Overcoming the challenges of designing efficient and specific CRISPR gRNAs

December 3, 2019

Hello, and welcome to this
integrated DNA technologies webinar– overcoming the
challenges of designing efficient and
specific CRISPR gRNAs. My name is Sean
McCall, and I will be serving as moderator
for today’s presentation. Today’s presentation will be
given by Dr. Matthew McNeil. Matt is a staff scientist in
the bioinformatics group at IDT. Matt earned his
PhD in neuroscience from the University
of Iowa and worked as a post-doctoral researcher
at the University of Illinois. at IDT, Matt is focused
on development of our NGS and CRISPR product lines. Matt’s presentation should
last about 30 minutes. And following the
presentation, he will answer as many questions
as possible from attendees. The question and
answer session will be conducted by Justin Barr,
senior product manager, functional genomics at IDT. As attendees, you
have been muted, but we encourage you to ask
questions or make comments at any time during or
after the presentation by typing into the questions
box located in the Go To Webinar control panel. Also in case you need
to leave earlier or want to revisit this webinar, we
are recording the presentation and will make the
link to the recording available on our website a few
days after the presentation. We will also post the
recorded presentation on our YouTube
and Vimeo channels and post the slides on
our slide share site. You will receive links to
these in a follow up email. So now let me hand it over
to Matt for his presentation. Thanks very much, Sean. Today we’ll be talking about
the implementation and display of our CRISPR-Cas9
design tool available through the integrated
DNA technologies website. In this presentation,
we will first review the CRISPR-Cas9
genome editing, challenges to identifying potent
guide sequences that have high on target activity
and low off target activity. In the final
section of the talk, we’ll review the Alt-R
guide design tool. We’ll review its features
and provide a demonstration of the tool itself. The CRISPR-Cas9
design tool is focused around designing
guides associated with the streptococcus
pyogenes Cas9, or S.p. Cas9. S.p. Cas9 is an RNA-guided
endonuclease that uses a 20 nucleotide spacer
sequence known as a guide RNA, or gRNA. The S.p. Cas9 PAM site is a
recognition domain directly adjacent to the guide
sequence found in the genome and is an NGG. In the figure below,
we show a two part system, which is native, coming
from streptococcus pyogenes. In this system, we have
the crRNA, shown in green, the 20 nucleotide
spacer sequence, shown in the thick bar,
and an unvariable region shown in the thin bar. It is annealed directly
to the tracrRNA shown in the orange bar, which
is also an invariant sequence. When complexed with the S.p. Cas9 protein, it forms a
rival nuclear protein complex. Again, S.p. Cas9 is shown in the
gray outline behind here. When all three elements
are complexed together, they can recognize
their sequence within the genomic
DNA directly adjacted to, in the three prime
end, the NGG PAM sequence, and three bases five prime
of the three prime end will create a double stranded
break using the RuyC and HNH domains within the protein. Intracellular
machinery can repair the double stranded
break through one of two typical pathways– non-homologous end
joining will create a series of insertions
and deletions at the target site for
each individual repair. Homology directed repair
can use a double stranded or single stranded template
to repair perfectly or nearly perfectly at the site
of the double stranded break. You can direct homology
directed repair using pre-designed single
stranded or double stranded [INAUDIBLE],, though
we won’t review that in this particular
presentation. Our presentation
today will focus on designing effective
and potent guides to target your S.p. Cas9 to a particular
target site and introduce the double stranded break. Guide RNA sequences
have varying activity at sites along the genome. Here we did an analysis of
guide activity and potency at a series of target
sites along the HPRT1 gene. In this graph, we show
the percent editing thing as observed through T7E1
cleavage assay on the y-axis and each of the individual
guide RNA sequence locations shown on the x-axis. Here in the x-axis bar, we
show representative names of some of the guide
sequences that we examined. As you can observe
from the graph, we saw between 3%
and 63% activity for varying guides along
this target sequence. As you can see, choosing
a guide with high potency is more likely to be
one that will give you good results in your experiment. Part of the goals
of this tool will be to help select guides that
are going to be highly potent. Off target activity is also a
problem for guide sequences. Prior literature and our
own internal analysis has demonstrated that
CRISPR-Cas9 can tolerate up to four mismatches between the
guide sequence and its target and allow up to one base gap. Off target activity
can cause inaccurate phenotypic results, depletion of
the CRISPR-Cas9 enzyme, or even cell death. We analyze the likelihood that
an individual guide sequence would target another
gene or coding sequence within the human genome. For this analysis, we
