Translational bioinformatics

bild på ett piller

Many currently used drugs are ineffective for treating complex diseases. This is likely due to a complex interplay between myriads of minor-effect genetic and epigenetic factors interacting in different combinations for each individual. By analyzing the governance of the immune system, we want to develop customized high-efficiency drug treatment without side effects.

Many common drugs work ineffectively for substantial sub-groups of patients. This results from the interplay between a multitude of small-effect genetic and epigenetic factors in complex diseases. In addition, treatment is started late when patients have received enough symptoms, which could take several years in common diseases as allergy, cancer and diabetes.

Recent decades have generated large measurement progress, which enables us to measure whole cell molecular state or imprint (also called omics) at a low price. One application of these imprints, called precision medicine, aims to, among other things, provide custom high-efficiency drug treatment without side effects. This could potentially also be utilised to use old drugs for new diseases. This new medicine requires new holistic analysis tools. Our research focuses on the analysis of the control of the immune system and there specifically:

1) Mapping the gene programs that control the healthy immune system. This can be used to understand the immunotherapy in the case of cancer, for example, and to understand which small errors in the gene programme that can cause various diseases.

2) Looking for synergistic bio markers by studying diseased individuals' footprint in the gene programme.

3) Implementing easy to use analytical tools based on 1) and 2) for general use in clinical medicine.

figur som visar generell pipe-line för prediktiv medicin

Our Project

The overarching goal of the project is the development of generally applicable system-based tools for dissecting complex diseases. To this end, we work to decipher healthy differentiation networks of immune defense and disease networks, as well as work on implementation of the advancements to the clinical community. In order to handle the computational complexity of genome-wide networks, in each sub-project, we simplify the problem into three coarse-graining levels, based on recently derived concepts, ranging from disease network modules, upstream transcription factors (TFs), to nonlinear modelling:
A) The most relevant genes for disease development tend to co-localise in so-called molecular disease modules. Disease modules shared across multiple T-cell diseases may have more relevance for pathogenesis of disease than disease specific modules.
B) The upstream TFs with most GRN targets in TH differentiation correspond to early regulators of T cell diseases, and can be used for predictive medicine.
C) Large-scale simulation based modelling (LASSIM, last author Gustafsson, PLoS Comput. Biol, 2nd revision) can perform mechanistic ODE expression modelling from the upstream TFs onto the entire genome in parallel using the national super-computer.

1) Identification of a validated multi-layered dynamical model
Understanding the control system of the normal immune defense would make it possible to perform accurate computer-based predictions of treatment effects that aims to manipulate the immune defense, which could drastically change the way drugs are developed. Examples could be immunotherapy that manipulates healthy T-cells to fight cancer, and immune-modulating treatment of allergy and auto-immune diseases. Moreover, as T-cells circulate through the whole body for the purpose of directing the immune defense, they can also serve as early detectors of disease.

2) Synergistically integrating multiple genomic layers for disease understanding
The background to this sub-project is that diseases are most often analyzed solely based on genetics or static epigenetics (gene expression and methylation). My recent results in multiple sclerosis (MS) and seasonal allergic rhinitis (SAR) shows that transcriptome dynamics derived from models of in vitro stimulation paired with genomics is of uttermost importance for identification of early biomarkers which predicts patient sub-types as they appear several years later [5, 16]. For example, in MS we defined an expression module from an in vitro model from one day of T-cell activation in patient vs controls, which remarkably could predict disease severity two years later by measuring corresponding CSF proteins, and also predicted response to treatment (Hellberg, et al Cell Reports).

3) Implementation of computation tools for precision medicine

figure som visar desease module


young woman in a wheelchair.

Severe MS predicted using machine learning

A combination of only 11 proteins can predict long-term disability outcomes in multiple sclerosis (MS) for different individuals. The proteins could be used to tailor treatments to the individual based on the expected severity of the disease.

Mika Gustafsson and David Martinez peeking into a server rack in the data center in Kärnhuset, NSC.

A step towards AI-based precision medicine

AI which finds patterns in complex biological data could eventually contribute to the development of individually tailored healthcare. Researchers have developed an AI-based method applicable to various medical and biological issues.

pregnant woman exercising in gym.

Why women with multiple sclerosis get better when pregnant

Women suffering from multiple sclerosis temporarily get much better when pregnant. Researchers have now identified the beneficial changes naturally occurring in the immune system during pregnancy. The findings can show the way to new treatments.

Group members