1. DRUG DISCOVERY
DD
Advancing
vancing
ADME
Edited by THERINE BAGLEY,
BAGLEY
Associate Editor
MA TROUTMAN, group leader in PHARMACOKINETICS
DYNAMICS and METABOLISM AT PFIZER INC., talks with an expert panel
on the future of high throughput absorption, distribution, metabolism, and excretion (HT ADME)
technologies. Panelists include: NICK LEVI , Senior Product Manager of AB SCIEX , and
GREG SAMOIL, Senior Product Manager at THERMO FISHER SCIENTIFIC.
2. WHITEPAPERS
MA TROUTMAN: My rst question is We’re able to look at 2B6, which is a lower abun- as being able to apply their own data analysis and
about the next generation of HT ADME pro l- dance cyp. algorithms. We then seamlessly apply it to an appli-
ing, which may require measurements beyond the We’ve taken this single analysis and compared cation that treats it like a built-in study type so the
2 it against the traditional method, against the users can publish their designs out to other users
analysis of a compound itself. For example, in order
mRNA method and against the western blo ing and enable access control to make it appear that
DRUG DISCOVERY
to gauge the propensity of a compound to elicit en-
zyme induction, one would measure the resultant method, and we’ve actually found there’s very good they’re using a standard assay experiment. By
changes to enzyme mRNA. correlation. e advantage of doing (analysis) doing this, we’re able to maintain consistency
with a mass spec is that you’ve already got the with established up running procedures across
WHAT NEW TECHNOLOGIES OR AP- equi ment. Secondly, you’re able to monitor many an organization, help to reduce complexity of the
PROACHES MIGHT BE ON THE HORIZON di erent peptides, especially with the latest mass operations, ensure compliance, and improve the
FOR QUANTIFICATION OF COMPOUND spec technology, simultaneously, so you can get a quality of the data for these assays.
DISPOSITION AND DRUG-DRUG INTE C-
TION OF THE NATURE DESCRIBED ABOVE,
lot more information a lot faster. irdly, it’s very
speci c. at’s one of the interesting advances that MT: I think data ow and information
management is certainly a key challenge in this,
FOR TIMES WHEN YOU’RE NOT NECESSAR- I think is going a ect the industry.
ILY ME SURING COMPOUND, BUT PER- We’re going to be providing a human-induc- and we’ll get into that a li le bit farther down in an-
HAPS CHANGES IN BIOLOGY RESPONSE tion kit, which has a set of heavy peptides, di erent other question. We talked in the previous question
INSTEAD? HOW ARE THESE (TECHNOLO- reagents, digestion, etc. for 100 assays for this cyp about non-traditional ADME pro ling, in particu-
GIES OR APPROACHES) BEING THOUGHT induction. We’re pre y excited about that. lar, measurements of biological response. In this
MT:
ABOUT FOR HIGH CAPACITY, SHORT next question, I’d like to bring it back to talk a li le
TURNAROUND TIME FOR THE SAKE OF In listening to your response, one idea bit more about compound measurement itself.
DOING EXTENSIVE PROFILING FOR MUL- that went o in my head was that this could be We’ve seen more of a drive to try and create
TIPLE COMPOUNDS OR SAR-GENE TION applied directly to measurements of other protein assays that produce higher content information
IN EARLY DISCOVERY? expression, like transporters. I think understanding that was previously unavailable in the realm of HT
of biology being pushed earlier and earlier into dis- ADME. An example of that would be early identi-
NICK LEVI : One of the interesting things covery to enable further knowledge of disposition cation of major metabolites. I think in part that’s
about drug-drug interactions are how the cyp and pharmacology requires abilities to quantify the been driven by some of the newer FDA guidelines
induction assays are done. Currently, we have a numbers of proteins, be they enzymes or transport- around monitoring major metabolites. We’ve actu-
couple of di erent techniques, like the mRNA ers; it’s quite exciting. ally found that we can generate a lot of metabolites
methods you mentioned, enzymatic activity with
the probe-speci c substrates, and western blot NL: Exactly. As you said in the question, it
is very interesting that we’re not looking so much at
that are in fact clinically relevant from in vitro stud-
ies and can produce some highly valuable informa-
techniques to actually measure induction. Each tion that discovery teams can use to make decisions
one of these techniques has their positives and the compound anymore, but actually at the cellular and plan future e orts. e one limitation that
negatives. With the mRNA, we’re only assessing environment in which it’s having its e ect. I think we’ve seen, however, is the analysis of the volume
the cyp induction from gene transcription changes. that’s really a fascinating advance. of data; the speed at which you can perform both
For the enzymatic activity, we pre y much rely experiment and data analysis is a bit limiting to do
on enzymatic conversion of probe substrates
GREGORY SAMOIL: From a more
in a typical discover-type environment.