did an in silico design of 620,433 guides
against 19,222 genes and analyzed their alignments
against the human genome using the allowable mismatch
and gap penalties that we described
in the prior slide. We found that on average 6.9%
of individual guides’ off target alignments landed within coding
regions of the target genes. We’ve also found that
3.95% of the human genome is covered by regions deemed
to be exonic by NCDI’s most recent GFF file. Dividing 6.9 by 3.95, we find
that an individual guide’s off target alignment
is 1.7 times more likely to occur at an exons
than elsewhere in the genome. Here in the plot to the right,
we show the amount of variation that can occur across guides. On the y-axis, we show the
frequency, and on the x-axis, we show the off
target percentage in exons of the
individual alignments. You can observe that
anywhere between less than 5% to upwards of 20% of the
off target alignments can end up within an exon,
given that exons that share homology and thus would
likely be similarly targeted by an individual guide in it’s
on and off target sequences are also ones that are more
likely to share function because similarity of sequence
might imply similarity of function. You could be
enriching for guides that would have an off
target sequence that could land in a gene that would
share function and thus cause an inaccurate phenotypic result. Our conclusion
from this analysis is that off target activity is
likely to occur in another gene and is something that
should be paid attention to when creating novel designs. Thus we reviewed that there are
two important characteristics of designing guides– high on target activity and
low off target activity. We want our design tool
to incorporate information about both characteristics. Our designed, thus, implements
both characteristics and allows customers
to access these results through a few different methods. We have a pre-designed library
of guides that is quickly accessible through our web page,
a set of custom guide designs– excuse me, we have a
custom guide designer that allows us to first
provide a FASTA sequence and will create a novel
CRISPR-Cas9 design against that FASTA sequence. Alternatively, we will
allow customers provide us with guides that
have been designed either through our tool manually
or through another base tool, and then we will check the on
target and off target activity for that particular guide. We’ll review more of this
later on in the presentation. Our on target
model was developed using a support vector
machine approach, which is a type of machine learning. We implement it as
a classifier method to predict whether a guide
will work or not work, and we’ll go into more as to
how we define those terms later. We showed our general approach
for how we created our model using this diagram. In our diagram, we show that
we first manually design training data, we map
features or sequence based information against each
of those individual guides, we select which
of those features are most predictive of
the guide performance, and we run 10-fold
cross-validation in which we use 90% of our data
to generate a model comparing against the other 10%. And we do that with different
subsets of 90% and 10% until we find the
model that we find is best predictive
of that other 10%. Selecting that
model, we manually designed another set of guides,
compared our model against it, and do final optimizations. The model that comes out
of the final optimizations was used to design novel
guides, and we used those guides in real world experiments to
validate the predictability of the model itself. In the next section,
we’ll review how we developed our training data. We designed the 560 guides
that walk five gene targets– four from human
and one from rat. For these guides
that were designed against the human genome, we
synthesized them as crRNAs and annealed them to tracrRNA. The annealed product
was lipofected into HEK-293 cells, which were
constitutively expressing Cas9. Rat guide RNAs were similarly
synthesized as crRNAs and annealed to tracr sequences. The crRNA plus
the tracr sequence was complexed with
the Cas9 enzyme to form a ribonuclear protein. The RNP was lipofected
into rat two cells. In all cases, DNA was extracted
after 48 hours and target regions where PCR amplified. We analyzed the performance
of any individual guide RNA using the T7E1 assay. The overall performance of the
guide set of the 560 guides showed a median
performance of 40%. We deemed the guides that
worked less well than 40% were those that did
not work, and ones that worked greater than 40% of
efficacy were those that did. We mapped features against
each of the 560 guides. We used n-gram
sequence composition. N-gram sequence composition
would be 2 and 3 based motifs that
occur at a given frequency throughout the each
individual guide sequence. We looked at position specific
based composition, the A, C, T, or G bases that are present at
anywhere from base 1 to base 20 along the guide. The probability of
self hybridisation and the hybridization
parameters between the crRNA and the tracrRNA. After we selected
those features which were most informative
and aggregate, we were left with a total
set of 1,432 features, which went into our model. After doing our 10-fold
