Klinik für Psychische Gesundheit

The lab for Functional Genomics in Psychiatry leverages cutting edge statistical and deep learning approaches, pluripotent stem cell based models, functional genomics and patient derived medical record data to empower personalized medicine in practice. In particular, we i.) develop and apply novel computational strategies to decode the personalized functional consequences of common disease associated genetic variants in individual patients, utilizing multidimensional –omics data types of large cohorts (e.g. UKBiobank) and machine learning, ii.) utilize standardized human induced pluripotent stem cell derived neural cell populations as a versatile platform to model the molecular and cellular polygenic basis of these diseases and test the computational predictions, iii.) employ high throughput genomic profiling using a variety of assays including ChIP-Seq, ATAC-Seq, HiC/capture HiC, single cell RNA-Seq/ATAC-Seq & SNP arrays of in vitro derived and primary, post mortem cell populations iv.) develop and apply different types of high-throughput functional genomic screening technologies such as massively parallel reporter assays (MPRA), shRNA/sgRNA as well as Perturb-Seq to functionally characterize (combinations) of disease associated non-coding gene regulatory elements in different in vitro derived human cell types and v.) apply machine learning techniques to mine electronic health records in order to connect molecular insights to patient (endo-)phenotypes.In this context, we closely collaborate with multiple clinical groups and large established research cohorts for patient recruitment and clinical characterization to apply our aforementioned analysis pipeline to distinct patient strata defined based on clinical and genetic parameters such as polygenic risk, disease course or treatment response.

It is our main goal to rapidly operationalize the molecular and neurobiological insights emerging from our experiments in order to effectively implement a workflow starting from the bedside to the bench and back.

At present, the main efforts of my lab are focused on (1-3) dissecting the polygenic basis of schizophrenia; (4) understanding the molecular genetic contributions to treatment resistance in schizophrenia; (5) developing novel methods for patient stratification and personalized medicine based on genetic and gene expression data and (6) pinpointing neural cell types and developmental stages most vulnerable to psychiatric diseases:

(1) Molecular and endophenotypic analysis of a large iPSC disease model library reveals polygenically driven alterations in key neuroplasticity related genes and cellular traits.

We have assembled a large cohort of pluripotent stem cells (Fig. 1) from more than 100 distinct donors suffering from schizophrenia or bipolar disorder as well as healthy individuals. Standardized 2D in vitro differentiation of this disease model library into cortical excitatory neurons followed by deep molecular and endophenotypic profiling revealed polygenically driven differences in key synapse related genes and pathways (Fig. 2). Moreover, these deficits translated in subtle, but highly reproducible differences in multiple cellular endophenotypes, providing new insights into genetically driven mechanisms that contribute to the emergence of SCZ. Moreover, we provide a roadmap of eQTL and chromatin QTL loci in iPSC derived neurons that are conserved between these in vitro models and the adult human brain. Lastly, we validate parts of these findings in in a large cohort of post mortem brain material from the PFC profiled by scRNA-Seq and scATAC-seq.

(2) Dissecting the genetic basis of neuronal hyperexcitability of iPSC derived neurons from schizophrenia patients

Here, we aim to deconstruct changes in molecular and cellular endophenotypes in human neuronal cells driven by polygenic risk load. To that end, we generated two cohorts of induced pluripotent stem cells from SCZ patients with polygenic diseases architecture, followed by their standardized in vitro neuronal differentiation. Deep molecular and cellular endophenotyping of the resulting mature cortical neuronal cultures revealed a subtle, but consistent alteration in the excitability patterns of the SCZ derived neurons across cohorts compared to healthy donors. Importantly, these physiological changes were accompanied by the consistent de-regulation of a class of genes previously not implicated in SCZ.

(3) Dissecting the polygenic basis of schizophrenia using functional genomics and patient specific induced pluripotent stem cell models

GWA studies were highly successful in linking hundreds to thousands of individual genetic variants to diseases such as schizophrenia or bipolar disorder. However, due to linkeage disequilibrium and the fact that most associated genetic variants reside outside the coding region of genes, it is unknown i.) which of the statistically associated genetic variants do indeed carry out a molecular function, ii.) which genes do they regulate and iii.) in what manner? Here, we provide a first draft of possible answers to these questions and develop a massively parallel variant annotation pipeline (MVAP) in SCZ relevant cell types (Fig. 3). First, we sought to identify those SCZ associated genetic variants that exert a molecular function in human neurons using massively parallel reporter assays. Subsequently, we link these variants functionally to their target genes in disease relevant cell types using chromatin conformation capture by HiC and massively parallel perturbation screening based on epigenome editing technology combined with single-cell RNA-Seq. These experiments provide a first roadmap of functional annotation of disease associated variants and allow to translate parts of previous GWAS findings into changes in molecular endophenotypes.

(4) Modular modeling of complex neuronal networks using iPSCs identifies disturbed excitation/inhibition balance as one neurobiological mechanism contributing to schizophrenia

(5) Identification of complex disease associated pathways using imputed cell type specific gene expression patterns across large patient cohorts

Genome wide association studies have unearthed a wealth of genetic associations across many common complex diseases. However, translating these associations into biological mechanisms contributing to disease etiology has been challenging due to the highly polygenic nature of complex diseases. This complexity is further exacerbated by the fact that all of the associated common variants carry very small effect sizes and operate in a highly cell type specific fashion with mostly unknown target genes. Moreover, most patients suffering from these highly polygenic diseases harbor different combinations of disease promoting common alleles, making the identification of altered biological process highly challenging.

Here, we hypothesize that the effects of disease associated genetic variants converge onto distinct cell type specific molecular pathways within subgroups of patients, contributing to changes in molecular and cellular endophenotypes.

To test this hypothesis, we develop a new computational method based on a machine learning approach to operationalize individual genotype level data for the discovery of altered biological processes within subgroups of patients and cell types. We show that this method offers significant advantages over previous approaches by i.) allowing for the integration of prior biological information within a rigorous machine learning framework to predict cell type and individual specific gene expression from genotype, ii.) aggregate power by integrating the small effects of many common genetic variants on the gene expression and pathway level and iii.) allowing for a biological mechanism based stratification of individuals into distinct subgroups.