general aspect, as far as technologies that we’re
and require a di erent probe substrate for each
utilizing, ( ermo Fisher Scienti c) has taken our CAN YOU DISCUSS SOME TECHNOLOGIES
isozyme, which isn’t always possible. Western
products and engineered our systems to be exible THAT MIGHT BE FORTHCOMING THAT
blo ing measures the actual protein levels of p450
and accommodating to the ever-changing needs of ARE EXPECTED TO ENHANCE THE LEVEL
enzymes usingisoform-speci c antibodies, but cur-
data management while still being able to incorpo- OF CONTENT OR INFORMATION THAT
rently there are only a few good antibodies that are
rate automation and ensuring standardization. CAN BE GENE TED FOR COMPOUNDS?
isoform-speci c.
We’ve done this by implementing user-de n- WHAT CORRESPONDING TECHNOLOGY
Ideally we’d like to do this with mass spectrom-
able study designs that allow users to de ne their IS BEING DEVELOPED TO HELP MANAGE
etry where we are able to address the challenges of
assay experiments with the ever-changing needs of THIS VOLUME OF DATA SO THAT IT CAN
assay selectivity and sensitivity; we’ve already put
ID experiments and in vitro experiments, as well
out some data on this type of application. We’re us-
ing a mass spec-based approach to provide the sen-
sitivity and speci city to detect individual peptides
from these speci c p450 isoforms. Using a mass
Matt Troutman|
DIRECTOR, HIGH-THROUGHPUT (HT)
spec that has a really fast scan speed and the ability
ADME CENTER OF EMPHASIS (COE)
to track several multiple reaction monitoring transi-
PFIZER
tions (MRMs), we’re able to actually take a look at
a set of cyp450 enzymes and be able to track their Mr. Troutman joined P zer in 2002 and leads the company’s HT
levels in samples in a single injection. We nd that ADME e orts within its Pharmacokinetics, Dynamics and Metabolism
rst of all, for the speci c p450 assays, it is quite department. e HT ADME COE has responsibility for providing
easy to make stable isotopic label peptides and to in vitro ADME data including metabolic stability, permeability and
be able to measure those against the peptides of the absorption, distribution and drug-drug interaction pro ling.
assay. Using the LC/MS method, we’re able to look Mr. Troutman has a Ph.D. in Pharmacy from the University of
at 12 di erent peptides representing the four most North Carolina at Chapel Hill.
common p450 proteins – 1A2, 2B6, 3A4, and 3A5
- simultaneously.
3. WHITEPAPERS
BE PERFORMED ROUTINELY IN HT ADME
APPROACH? “We’ve seen more of a drive to try and create
GS: Here at ermo Fisher, we’re employing
the technologies that streamline the process and
assays that produce higher content infor- 3
DRUG DISCOVERY
maintaining only the necessary data required for
that decision-making process. With the amount
mation that was previously unavailable in the
of data that’s being acquired through high-content
screening, we’re only keeping data that’s pertinent
realm of HT ADME… (but ) the speed of
to the decision-making process. So we’re constantly
utilizing the latest versions of the data repository
data anal ysis is a bit limiting.”
so ware to take advantage of their technologies
and their advances. Our policy is to use those latest
especially in relation to this question in which CAN YOU DISCUSS SOME NEW AP
technologies and develop our products, which are
conduct of the assay is actually the quick part. Ana- PROACHES AND TECHNOLOGIES THAT
in turn used by our customers.
lyzing data, being able to si through and manage WILL SIGNIFICANTLY ENHANCE THESE
MT: I think that’s de nitely a key notion;
that si ing through and highlighting just the key
it, and ge ing it out to teams is becoming the real
challenge.
ELEMENTS, AND MAY ALSO AFFECT THE
EFFICIENCY BY WHICH DATA IS GENER
data needed to make a knowledge-based decision is
NL: ATED?