cross-validation and creating our final model,
our final model at this step, we compared the
model’s performance against a novel set of guides. We tested the model
with 215 guides, which had a median overall
performance of 63% editing. And if you remember,
we had our prior not work category had
a dividing line at 40%. That means that our median
performance is now higher, and we would anticipate
that more of our guides would work in the
testing set than those in the initial training set. We found that we had a precision
of 94%, a recall of 51%, and a specificity of 75%. The precision is calculated as
the number of true positives over the total set of true
positives plus false positives, which is our rate of being right
in those that we call working. Our recall is our number of true
positive events over those that are true positive
plus false negative, and that’s our rate of
finding the ones that did work within our testing set. Our specificity is the
number of true negatives, or true negatives plus
false positives, of 75%. And that’s our rate of
correctly not selecting those that did not work. We used the model after
some final optimization to pick guides that
we then would validate in experimental conditions. We designed three guides
against each of 32 genes for a total of 96 guides, and we
analyzed how many of the guides worked per gene. In 24 of the 32 genes, we found
that 33 out of three guides worked. In six of 32 genes, we found
that two out of three guides worked. In two genes, one out
of three guides worked. In total, this means
that if you choose to design three guides
against a target gene, we would anticipate that most of
the time all three would work. But this would allow
you to hedge your bets to find in general at least one
that should work in your case. We also considered
off target activity. To score off target activity
for an individual guide, we designed a position
specific scoring matrix. PSSM represents values that
quantify the relative value of a mismatch or a variant
at each individual position along a guide sequence for
each off target alignment. We used it to calculate a weight
for each off target alignment. An example of how to use
the PSSM would be here. This is a guide
sequence that’s oriented from five prime on the left
to three prime on the right. We know from prior
experimental results that the three prime end is
more important for overall guide activity than the
five prime end. In the sequence below, we
can examine an off target alignment. The difference between
these two sequences is highlighted by turning
the individual bases red to show the mismatch. At position 20, where we
know that the influence of an individual
mismatch is less important than at
the three prime end, we might score an A
to T mismatch as one. At position of
13, getting closer to the end of
greater importance, we might score a T to
A mismatch as three. We developed our PSSM by
pulling sequence and performance data from published and
internally generated sources. We built linear regressions per
mismatched type and position to predict overall performance
of an individual guide. And we constructed PSSMs per
sequence position accounting for the differences in
each type of mismatch. An example of the full
table might look something like this, where a
guide sequences oriented from position 20 to position 1
and the NGG PAM sequence would be beyond the end of the table,
and each of the variant types would be across the
x-axis at the top. As we talked about before, green
would be a particular mismatch that is of lesser value
and up to the heat map on to red, which would be a
mismatch of a greater value. Some mismatches would
have overall less value than others, which would
have overall greater value. We used our PSSM by first
aligning each guide sequence against the genome of interest. Offh target alignments
were then scored using the PSSM of values. And we aggregated
those off-target scores to overall represent
the guides the guides’ off-target specificity. We represent these
values on our website on a scale from 0 to 100. For on target sequences, guides
that worked had a score of 100. Those that did not work
had a score of zero. The dividing line would
be between the case of work versus not
work and is divided at the line of about 50– or excuse me, exactly 50. The off-target sequences,
those with high specificity, would also have a
high score of 100. And those that have
poor specificity would have a score of 0, and
then any individual guide can have a score ranked
anywhere in between. We validated our PSSM
method by designing for guides against the AAVS1. We co-nucleofected
HEK-293 cells, stably expressing Cas9 with
pre-annealed crRNAs containing any of the four guide sequences
and yielded to the tracrRNA. We co-nucleofected the annealed
product with the GUIDE-Seq tag as described in the
Tsal et al 2015 paper. After 48 hours, we extracted DNA
and constructed DNA libraries following the protocol
in Tsal et al. This involves amplifying off
of the tag sequence, which has been integrated
into edited target loci using nested primers. We sequenced the individual
molecules on an Illumina MSeq platform and aligned them
back against the human genome. Loci containing a large
