Large-scale biology projects such as the sequencing of the human genome and gene expression surveys using RNA-seq, microarrays and other technologies have created a wealth of data for biologists. However, the challenge facing scientists is analyzing and even accessing these data to extract useful information pertaining to the system being studied. This course focuses on employing existing bioinformatic resources – mainly web-based programs and databases – to access the wealth of data to answer questions relevant to the average biologist, and is highly hands-on. Topics covered include multiple sequence alignments, phylogenetics, gene expression data analysis, and protein interaction networks, in two separate parts. The first part, Bioinformatic Methods I (this one), deals with databases, Blast, multiple sequence alignments, phylogenetics, selection analysis and metagenomics. The second part, Bioinformatic Methods II, covers motif searching, protein-protein interactions, structural bioinformatics, gene expression data analysis, and cis-element predictions. This pair of courses is useful to any student considering graduate school in the biological sciences, as well as students considering molecular medicine. Both provide an overview of the many different bioinformatic tools that are out there. These courses are based on one taught at the University of Toronto to upper-level undergraduates who have some understanding of basic molecular biology. If you're not familiar with this, something like https://learn.saylor.org/course/bio101 might be helpful. No programming is required for this course.
Lecture
Loading...
來自 University of Toronto 的課程
Bioinformatic Methods I
679 個評分
從本節課中
Selection Analysis
In this module we'll take a set of orthologous sequences from bacteria and use DataMonkey to analyze them for the presence of certain sites under positive, negative or neutral selection. Such an analysis can help understand the biology of a set of protein coding sequences by identifying residues that might be important for biological function (those residues under negative selection) or those that might be involved in response to external influences, such as drugs, pathogens or other factors (residues under positive selection).
與講師見面
Nicholas James ProvartProfessor
Cell & Systems Biology
[MUSIC]
Alright.
In this week's lab, we'll be doing selection analysis, which
attempts to identify if natural selection is acting on DNA sequences.
And, if in fact natural selection is acting, is
the selection in a positive direction or a negative direction?
And that means is the selection for or against a site.
And we're also trying to address how the
sequences are changing because of natural selection.
So there are three main flavors of natural selection.
Positive or Darwinian selection occurs when a beneficial
mutation arises in a population and increases in frequency.
So an example of this would be
antibiotic resistance rising in a bacterial population.
And that would quickly spread throughout the population.
So that bacteria, if
someone's being treated with an antibiotic,
all the bacteria would quickly acquire that mutation.
The other kind of natural selection is negative,
or purifying selection, which is the opposite of positive selection.
And this occurs when a detrimental mutation is selected out of a population.
A detrimental mutation is typically anything that
would inactivate a protein or an enzyme.
And these kinds of mutations quickly
become eliminated from a population.
The third flavor of natural
selection is balancing or diversifying selection.
And this is a kind of selection that favors
the maintenance of genetic variation at a locus.
You can imagine that this kind of selection might occur in a
population that is subject to or exists in different environmental
niches, where one allele might be
advantageous under one condition
and the a certain other allele might
be advantageous under another set of conditions.
How do we measure selection?
There are two commonly used measures: Tajima's D and the dN/dS test.
And there are many ways to quantify genetic variation.
Typically population geneticists focus on two metrics,
those being theta, which is based on the number of segregating or variable
sites in sample, and pi, which is based on the average number of differences
between all pairwise combinations of sequences.
This is also called pairwise nucleotide diversity.
So, both theta and pi measure variation.
Pi is more sensitive to the frequency of
genetic variants, while theta considers all variants equal.
Regardless of whether they're only found in one sequence, or half
of all of the sequences.
And pi takes these differences into account.
So this is just a cartoon of what I just said.
Basically we can have rare polymorphisms occurring just in one sequence.
We can have frequent polymorphisms occurring in
several of the sequences from a population.
And
Theta just counts the number of polymorphims
across the sequences, whereas pi measures the pairwise
difference between all possible pairwise comparisons of the sequences.