NL:
It is easy to go wrong on both sides. To
becoming fairly critical. So any added ability in that
have a team going very carefully through all the me- One of the interesting things about
realm would be very helpful.
tabolites for compounds that don’t actually make it the question is it addresses the idea of turnaround
NL: Greg’s thinking very carefully about
the Laboratory Information Management Systems
any further into the pipeline, or to not collect any
information and then not have that.
time versus the idea of throughput. “Turnaround
time” being how fast I can get the sample back.
(LIMS) and about the transfer of data, which is
extremely important. From AB SCIEX’s side, we’re
MT: Absolutely. Going back to Greg’s
point, noting some of the potential regulatory
“ roughput” being how many samples I can do
a week. ey’re related, but di erent concepts.
thinking more about generating that data, speci - We’ve been focusing on both, but I think the turn-
concerns there, what to keep and what to discard is around time is really the key thing to be addressed.
cally about the metabolic stability assays. We have a
a key challenge. First, we’re doing a lot of work trying to get be er
lot of metabolite information in the samples; the
key is to get that information out quickly and then
use it as necessary, as Greg said.
NL: We also believe that an accurate mass
spec approach is going to be very useful in the fu-
quality information into the system. We know
labs don’t have enough time to optimize each
Technologies are ge ing be er and be er at compound manually, those days are long past, but
ture. I think that it’s really important to have a ver-
extracting information, but obviously we don’t many of the automatic optimization algorithms
satile instrument that can perform these multiple
want to keep every single last bit of data that we that people are using aren’t good at nding all the
functions. In some sense, the more information
collect from very early on in the pipeline all the way compound parameters early on in the process.
you can get beforehand in order to distill it earlier
to the very end. With our latest mass spec, which AB SCIEX and MDS Analytical Technologies
in the process the be er. We’d like to see a box that
is 5500 Q-trap system, we have the speed and the have been working to have be er optimization
has both excellent qualitative and quantitative data
quantitative ability to perform rough untargeted algorithms to nely tune the compound. We’ve
become used throughout discovery. In the last
screens for metabolites as well as exact targeted been looking at some timesavers, like being able
couple of months, we’ve been pu ing out research
screens for metabolites. We’re able to produce to do on-the- y saturation control, which is in our
papers on some potential ideas for a box like this.
upcoming product DiscoveryQuant 2.1, where
the data very easily without sacri cing any time
doing the injections. e rst real bo leneck then MT: Very good point, Nick. e next ques-
tion is really going to deal with some of the more
you can get compounds that you’re saturating
and change parameters to x the optimization by
becomes how to quickly instruct the instrument to
collect all data. Second, how do you automatically logistical aspects of what we do with HT ADME turning down the signal on the detector, which
evaluate all the results so you can whi le pro ling. As you well know, a lot of what we do we allows you to then be able to perform a really good
them down into a smaller amount of data that can do is basically to provide data to enable discovery. I optimization.
be transferred through the organization and can be think a key aspect of a study’s success is being able We’re also looking into performing chromato-
saved in the LIMS system? to provide su cient data to be able to make good graphic tests beforehand, when the compound
We’re working with our DiscoveryQuant decisions, either by enough pro ling or enough meets the mass spec, to characterize everything
so ware to be able to collect more data. We’re generation of SAR, what we might call capacity. before doing the analysis. Key to actually being
using the predictive MRM approach, which takes Certainly, we want to have a given level of quality able to go fast is being able to network this quality
the major metabolite transitions and adds them to and granularity to enable those decisions as well. information with a global database. e philosophy
sample, and then we’re using that in the Discovery- One key challenge we always face is trying to do is that the rst time the compound meets the mass
Quant so ware to generate the lists of metabolites this within the right time scale, something we typi- spec, you do the optimization with quality so ware
right o the bat. We think it will be very useful and cally term throughput. and then dump optimization into a global database
will really give the projects a lot more information Most o en what we try and shoot for is that every lab worldwide has access to. e labs
to be able to design drugs be er. generation of data on a week cycle, which is driven can then use that information next time they want
MT: I think we certainly have some experi-
ence with this and are quite excited by a lot of this
typically by a design team’s cycle. A lot of times, a
limiting factor is the bioanalytical methodology
an assay on that compound; they don’t have to
optimize it again.
applied. We’ve also been looking at some high through-
technology. It’s becoming an interesting situation, put approaches for actually doing the assays, for
4. WHITEPAPERS
both turnaround time and throughput. One has really bene ted our users by automating their for example.
