Does circulating hepatocyte growth factor (HGF) have a causal effect on cardiovascular disease and type 2 diabetes?
Responsible: Christoph Nowak
Introduction
Hepatocyte growth factor (HGF) is a ubiquitous paracrine molecule that acts on many organs and tissues. Signalling via HGF has important functions in growth and angiogenesis, and circulating HGF levels are associated with health conditions including cancer, inflammation and metabolic disorders. There is observational and animal model evidence that raised HGF play an important role in the compensatory response against cardiovascular disease, type 2 diabetes (T2D) and insulin resistance. For instance, in mice fed with a high fat diet, cardiac-specific overexpression of HGF protects against obesity and insulin resistance, whilst pancreatic beta cell-specific knockout leads to the development of T2D. In cross-sectional epidemiologic studies, raised circulating levels of HGF are generally associated with adverse metabolic and cardiovascular traits such as obesity, increased systolic blood pressure, insulin resistance and asymptotic cardiovascular pathology. In longitudinal studies, raised baseline HGF has been associated with an increased risk of T2D and insulin resistance (e.g. PMID 28572400, 28273932, 20980460, 26892517, 26420861, 23024263, 22788978, 22318499).
An important question arising from these correlational studies is whether HGF has a causal effect on cardiometabolic impairment - in other words: if targeting HGF levels by therapeutic intervention might prevent or treat chronic cardiometabolic disease. Causality cannot be assessed in observational studies because of potential bias from confounding and reverse causation. For instance, higher HGF levels are associated with insulin resistance - this could be because (i) HGF leads to insulin resistance; (ii) insulin resistance causes raised HGF (reverse causation); or (iii) because of related (un)measured factors, so-called confounders.
In this project, we will use a recent method called Mendelian Randomization analysis that uses genetics to assess causality in observational associations. At conception, parental alleles are essentially inherited at random. Mendelian Randomization uses alleles that are associated with a risk factor (e.g. HGF) to test for independent effects on an outcome (e.g. T2D). Because the "quasi-randomization" to different levels of the risk factor happens before birth and thus before environmental and other confounders occur, risk factor-associated alleles can be used as instruments to test for causation. Causal estimates from Mendelian Randomization analysis are not equivalent to experimental manipulation, but can give a reasonable picture of the expected long-term causal effects of specific exposures. We will use Mendelian Randomization analysis with existing summary-level genetics data to assess whether circulating HGF has a causal effect on cardiovascular and diabetes-related pathology.
An important question arising from these correlational studies is whether HGF has a causal effect on cardiometabolic impairment - in other words: if targeting HGF levels by therapeutic intervention might prevent or treat chronic cardiometabolic disease. Causality cannot be assessed in observational studies because of potential bias from confounding and reverse causation. For instance, higher HGF levels are associated with insulin resistance - this could be because (i) HGF leads to insulin resistance; (ii) insulin resistance causes raised HGF (reverse causation); or (iii) because of related (un)measured factors, so-called confounders.
In this project, we will use a recent method called Mendelian Randomization analysis that uses genetics to assess causality in observational associations. At conception, parental alleles are essentially inherited at random. Mendelian Randomization uses alleles that are associated with a risk factor (e.g. HGF) to test for independent effects on an outcome (e.g. T2D). Because the "quasi-randomization" to different levels of the risk factor happens before birth and thus before environmental and other confounders occur, risk factor-associated alleles can be used as instruments to test for causation. Causal estimates from Mendelian Randomization analysis are not equivalent to experimental manipulation, but can give a reasonable picture of the expected long-term causal effects of specific exposures. We will use Mendelian Randomization analysis with existing summary-level genetics data to assess whether circulating HGF has a causal effect on cardiovascular and diabetes-related pathology.