Theta is the number of
segregating sites and pi is the nucleotide diversity.
So natural selection acts directly on mutations that change
fitness, so this typically manifests at the protein level.
But genetic regions that are
surrounding a selected site are also influenced.
And the reason for this is that genetic regions surrounding
a site are dragged along, due to genetic hitchhiking.
And that, the size of the piece that hitchhikes along, is dependent on
the rate of recombination. So most of the genetic variation in
the surrounding regions is neutral.
So these are third position substitutions in the codon.
We'll talk about this in a little bit. That don't change the coding sequence.
But we can still see signs of selection due to linkage to to a selected
polymorphism. Tajima's D looks at theta and
pi and asks which one predominates. So if a gene is,
is neutrally evolving where there's no selection acting,
theta would be roughly equal to pi. In the case of positive selection
what happens is the genetic variation is swept from a population.
This is called a selective sweep. Very little genetic variation
will remain in a population after a beneficial mutation spreads through it.
But then new genetic variation will begin to accumulate and
this new genetic variation is accumulating in a homogeneous background,
and therefore most mutations will be at a low
frequency and in this case, theta is greater than pi.
And in the case of balancing selection, what is
happening here is that we're retaining alleles, genetic variation,
longer than would be expected. So, the allele is neither...
it doesn't sweep through the population or it's not eliminated from the population.
And since that's the case, those
alleles will arise to intermediate frequencies.
In this case, theta is less than pi.
So, this is just a cartoon again of what I was
just saying.
basically, if the genes are neutrally
evolving our theta is roughly equal to pi.
And our Tajima's D is approximately zero.
If there has been a selective sweep
due to positive selection, what we'll see is a lot of accumulation of
the rare variants in members of the population at a given locus.
And in this case theta is greater than pi and D
is less than zero. And if alleles have been retained for longer period
We'll actually see accumulation of variation in
multiple branches of the locus and theta in this
case is less than pi and D is greater than zero.
So, there are some caveats when using Tajima's D. You can see
that it could be a powerful statistic for detecting seleciton.
But it can be fooled by a couple of factors.
One of these is the population having
gone through a bottle neck, and this bottle neck
would very much resemble the case of positive selection.
So it's very difficult to distinguish between
a population having gone through a bottleneck versus positive selection.
And the other issue could be if
the locus is in a region of
of low recombination.
So basically there are some other tests that we need to apply,
to look at the population substructure with other tools
in order to be able to interpret Tajima's D properly.
The other kind of test we can do for natural
selection is to use the dN by dS ratio test.
So this is also known as the Ka by Ks ratio test.
And its perhaps the most widely used method for detecting the
pattern, the kind of natural selection
that is happening from nucleotide sequence data.
And its quite useful because we can actually infer selection
all the way down to the codon to ask the question
whether or not there are specific sites that are being selected for.
In order to be able to understand the dN by
dS test, we need to
understand non-synonymous vs synonymous nucleotide changes.
Now, non-synonymous substitutions result in a change in the protein sequence.
Synonymous substitutions change the DNA sequence, but not
the protein sequence
due to the degeneracy of the genetic code.
So if we look at our codons, The codons for amino acids,
We can see that for instance, in the case of leucine, there are actually six
codons that encode leucon leucine. TTA, TTG
TTT and so on. And typically it's the third position
(sometimes the second position) that can change without
changing the amino acid that's being coded for.
And typically the first position of codon
is the one that causes a non-synonymous change.
The dN by dS ratio test actually calculates the
ratio of the rate of non-synonymous substitutions, which is dN
(the number of non-synonymous substitutions per non-synonymous site, so
we look at all possible non-synonymous sites)
To the rate of synonymous substitutions, dS, which is
the number of synonymous
substitutions per possible synonymous site.
Now synonymous subsutitutions are not
exposed to strong selective pressures because
they don't result in a change to the protein sequence.
And typically selective pressure acts at the level
of the proteins, which are the manifestation of the phenotype
that arises from the DNA sequence,
and thus they accumulate at roughly a constant rate.