4
approach is to completely automate the sample
setup process-to be able to have plate templates
processes and allowing them to do these multistep
processes. We’re automating the more complex MT: at’s a great segue into the last ques-
tion. ADME data is being produced on a large scale
and very quickly load in a compound le, which is experiment templating. e interfacing instrument
a list of compounds and their well locations, and is fairly extensive in Galileo. e data analysis, for multiple endpoints, in some cases from various
DRUG DISCOVERY
have the methods already set up so that you’ll be review and acceptance tasks are facilitated by easy- sources, both internally and at CROs that may
able to run those (samples) just a few minutes a er to-understand user interfaces that promote rapid leverage to do some of the work.
you start the loading process. data review. So essentially we’re allowing users to Consumption of the data is usually in a short time
Second, we’re taking a closer look at pooling set up their criteria with which the data will pass cycle, and it goes back to what we talked about a
strategies. Some companies are using advanced or fail and we allow them to centralize or focus in li le bit in the last question with turnaround usu-
pooling strategies for their in vitro analysis, and on the data that may be questionable. Good data ally facilitating design cycles. HT ADME data is
some companies are not. We think that one of the is accepted, poor data is rejected, and only that being increasingly used to inform design decisions
main reasons companies aren’t using this strategy is questionable data is aimed at achieving the right PK pro le rather than
that it’s just too complicated. One of the great
things about so ware is that it’s good at reducing
complexity if wri en correctly. So we’ve been look-
“It is very interesting that we’re not looking so
ing very closely at pooling strategies and our dis- much at the compound anymore, but ac-
covery so ware. We are looking at pooling based
on di erent compound characteristics. Advanced tually at the cellular environment in
pooling strategies allow for really high quality
data from your pooling. which it’s having its effect . I think that’s re-
e other side of this issue is processing all
this data. We’ve had labs come to us and say, “it’s ally a fascinating advance.”
great that you can give us all this data, but we can’t
process it quickly.” We’ve come out with a package
called MultiQuant so ware, which is able to both reviewed by the analysts. just collecting data on particular aspects of the
MT:
process data faster and nd any anomalies in the compound itself. In addition to that, we’re more
data. I think that hit on a really key point: routinely seeing that the data is also being used to
spending the time only on the data that needs create in silico models for the endpoints to predict
MT WHEN YOU TALK ABOUT DATA, human intervention is very important to increasing likely properties of compounds, which actually
YOU’RE TALKING ABOUT e ciency and reducing complexities. inform the design decision to make or not make a
LIQUID CHROMATOG PHY MASS SPEC
GS: compound.
MT:
TROMETRY LCMS DATA, CORRECT? Right. In a system like Galileo, it is set
NL:
up so that the users feel like they have control over One critical aspect of this is data man-
I am. We are very focused on the data and that there’s integrity to the data. agement, and in a lot of cases, how we e ectively
LCMS part of the process.
MT: manage the data will drive how we can actually do
GS: I think also from an organizational these activities.
From the ermo Fisher side, looking perspective, to know that people are review-
at the amount of data that’s coming across, the rst ing data, especially when it needs to be, is very THE QUESTION IS WHAT NEW DATA
thing that we address is trying to get that data into important. MANAGEMENT SOLUTIONS ARE ON
the system using a data management solution such
as Galileo from ermo Fisher Scienti c, which is GS: It’s also good to mention the fact that
how we’re presenting the data is important. To
THE HORIZON THAT WILL ENABLE THIS
PROCESS? FIRST, HOW DO WE DEAL
importing data directly from the instrument. Once WITH THE SHEAR VOLUME OF END
that data is in, it is automatically analyzed based on visualize the data is one thing, however the users POINTS? SECOND, HAS THERE BEEN ANY
a series of experiment templates. must be comfortable with reviewing the data as THOUGHT GIVEN TO HAVING A MORE
As Nick mentioned, the use of templating well, giving them the option to see the status ags, STREAMLINED PROCESS TO HAVE THIS
BE BUILT AS IN SILICO OR COMPUTA
Nick Levitt| TIONAL MODELS? AND THIRD, HOW WE
CAN POSSIBLY PUT DATA TOGETHER
SENIOR PRODUCT MANAGER FOR MORE HOLISTIC UNDERSTANDING
AB SCIEX OF PK THER THAN JUST DATA ON AN
Mr. Levi manages mass spectrometry and related products for the ENDPOINT BY ENDPOINT BASIS?