Project details
Inst f neurobiologi, vårdvetenskap och samhälle (NVS) | |
- | |
Other | |
2 | |
English | |
All data have already been collected and only need to be analyzed | |
Ethical permit is required and exists |
Supervisor/Contact
Christoph Nowak
0739806535
christoph.nowak@ki.se
Contact 2
Aims
We will use summary-level two-sample Mendelian Randomization to assess if
(A) HGF affects the risk of coronary heart disease, stroke, and raised blood pressure (Student 1)
(B) HGF affects the risk of type 2 diabetes, obesity and other metabolic traits (Student 2).
(A) HGF affects the risk of coronary heart disease, stroke, and raised blood pressure (Student 1)
(B) HGF affects the risk of type 2 diabetes, obesity and other metabolic traits (Student 2).
Design
Mendelian randomization study using summary-level data
Material and methods
We will use the largest available summary association datasets for genetic effects on cardiovascular phenotypes (coronary heart disease, blood pressure, stroke) and a group of metabolic phenotypes (T2D, insulin resistance, obesity and others). We will construct a genetic instrument for circulating HGF levels based on published genetic associations and public data repositories. In order to estimate the causal effect of HGF on the outcomes, we will use several Mendelian Randomization methods, all of which can be implemented with existing software (and simple R programming). We will also investigate if major assumptions of Mendelian Randomization are violated and adjust for bias whenever possible. No particular experience in bioinformatics, programming or genetics is needed - as long as you are interested and motivated, we'll take it from there!
Project time schedule
In the first two weeks, we will schedule introductory meetings to plan the project(s) based on your existing skills and interest. One student will focus on the cardiovascular outcomes (Aim A), the other student will work on the diabetes-related outcomes (Aim B). You can either work collaboratively or completely separately, provided you stick to your aim and prepare your thesis independently. During the project, we will have meetings whenever needed and in the final phase, we will have a practice session for your defence.
Following successful defence and if you are interested, we can aim for publication with yourselves as first authors - a good way to experience the travails of (trying for...) publication.
Following successful defence and if you are interested, we can aim for publication with yourselves as first authors - a good way to experience the travails of (trying for...) publication.
Backup plan
Data and methods have been collected.
Teaching/Supervision activities
Depending on your background, I will give you an introduction and practical supervision in molecular epidemiology, Mendelian Randomization and other epidemiologic study designs, using genetics in large cohorts, basic R programming ... Your call! We will use R programming language for analyses - but no worries if you've never used R or any other programming language before. I am not usually physically based on campus (rather, in Uppsala). We will have in-person meetings in the beginning and before handing in the thesis and whenever needed. Otherwise: skype, phone and email.
Resources
All the data needed for the project are available and have ethical permission. With regard to analysis, software etc. - we'll take it from whatever experience you have.
Miscellaneous
Learn:
- Intro to programming in R
- Using "big data" for clinical insights
- Principles of molecular epidemiology - particularly using genetics to tease out causality from correlation
- HGF and its role as a potential drug target/biomarker
- General idea of academic research: designing and writing a study, co-author and reviewer feedback, peer-reviewed publication...
Needed:
- Motivation
- Computer and somewhere to work
Please note: I am not usually physically based on campus (rather, in Uppsala). We will schedule in-person meetings in the beginning and just before handing in the thesis and whenever needed. Otherwise - skype, phone and email will be our primary way of communication. I cannot offer you an office space. On the upside, provided you have access to a computer, check your emails regularly and are aware of our deadlines, you can work on the project in your own time.
- Intro to programming in R
- Using "big data" for clinical insights
- Principles of molecular epidemiology - particularly using genetics to tease out causality from correlation
- HGF and its role as a potential drug target/biomarker
- General idea of academic research: designing and writing a study, co-author and reviewer feedback, peer-reviewed publication...
Needed:
- Motivation
- Computer and somewhere to work
Please note: I am not usually physically based on campus (rather, in Uppsala). We will schedule in-person meetings in the beginning and just before handing in the thesis and whenever needed. Otherwise - skype, phone and email will be our primary way of communication. I cannot offer you an office space. On the upside, provided you have access to a computer, check your emails regularly and are aware of our deadlines, you can work on the project in your own time.