So, we can use this, the rate of
synonymous substitions as a baseline
to compare the substitutions that do change
the protein sequence, which are the non-synonymous substitutions.
So when we do a dN by a dS ratio
test, it's critical to have our DNA actually
broken up into codons and aligned by codons.
And we talked about this when we were doing
earlier labs. We can actually translate our sequence from DNA
into proteins and then do the alighnment at the protein level.
And then backtranslate into the DNA sequence.
And once we've got the DNA sequence in codon-based alignment, then we can
do the dN by dS ratio test with those data.
So how do we interpret the dN by dS ratio? In the case of a completely neutral
sequence it would be free to change with no constraints.
You would expect the dN to be the same as dS, so the
rate of non-synonymous substitutions to be the
same as the rate of synonymous substitutions.
In this case, the ratio of dN by dS would be equal to one.
And that's also
shown over here in this cartoon.
Ka by Ks would be equal to one.
When there are selective constraints on a sequence such that there's
negative selection you'd expect fewer
substitutions that changed the protein sequence.
So we would have a lower dN number, and therefore dN by dS would be less than one.
So that's shown over here, our Ka by Ks would be down
here, closer to zero.
And in the case of positive selection, we'd expect to see a greater proportion
of amino acid substitutions in the
population because they're being increased by positive selection.
So we would have a higher number of non-synonymous substitutions,
relative number of non-synonymous substitions,
therefore our dN by dS ratio would be greater
than one.
Now, it's important to keep in mind that most functional
genes, protein coding genes are under some level of selection constraint
so
the dN by dS ratios are typically well below one
for coding regions. Certain sites within those sequences,
however, can be under positive selection. So this figure over here to the right just
shows mouse-rat orthologous genes plotted in bins
according to their Ka by Ks ratio so Ka by
Ks ratio of one is over here, and it goes down to .05 on this side.
We see that by far the greatest
number of different values for Ka by Ks falls
into this bin here with a strongly small
kN/kS or dN/dS ratio.
So how can we use this information to study the
world around us? Well, basically evolution is an
arms race. And certainly a lot of the
drugs that we take are applying selective pressure onto
bacteria or viruses that infect us.
Or the plants that we like to eat.
And this is of concern. Certainly
changes to viral DNA sequences can happen overtime.
Changes to bacterial DNA sequences can happen overtime.
Simply through new mutations that arise.
And then these mutations, if they are beneficial, can spread to other
members of the population and could lead to outbreaks.
So for instance, this paper here actually
models the effect of anti-retroviral drugs.
On HIV and shows that certainly
some of the strains of HIV that are now resistant to
anti-retroviral drugs could pose a threat as
to cause a new epidemic which would be not
good. And so perhaps we can use this information
to actually understand the kind of selective pressure that's being applied
to DNA sequences. And to get a better understanding of how
we can perhaps modify drugs so that they
can still work against an evolved strain of HIV, for instance.
And here's an example with an HIV-1 protease using the SELECTON server of
the Tel Aviv University.
And basically the HIV-one protease is an essential enzyme for viral replication.
And several drugs target it. One of these is Ritonavir
and what the authors/creators of this tool, the SELECTON server,
did, Pupko and colleagues, they took 70
HIV-1 protease gene sequences from Ritonavir-treated HIV-1
patients extracted from a database, the Stanford HIV Drug Resistance Database.
And what they did was, they then fed those sequences, along with the PDB record (the
tertiary structure of the protease bound to Ritonavir) into the SELECTON server.
And what the SELECTON server
does is actually examine on a
residue by residue basis, whether or not the sequence of
the HIV protease is under positive selection or purifying selection.
Positively selected residues in this
case are shown in this orangey color.
And negatively selected residues or
those residues under purifying selection are in this purple color.
When we map that onto the 3D
structure of HIV-1 protease complexed with Ritonavir
we can see quite nicely actually that residues that are in the area
where the Ritonavir binds are actually being positively
selected for
So they're selected to change from something that they were
before, perhaps so that this Ritonavir can't bind as well
in the patients that are now resistant to Ritonavir.