worldwide Pharmaceutical/CRO market, focused on the quantitation
of compounds, with particular focus on the ADME and clinical trials
GS: When thinking about data management
needs, it’s di cult to predict what future data will
portion of the pharmaceutical drug pipeline. In past positions, Mr. look like and therefore the requirements around
Levi has managed and developed successful so ware products and managing it. At ermo Fisher Scienti c we’ve
strategies for the pharmaceutical and medical industries. He holds an architected our solutions to be able to address
MBA from the Wharton School of Business and a degree in Nuclear future needs of data management. We’re already
Engineering from the Massachuse s Institute of Technology. doing this at our customers’ sites with Galileo by
storing and organizing the data in a central and
5. WHITEPAPERS
secure repository. New generations of so ware
solutions utilize databases that control access and
enable a one stop shop for data and data mining.
“New genera-
Also, by guaranteeing this secured layer, scientists
are able to spend less time and resources reviewing
tions of software 5
DRUG DISCOVERY
the data. e controlled access can be viewed as solutions utilize da-
key to bene ing and maintaining the integrity of
the data and ensuring accuracy. tabases that control
NL: I de nitely agree with Greg. I think
there’s some very interesting things happening in
access and enable a
the in silico model world. I nd that there’s not
enough collaboration and partnership between
one stop shop for
the emerging in silico companies and the more
established data-driven, instrument-producing
data and data mining.
companies, partly because the incentive structure
is a bit reversed. If you’re selling a really expensive
Also… scientists
instrument, you don’t necessarily want to enable
a be er in silico model, but I think in the end run,
are able to spend less
the in silico models will only get be er and are
worth using more if you’re doing drug discovery.
time and resourc-
We de nitely see the Galileo system as being
one of the standards for drug discovery, but we are
es reviewing the
also looking at things broadly. As Greg says, the
LIMS systems need to have the exibility for dis-
data.”
covery. Some of the traditional LIMS systems are
very good at tracking samples, but not as good for
dealing with exibility, which is needed in discov-
MT: Absolutely. I think with the one
platform, you’ll be able to really use the data in
ery. combination, for a PB/PK model, like SimCyp or
On the other side you have lab notebooks, Gastroplus, to be able to have all that data, which
which are very exible, but they’re kind of scientist- might be stored in Galileo and Watson, to just go
based. ey come from the genesis of being able to ahead and populate to get a good prediction for
get everything from an individual scientist-note- wholebody PK.
book into a secure format so that you can look at Coming back to that notion on in silico. I guess
it 10 years later. at doesn’t always work when it’s a very good point and actually, you can probably
you have many molecules for many projects going make the same claim to in vitro experimentalists.
on. We’ve de nitely seen companies like ermo For the stu that we’re already good at predicting
become more exible in this realm. via in silico, you’re able to do a li le more quanti -
One of the things I’ve personally found cation of proteins around induction, or things that
very intriguing is the work being done at Indigo you can’t currently do because resources are lim-
BioSystems, which is a startup company run by ited. Certainly, I think, if we are successful with in
Randy Julian. ey have an R&D data management silico, the compounds being made have less of the
system that is using a blend of web technologies to traditional liabilities, but there’s other non-tradi-
create a really exible data framework within the tional stu that one has to worry about. It’s quite
discovery enterprise, which allows companies to interesting.
very easily customize both the calculations and re-
porting. As Greg says, that’s kind of the name of the
game in discovery because things are always going Greg Samoil|
to change, and the pace changes pre y fast. SENIOR PRODUCT MANAGER INFORMATICS
GS: Good point Nick. To further that, when
we talk about looking at the data and taking the
THERMO FISHER SCIENTIFIC
data from our systems, if you look at our Galileo Mr. Samoil joined ermo Scienti c Pharmacokinetic Services team
system and our in vivo system from Watson, our in 2000 and became a Product Manager in 2004. As the Galileo LIMS
customers are asking more and more of us to be Product Manager, he is involved extensively with the development of
able to combine that data. So, we’re moving to a new so ware features and the evolution of the Galileo LIMS to meet
more common platform across all of our applica- customers’ evolving informatics needs.
tions that ultimately will bene t a customer so that Before joining ermo Scienti c, Mr. Samoil conducted plant
they’ll be able to access the data in a single level, to immunology research at the University of Concordia, Montreal.
do their data mining in there, and be able to easily
analyze and access the data.