number of reads piling up indicate that that
was an edited site. I first want to walk
you through the scores down here at the bottom. On target scores are listed
anywhere from 46 to 94 for each of the four
guide sequences. And off-target scores
are listed from 22 to 87. As a reminder, low off-target
scores and low on target scores are ones that would be less
well-performing and less desirable. For our positive
control sequence, we had a low on target
activity score of 46 and will be one that would
be predicted to not work. It also has a low
off-target score and is one that is predicted to
have many off-target alignments that are of higher value. Consistent with that, we
find that reads piling up at the off-target locations,
shown in the light blue, overwhelm the overall
total number of reads piling up across the genome
for the control sequence compared to those that are
aligning to the on target sequence. For each of the other
three guide sequences, we find that the vast
majority of the reads line to be on target
sequence as compared to any of the off-target loci. Looked at overall, we can
see that the low scores are associated with the positive
control sequence, which was poorly performing, compared
to the higher scores, which are present for the other
three guide sequences and are much better performing. Looked at the total
number of sites that were identified
through GUIDE-Sec, we find that there were a
total of 278 off-target sites for the positive
control and between 1 and 5 off-target sites,
which were identified for each of the other three. We implemented both the on
target and off-target models into our design tool. And before we go on the
demonstration of the tool itself, I want to review
some of the features available in the tool. Our tool has a customized on
target model for IDT’s system, and it’s available for
pre-designed libraries and off-target screening
for five species– human, mouse, rat,
zebrafish, and nematode. It provides rapid access
to pre-designed library for all these five
species, which can be looked up using genes
symbols, transcript accession numbers, or design IDs. Additionally,
customers can provide FASTA sequences, for which we
can provide custom Cas9 guide RNA designs and off-target
checking against any of these five species. Additionally, our
tool will report whether an individual
guide alligns against a region
with a known snip, that it can occur within
the genomic population. We also allow for rapid off
target screening and a checker for guides that have
been previously designed. This tool allows customers to
identify off-target locations that land within a gene exon
for any of the individual five species. The off-target alignments also
allow for up to one gap open, and we’ll find
alignments that are adjacent to a
non-canonical PAM, NAG. Finally, our tools allows
for easy visualization of the on and off
target activity scores so that customers
can quickly select guides that have the
highest combination of both. Our tool’s available on
three separate links, and we’ll review each
component of this tool in a live demonstration. You could find a tool either
through the links that were provided in the
slides, and those slides will be available
after the talk. Or you can find it
by going on Google and searching integrated
DNA technologies followed by CRISPR. By clicking on
that first link, we can scroll down the
page, passed a number of the different products
for which this is compatible, and select on any
of the three buttons which will take you
to those same links. Our tool is available
on our page. We can design against
custom sequences because an off-target check
does take a little bit of time. I’m going to go ahead
and start this now. I’ve timed this before. I ran it over the last few
days five or six times. And in general, it
was taking somewhere between a minute
20 and a minute 30 to run for this
particular sequence. So I’m going to let this
run in the background while we review other
parts of the tool. So this particular target
sequence, you can see here, is relatively large. All we have to do is enter that. The species that we’re
going to be comparing against for off-target
checking is Homo sapiens, though there are other
species that are available or you don’t have to run
an off-target analysis. And you can, alternatively,
put in a pre-designed design ID as an input. Here we’re using FASTA. Also available is
the ability to upload the entire sequence as
a file if you prefer not to copy and paste it in. We’ll go ahead and
start that running, and we’ll open up a new tab
and go to the web page again. And we can come back to
this particular sequence here in a little bit
once it’s finished. In the meantime, I want to show
you our pre-designed library. Here we’re going to be pulling
designs from Homo sapiens. We’re going to be looking up
through gene symbol, though, as I mentioned before,
accession number and design ID are alternate ways
of looking them up. And we can ask
the tool to return a specific number of guides for
the genes that are of interest. You can alternately upload a
file of gene IDs into the tool. Here we’re going
to have a search, and you’ll note that this
comes back very fast. We can select which of the two
genes that we want to examine. Individual guides are shown