So, by understanding which residues are being acted upon, you can actually start
to think about how evolution is acting and perhaps modify
the drug so that it could work in the cases where
it currently doesn't work.
So another site specific dN by dS analysis
comes from Professor Guttman who's a co-author on this course.
And his interests is in plant pathogen
interactions and plants are pretty interesting.
They do have actually have an immune system.
There is,
there is an innate immune system that basically
prevents most bacteria... the vast majority of bacteria actually
can't infect plants and this innate immune system is triggered
by molecular patterns, called MAMPs (Microbe-Associated Molecular Patterns).
These molecular patterns are found on structures like peptidoglycan cell
wall of bacteria, or the flagella of bacteria.
And these trigger an innate immunity.
This is commonly callose deposition,
an increase in a kind of cell wall component, and it makes it difficult for
those particular bacteria to enter the plant cell and attack it.
Sometimes the bacteria, however, have evolved mechanisms that
actually shut down this innate immunity. By
injecting components through a secretion
apparatus, effectors that actually block this innate immunity, and
plants then in turn have evolved mechanisms
to actually shut down the bacterial mechanisms.
And then the arms race continues.
These other bacteria have in turn acquired
new mechanisms for blocking this compensation mechanism here.
So it's really, it really is an evolutionary arms race.
And we can, we can start
to use sequences across many bacterial varieties, different strains of bacteria
and ask the question, are there any particular sites that are under selection,
positive selection, which might indicate that those sequences
are being forced to change by the action of
plant defense systems?
And there's a very nice paper published a couple years ago
by Honour McCann, one of Dave Guttman's grad students, and Professor Guttman,
where they took the genomes at three Peudomonas
syringae (a plant pathogen), and three Xanthomonas campestris
pathovars, scanned the 1322 orthologous core genes
for evidence of positive selection, and came up with several candidates.
So what we're looking at here
are four genes,
and the signatures of positive selection along those genes.
So the line, the red line here is a Ka by Ks ratio of one.
Anything above that line indicates a positively selected residue or region.
And anything below that line indicates a negatively-selected region,
or a region under purifying selection.
And as you can see as I mentioned before, for the rat/mouse orthologs, many of
the residues are indeed under purifying selection,
as you would expect for proteins that are
important for certain core functions: DNA processing protein,
an Elongation Factor Tu, the flagellan (you need
to have the flagellan to be able to move around).
But, however, there are some residues that do seem to be under positive selection.
And what McCann and her co-workers did is they
took those residues and actually asked the question, are
those residues being detected by the plant and are
they able to initiate this kind of innate immune response?
And if that's the case then you could imagine that those things would be
under positive selection to change, to
actually reduce the innate immune response.
And indeed, several of the peptides that
they chose from this positive selection analysis
were able to initiate a positive response in the innate immune assay that they used, which
is accumulation of callose, and which you can see here in these panels here.
So, this is a callose deposition that
we're seeing, it's measured
in the graph below.
So in today's lab what we'll be doing is examining one of the components that helps to
create this particular structure that allows bacteria to insert
a set of proteins that can actually subvert the innate immune response.
It's not clear whether or not parts of this should
be under positive selection but one suspects there might be and
we'll be examining that.
Now in order to be able to do this kind of analysis
we do need to first perform a phylogenetic analysis, and to
also of course have our sequences aligned as per their codons.
We also need some evolutionary models that
we touched on briefly in the phylogenetics lecture.
And here we will let a tool call Data Monkey help us choose the correct
model and we will also upload our phylogenetic information to this tool,
Data Monkey, in order to be able to do the positive selection analysis.
So, I hope you enjoyed today's lab, and really it's it's becoming a
bit of detective game here, trying to figure out what's going on in biology.
And the
availability of large numbers of sequences
is really allowing this detective game to become easier.
[MUSIC]
[SOUND] Hope you enjoyed the lab. [SOUND]
