolouri lab is interested in developing and applying computational systems immunology methods to better understand immune dysregulation in autoimmunity, cancer, and infection (most recently COVID-19). Our approach is inherently multi- and cross-disciplinary and is predicated on:

• Tight integration of biological and immunological insights
• Use of high-throughput technologies
• Methods derived from computer science, engineering, mathematics, statistics, and physics

We typically integrate data from diverse assays (e.g. single cell RNAseq, flow/mass cytometry, proteomics and epigenomics) with patient clinical data as well as curated data from databases/publications.

Dysregulated immune processes in pediatric Acute Myeloid leukemia (AML)

In spite of its heavy toll on families and young lives, treatments for pediatric AML lag far behind other leukemias. Healthy hematopoiesis relies on the coordinated birth, differentiation, and movement of dozens of bone marrow (BM) cell types, signaling pathways, and developmental gene regulatory interactions. The dysregulation of these pathways and processes in AML has been the subject of intense research for more than forty years and resulted in the development of more than 65 molecularly targeted drugs. Nonetheless the long-term survival rate for about 1 in 3 children with AML remains at approximately 30%.

Using multiple complementary assays and a cohort of nearly 1000 children with AML, we previously showed that childhood AML is distinct from adult AML in terms of the types and frequencies of somatic genomic alterations, the number of leukemic sub-clones per patient, and epigenomic and transcriptomic profiles [Bolouri et al, Nat Med. 2018; 24(1):103-112]. Continuing this collaboration with the Meshinchi laboratory at Fred Hutch, we recently established that a B-cell developmental gene regulatory network (see interaction network figure) is epigenetically activated in <3-year infants with AML. As shown in the heatmap below, the expression of genes associated with this network is sharply up-regulated in infant AML samples carrying distinct genomic alteration patterns. These findings suggest infant AML may respond to existing drugs, and immune therapies targeting B-cell specific genes such as CD19, CD20, CD22, and CD79A. Moreover, our 56-gene infant-AML gene signature can potentially be used to identify infants that would benefit most from such therapies.

In ongoing research, we are using large-scale RNA-seq, miRNA-seq, DNA methylation arrays, and whole genome sequencing data to further pinpoint the distinct and overlapping roles of dysregulated developmental and immune pathways in childhood AML.

Multi-scale regulation of immune homeostasis and steady-state maintenance

The mammalian immune system is constantly undergoing self-renewal, while also interacting with environmental cues, self-antigens, the central nervous system and commensal microbiota. How does the healthy immune system maintain a balanced steady state in the midst of all these changes? And how are the regulatory mechanisms that maintain a healthy immune balance impacted in disease? Homeostasis and steady-state are well known to be regulated through combinations of self-reinforcing and stabilizing feedback loops. In the immune system, these feedbacks occur at multiple scales, within individual cells, through cell-cell interactions, and via organ and organism level regulatory signals.

In collaboration with the BRI Sound Life Project, we are leveraging recent developments in single cell omics and high-throughput cytometry to characterize the intra- and inter-cellular immune regulatory interactions in healthy humans.This collaborative project has two components:

1. Using integrative bioinformatics tools to reverse-engineer regulatory interactions from single cell and bulk blood measurements from multiple individuals at multiple time points.
2. Development of novel multi-scale and multi-modal data integration methods that optimally exploit existing knowledge from the literature, collaborating domain experts and curated databases.

Unbiased semi-supervised cell population labeling

Currently, automated data processing pipelines for high-throughput flow/mass cytometry and single-cell omics (e.g. scRNAseq, scATACseq) typically require ad-hoc, manual labeling of clusters of cells into known cell types. Automated approaches are attractive because they can enable unbiased/uniform processing of large numbers of samples while addressing batch to batch variations and other potential confounding factors. Existing approaches to automated cell labeling require either cell-type-specific markers or sets of differentially up/down regulated genes per cell type. However, in diseases involving immune dysregulation, changes in the transcriptional states of immune cells can affect signature and marker gene expression, confounding cellular identity and state. We are exploring semi-supervised and hybrid machine-learning methods that leverage domain expert knowledge and expert guided training to provide reliable cell clustering and labeling in specific immunological settings.

This collaborative project currently has three components:

1. Clustering and labeling of single-cell omics data in disease settings where disease-related transcriptional changes such as activation and exhaustion complicate cell type clustering/labeling
2. Automated population gating and removal of unwanted variation in mass cytometry (CyTOF) data
3. Identification of novel cell populations/states in specific diseases using automated methods