against these CDS regions, which are displayed in
green, and individual guides can be selected either
through the visualization here at the top,
which will provide a checkbox next to an
individual design shown here. This is an easy
way for a customer to take a look at
which guide they’re most interested in
based on its placement within the genomic sequence
or the transcript sequence. Select an individual guide– this is your design ID if ever
you want to look it up again. You’d also list the
genomic position for the guide sequence– the strand, the sequence
itself, and it’s adjacent PAM. It has an on target score of 62
and an off-target score of 33. 62 indicates that
this particular guide is likely to work. 33 means that it has a higher
off-target profile and has a less favorable on
target– excuse me– off-target specificity score,
but is one that we would still provide through our database. You could look at more details
of the off-target locations by expanding this window below. The on target location
is listed first, and you can see
it’s intersecting with our gene of interest. We also show individual
off-target locations and whether or not they
land within an exon for an off-target gene. Here we also show the sequence
of that off-target location, it’s adjacent PAM sequence,
the PSSM score associated with that particular alignment,
the number of mismatches, and the one based coordinate
of the individual guide sequence listed on the
chromosome, strand, and position number. You can download all of these
sequences in these locations to an Excel file or to a CSV. The custom design has completed
running while we were waiting. On the left, I’d like to show
you the visualization that allows you to quickly see
the comparison of on target potential and off-target risk. As you go up on the y-axis, we
go from high off-target risk to low off-target risk. On the x-axis, we go from
low on target potential to those that are very
likely to work having a high on target potential. You can select guides that
fall into any of these bins by clicking on the
particular box. The number within
the box indicates how many guides are associated. And in the visualization
to the right, you can see that there
are now one, two, and three guides that
are selected by checking on that particular box. You can see that the
individual guides were oriented along the length of the FASTA
sequence that you provided, and you can further select an
individual guide from that’s set by clicking on it. Because I selected a
genomic region that included the sequence five
prime of the first CDS region of TRPM7, we
can pull one design from the database, which was
this one right over here– and you can see that one get
checked as I click on it– indicating that that’s one
that was previously designed and found in the database. We also know that
these are now ones that were not pre-designed
in the database but were designed against
our FASTA sequence. The tool quickly differentiates
between these different options, or these
different sources, by showing in this orangey color
that it’s a pre-designed Alt-R CRISPR-Cas9 RNA– additionally, includes the
gene name in the design ID. For these custom sequences,
the custom sequence ID is listed here followed by
the fact that it is a custom Alt-R CRISPR-Cas9 gRNA. Just like the on
target sequences that we pulled from the
database in the prior option of the tool, we provide access
to the off-target landing alignments for the particular
guide sequence in the dropdown, and you can again download
all of these to an Excel or export it to as CSV file. And like before, we see
that some off-target allignments land within
exons while others do not. So quickly, if we wanted
to look up a custom design ID that had been generated in
a prior custom analysis run, we can select an
individual design ID, change the input type up
here into a design ID, input that here, and search for
that particular design idea. And it comes right
back as the last one. We also do a quick
order, which will order this particular
guide sequence and go directly into our
ordering tool for easy access. Finally, I want to
show doing a check. So here we’re going to
check the same guide sequence against our tool. We’ll import this
as a FASTA sequence. We’ll do off-target training
against Homo sapiens. And this will take
about 15 seconds based on my prior analysis. Here we have the opportunity to
upload a lot of guide sequences simultaneously. As you can see in the
screen, we can upload up to 99 of these FASTA
sequences simultaneously. Because this is a
checker tool, we provide a bit more
information on the screen to help guide you to whether
or not this is a good guide. Here we note that this is a cRNA
that is expected to be good. If it had a particular
issue, which we identified, such as having multiple
perfect off-target alignments, that would also be listed here. Similar to the
prior page, we have all of the off-target
alignments listed down below. We still retain the on target
score and the off-target score. Now return to the presentation– again, all of
these links will be provided in the slide deck
that will be available to you after the presentation. Our design tool will work for
any of our guide products– the crRNA, the crRNA
