What the thesis ended up being was an exploration on the failures of heurestic methods in doing this kind of work. Not only for this specific
task and data, but it raises skepticism regarding the work done at JCVI as well. The results cannnot be trusted without validation, and
there needs to be an accurate, unified method that work on all levels of data (blood/LCL, high depth/low depth, etc.). Personally, this work
proved to me the need for mathematical rigour every step of the way; agreed-upon conventions and arbitrary measures should always be taken
with a heavy grain of salt, and only really useful as a starting point, not as an end in itself, especially when it concerns heurestic methods.
Below is the introduction and link to the full paper:
"Over the past several years interest and active research in viral metagenomics has been steadily
increasing, and not without due cause; in any given environmental habitat, including the human microbiome,
at least 1031 viral particles exit at any point in time. Humans in particular carry a 10-100
fold viral particles to human cells, including the gastrointestinal tract where extensive research on
the bacterial microbiome has already been under way for more than a decade [64]. This may in fact
be less innocuous that it sounds; firstly due to immune systems’ response to viral infections, and
its role in regulating transkingdom interactions within a microbiome through immunomodulation.
Viral immunomodulation has been implicated as a significant contributing factor in chronic autoimmune
diseases (lupus, rheumatoid arthritis, multiple sclerosis and asthma and several more). It has
also been shown to confer beneficial effects in the form of protection against bacterial infections
in the gastrointestinal tract (as shown in mice) and a heightened antiviral immunity and superior
response such as the case in young human adults carrying herpesvirus[9]. On an individual level,
there are also widely disparate variations, even among immediate relatives [32] [42] .Furthermore,
the heavily underestimated diversity of viruses, the number of novel viruses being continuously
discovered within the human body, as well as the shortcomings of fully sequenced viral genomes
with ”Hallmark” universal genes remain open issues in profiling the human virome, and a challenge
in taxonomy in general [48]. Efforts have been made to probe DNA viral diversity within humans
however, and a recent study found 3,761 novel DNA viral sequences circulating as cell-free circular
DNA in 1,351 blood samples [22].
The aforementioned not only highlights the significance and prevalence of the human virome
within the body, but also the difficulty in fully characterizing what would be a typical virome
of a healthy individual. Towards that end, a recent metagenomics study (JCVI) with a focus on
homology-dependent methods rather than novel discovery, explored the unmapped sequences of
8,420 deep sequenced individuals [53] in an attempt to profile a normal blood phenotype. Sequences
were obtained from whole genome of sequencing of blood, and the study found 94 viruses (19
human DNA viruses with the rest being concluded as reagent contaminations) in 42% of all -disease
free- individuals. The researchers also found significant contributions of age, sex and ancestry in
viral infections [34].
Aim:Our goal is to apply similar methods of the latter study on the freely available 1000 Genomes[4]
to capture a profile of the human virome. We futher aim to conduct Genome-Wide Analysis (GWA)
on the most prevalent viruses found to test for associated loci with viral abundances. There are
several marked differences in contrast with the methodology of JCVI study with this study; firstly,
samples from the 1000 Genomes Project are obtained from cell cultures (lymphoblastoid cell lines)
and are of low-coverage (4x – 8x).Furthermore, because the samples are publicly available, health
related records including age and gender of the indivudal was not recorded. Individuals sequenced
in the 1000 Genomes Project self-reported as ”healthy” however. These factors combined would
lead us to expect significant differences between the results of this study and the previous one, in
particular lower viral abundances, but we hope to nonetheless ascertain the validty of the method
within a low-computational setting. To add another layer of complexity and to exclude atleast one
source of variance, the same methods were applied on low-coverage (11x) whole-exome sequence
data of 2,000 Danish individuals.
Grab the full project report here: https://github.com/Adhamkmopp/Thesis