In this culminating project, you will deploy the tools and techniques that you've mastered over the course of the specialization. You'll work with a real data set to perform analyses and prepare a report of your findings.
The Results - Andrew Jaffe
Loading...
From the course by Johns Hopkins University
Genomic Data Science Capstone
27 ratings
From the lesson
Understand the Problem
The purpose of genomic data science is to answer fundamental questions in biology. Before starting on the data analysis process, the first step is always to understand the scientific question you are trying to answer. Don't forget to stay on top of the alignment task due in Week 4; it will take a long time to accomplish and shouldn't be put off.
Meet the Instructors
Jeff Leek, PhDAssociate Professor, Biostatistics
Bloomberg School of Public Health
Kasper Daniel Hansen, PhDAssistant Professor, Biostatistics and Genetic Medicine
Bloomberg School of Public Health
Hi, I'm Andrew Jaffe, an investigator at the Lieber Institute for
Brain Development and an assistant professor in the Mental Health Department
at the Bloomberg School of Public Health.
And today I'm going to be talking about some of our RNA sequencing
analysis results, exploring developmental regulation of human cortex transcription
and its clinical relevance at base resolution.
You can read more about our work at my website aejaffe.com or
follow me on Twitter at my handle @andrewejaffe.
In this project, we tried to better characterize gene expression changes
across development because we believe that transcriptome analysis can provide
fundamental insight into development and disease.
However, previous approaches have relied largely on existing gene adaptation.
This figure contains an overview of our study.
We first used RNA sequencing from the dorsolateral prefrontal cortex, or
DLPFC, in 36 samples across the lifespan.
We applied an approach called DER Finder, or Differentially Expressed Region Finder,
which performs differential expression analysis at single base resolution.
On the top right corner, you can see an example of differentially expressed region
identified by this algorithm.
In this region, you can see fetal samples have high expression, infant samples have
medium expression, and samples after infinite have low expression.
This region overlaps SOX11.
This approach works by identifying as single base resolution differences in
expression merges adjacent bases that show differential expression signal.
And then perform the permutation test to assess the statistical significance of
the identified differentially expressed regions.
Once each differentially expressed region, or DER, is ranked by significance,
we than annotate these to the known gene annotation.
What we did was we performed this approach in those 36 samples, and
then replicated the differentially expression regions we identified in
an independent set of 36 samples.
We, therefore, focused our attention on the set of about 50,000 differentially
expressed regions that were genome-wide significant and independently replicated.
And took these differentially expressed regions and
characterized them in the context of other data sets and clinical gene sets.
One of our key findings was identifying widespread developmental regulation
outside of annotated gene sequences.
We found that about 40% of these differentially expressed
regions were annotated to intronic or
intergenic sequence regardless of which gene annotation we compared.
For validation we took independent samples, three infant and
three adult samples, separated the RNA into nuclear and cytosolic fractions.
And confirmed that the changes we saw in total RNA were also present
in the cytosol, including the intronic and intergenic annotated sequence,
suggesting that these sequences were exported out of the nucleus.
To better ascertain the clinical relevance of these regions,
we took genome wide association study linkage to equilibrium blocks
from many common diseases in the central nervous system like Schizophrenia,
Alzheimer's disease, and Parkinson's.
And asked how many of these GWAS regions overlapped
differentially expressed regions in brain development.
We found statistically significant overlap between the 108 schizophrenia GWAS loci
and these differentially expressed regions, both when you consider
all regions and when you stratify it by whether the region was exonic or intronic.
We also found significant association between regions identified as
Alzheimer's disease and Parkinson's disease.
Interestingly, the regions we found that overlap the schizophrenia GWAS loci
were most highly expressed in fetal life
in line with the etiology supported by the neurodevelopmental hypothesis.
Conversely, the regions identified in the Alzheimer's disease and
Parkinson's disease loci most highly expressed in later life
in line with the etiology of those disorders.
As a negative control, we explored the overlap between Type 2 diabetes, a non-CNS
disorder and our brain associated changes and found no statistically significant
overlap, suggesting that these might play a role in disease risk in the brain.
We then took these 50,000 regions and asked how they looked in
other publicly available data sets like BrainSpan project.
This was a similarly sized sample size as our study, looking at about
40 individuals across the lifespan, but included many more brain regions.
There are 11 neocortical regions to the left of the vertical line and
5 non-neocortical regions to the right of the vertical line.
And you can see that at these regions, regardless of what brain region you look
at, the majority of the signal is about age and not brain region, suggesting
that they lack regional specificity and see more about development.
We then were able to show using another publicly available data set, in mouse,
that by lifting over the coordinates of our regions to the mouse genome and
quantifying the expression and their samples.
That many of these differentially expressed regions were conserved in
the mouse, including intronic and intergenic regions.
We lastly were able to show that in other human samples outside of brain,
we found evidence for expression and also regulation.
For example, if you look at the differentially expressed regions
that includes exonic sequence,
we found that our fetal samples look very similar to stem cells.
And that the first principal component shown on the left can separate
the postnatal brain from the other samples.
We lastly were able to show that fetal brain had the largest fraction of
the genome expressed by tabulating contiguous
regions that were expressed above a particular threshold.
As you can see here, we found that about 4% of the genome was expressed in fetal
samples compared to about 3.5% in infant life and about 3% later in life.
We believe many of these expression changes arose due to a shifting neuronal
phenotype that we're able to quantify using the DNA methylation data.
Using existing algorithms that can take homogeneous tissue like brain and
classify the relative proportion of different cell types, we were able to show
an expected rise of non neuronal cells occurring in postnatal life.
And interestingly, a loss of a progenitor-like cell that was largely
present only in fetal samples in the bottom left corner.
We're able to take these estimates of the relative portion of these
neural progenitor-like cells and
correlate them with the expression of our differentially expressed regions.
And in the bottom-right corner,
you can see that the majority of our differentially expressed regions were
statistically significantly associated with composition,
suggesting that many of our changes arise from the shifting neuronal phenotype.
This work was made possible through collaboration at the Lieber Institute for
Brain Development and the Johns Hopkins Bloomberg School of Public Health.
You can read more about our analysis and
results in the Nature Neuroscience paper that came out in December of 2014.
Funding was provided by the Lieber Institute for Brain Development and
an R21 grant to myself.
Thank you very much, and I hope you enjoyed my talk.
Explore our Catalog
Join for free and get personalized recommendations, updates and offers.
Coursera provides universal access to the world’s best education,
partnering with top universities and organizations to offer courses online.
© 2018 Coursera Inc. All rights reserved.