XT, or the sgRNA. And you can read
more about these after the presentation in
the notes in the slides. In conclusion, our on
target and off-target PSSMs have been validated using
internally generated data to perform well, and we’ve
demonstrated that they predict high quality guide RNAs. The Alt-R design tool has
many features, some of which include custom designs
against any FASTA sequence, the ability to check
generate guides elsewhere. It’s an easy to use guide RNA
quality visualization tools as well to help you
select the guides that are going to have the highest
balance of on target likelihood to work as well as
high specificity. We also allow access to
a pre-designed library, which allows rapid access
to previously designed, high quality guides. And these guides are available
for five different species. Of course, developing
these tools wouldn’t be possible without
a lot of other people here at Integrated
DNA Technologies, including people in our
web lab, production, web development, bioinformatics
teams, and elsewhere. And I want to express
a deep set of gratitude for all of them who have
been able to help out on this project. And at this time, I’ll take
any additional questions that are available to the audience. Thank you very much. Thank you, Matt. That was excellent. So at this point, we’ll
switch over to some questions. As a reminder, if
you have a question and have not
already done so, you can go ahead and type
it into the questions box, located on the right
hand side of your screen. So we have a few
questions already, and I’ll just start from
the top of the list. The first one is– can I specify a
different PAM site in your tool for Cas9 variants? Unfortunately, that
is not a feature that we currently allow. OK, the next question is,
if your training data use human and rat genes,
how applicable is your tool to other species? We believe that
sequence based features are the most
informative information. And given that sequence based
features are biochemistry as opposed to biology so
much, our current belief is that this will work
well across any species. There are some other features
that, as the literature has indicated, may be predictive
that I didn’t list here, such as chromatin structure
and methylation patterns. Those would be
cell type specific and would also vary by species. Our results so far
indicate that guides that our tools
designs work well. And we’ve tested these guides
in a number of different cell lines and across
several species. Great, thanks. Another questions– just
to clarify about off-target scores– is a high off-target
score good or bad? A high off-target score is good. It indicates a high
level of specificity for that particular guide. Great. This is similar to the
question that you just answered, the prior question,
about specifically designing targeted guide RNAs
for use in bacteria. Now you’ve already said
that the same design tool can be used for
multiple species, but is there any general
advice that you’d give for the design of
RNAs specific to bacteria or for CRISPR interference? For CRISPR interference,
the placement of your guide is going to be important
to be able to make sure it’s going to interfere
with the particular sequence space features you’re
most interested in. Off-target activity
is unclear with that. I really haven’t looked
into that too much. So as I indicated when we
were at the very beginning of the talk, when we were
discussing off-target activity and the importance
of it aligning to other exons that share
homology, with interference you might run into issues where
the off-target alignment isn’t so much in the exon so much
as in a regulatory region that shares homology with
your own target location. While our tool
doesn’t explicitly do that intersection,
you could download the off-target alignments
that our tool generates and determine whether or not any
of those off-target alignments are also in regulatory
regions of interest to you. Sorry, Justin, what was the
other part of that question? So it was just about any
specific design recommendations for use in bacteria. For bacteria– yeah,
unfortunately, I haven’t done much
with bacteria myself. So the next two
questions are related. So I’ll try to
combine them together, and it’s about doing off-target
analyzes in species currently not supported or
listed in the tool. So the question is,
are there other options for doing off-target
analyses in other species, and do we plan to expand
the off-target analysis for additional species, such
as Chinese hamster ovary cells. I have not been part
of any discussions where we have
discussed expanding our off-target analysis for
currently unsupported species. I’m sorry. I think that’s definitely
something we’re going to continue looking into. SO I’d would say, stay
tuned for other species and hopefully we can
expand beyond that. We can also take another
look at the design requests. If you’d like to email
[email protected], we can continue
that conversation. Another question here is,
what is the general cut off of off-target scores
or on target scores. And do we have a
recommendation for that? Yes, so for on target
scores, we determine guides that are likely
to work as those having scorers greater than 50. For off target– so for
the on target score, our recommendation is
for an individual guide to have a score greater than 50. And as you’ll remember, our
tool has a recall of 51%. That indicates that
there are guides that do work that
are tool will miss, but there are generally a
lot of guides out there. And so when we
developed the tool, we determined that
this was OK so long as that the guides we do
we indicate are good really are good. And our tool does an
excellent job of that. So if you do a custom
design and your guide– if the only guide
that’s available to you has a score of less
than 50, while we can’t say that it’s good and
our tool isn’t predicting that it’s good, it doesn’t
necessarily mean it won’t work. So if that’s the only option
to you, you can use it, but it would be
at your own risk. For off-targets scores, we
scale our off-target specificity between 0 and 100. There are guides that have
an off-target specificity score that would effectively
be negative on that scale. Those are ones that we
definitely do not recommend, and they might have lots
of off-target landing sites, many of which might be
perfect off-target alignments against the genome,
indicating that each of those targets in
that particular case might act more like an on
target site with high activity. In that case, we
would definitely not recommend any guides that
have a score of 0 or less. Excellent. Thanks, Matt. So here’s another question
on a similar topic. This is about modifying
mouse embryos. And the question is,
is it necessary to do further validation if a
guide RNA has a good score. You’ve already spoken
about the cutoff line. But perhaps before going
into direct delivery to a mouse zygote, is
there any other validation you might recommend? It depends on what
you care about. So having a guide that has
high on target activity is always desirable. Validating that
the guide doesn’t have off-target
activity in regions that are of importance
to you is something that different customers have
different feelings about. Given that an individual guide
could have off-target activity in a gene that will
cause a phenotype that’s similar to the on
target guide, I would recommend that
you would, minimum, want to look to see
whether or not that’s likely to be true by examining
the off-target alignments, particularly those that align
within genes that might share function, or at least be within
the same pathway as those that you care about. For additional work, you
could, if you’re interested, test your guide in
subculture first and run a tool,
such as GUIDE-Seq, to identify real world
off-target editing, as opposed to the insilical
predictions generated by the tool, which
should get you closer to what will actually happen
when you inject it into it embryo. That’s great advice. Thanks, Matt. So just a couple
more questions– the first is, if
I’m performing HDR, should this impact how
I use the design tool? Right. So our results indicate that
having high on target activity is going to give you
better HDR performance. Off-target activity is going
to vary depending on needs. If you’re doing a
single guide edit to be able to do
homology directed repair, you would probably also
want to pay attention to off-target editing because,
while you’ll repair the on target site with your homology
directed repair construct, you won’t have a similar
ability to repair those off-target sites landed
on by an individual guide. If you’re doing this through
a nickase approach, where you only nick the DNA, you
would design two guides through our tools, and
you would design them on either side of your
edit your sequence. But because these
are individual nicks, it’s possible that
off-target editing may be less important in
this particular scenario. Excellent. Thanks. And one last question
here, and this is about doing very
large sized projects. The question is about
screening and needing thousands of designs. Is that something
still supported by the Alt-R design tool? It is. You can look up
up to 100 gene IDs through the online interface,
and there’s also the ability to upload a list of gene
IDs through the interface as well using a file format. If, however, you are unable to
get the kind of performance out of the online
design tool that you need for your
particular design, I would recommend that you contact
our customer support team, and they would be happy to help
you with your project needs. Thank you, Matt. OK, there’s all the time
we have for questions. I want to thank all of you for
attending today’s presentation. I also would like to thanks
Matt for his informative presentation, as well as Justin
for conducting the question and answer session. This is one of a
series of webinars we’ll be presenting on CRISPR,
as well as other topics. We will email you about
these future webinars as they are scheduled. Also as a reminder, a recording
of this webinar will be posted shortly on our website and
at There you will find several
other educational webinars on such topics as next
generation sequencing, genotyping, qPCR, and
general molecular biology. Also the presentation slides
for today’s webinar have been published on our SlideShare
page at Thank you again for attending,
and we wish you the best of success in your research.

You Might Also Like

No Comments

Leave a Reply