Going with the Information Flow
Data acquisition and instrument control technology,
long buried inside of automated laboratory instruments, can't be
The following is a manuscript for an article published in R&D
magazine. R&D magazine holds the copyright for the finished
C.G. Masi, Contributing Editor
Researchers involved in discovering and developing new pharmaceutical
compounds generally don't have a lot of background in advanced semiconductor
electronics. This we know.
They are chemists, biologists, medical doctors--almost anything
but electrical engineers. They do what they do, and do it well,
because they care about biochemistry, medicine, etc., not microelectronics.
So, it's a bit unfair to expect them to dig into the bits and bytes
of data acquisition (DAQ) systems, but that is just what the Universe
is forcing them to do.
DAQ is nothing new in the biochemical laboratory, or any other
laboratory for that matter. In the approximately 40 years since
the folks at Digital Equipment Corporation (later absorbed into
Compaq Computers, Houston, Tex.) made computers compact, affordable
and easily programmable, data acquisition (DAQ) technology is what
has made automated instrumentation feasible. (For more information
on minicomputer history, see the PDP-11
Up to now, however, vendors of automated laboratory instrumentation
have managed to hide their DAQ technology under the covers of stand-alone
instruments. They did this for the simple reason that they were
using DAQ to make life easier for pharmaceuticals researchers, not
more difficult. Why force researchers to deal with the ins and outs
of data acquisition systems when they have more useful things to
think about, such as which compound does what and how well?
"That was okay," Wolfgang Winter, Product Manager for
Data Systems for the Life Science Business Unit of Agilent Technologies,
Waldbronn, Germany, points out, "as long as the DAQ systems
were islands of information in the environment. Now, the pharmaceuticals
companies have to relate all of these islands of information into
their ocean of information."
That change is forcing pharmaceuticals researchers to come face-to-face
with DAQ systems, and forcing instrument vendors to find ways to
help them do it with as little pain as possible.
What is this DAQ, anyway?
Data acquisition is, quite simply, the art of using digital computers
to collect and record measurement results electronically. It has
two major advantages: it can note instrument readings much more
precisely and orders of magnitude more rapidly than any human laboratory
assistant could ever do; and it can store those readings in permanent
digital electronic records that are readily accessible to any analysis
software. A third advantage, which is often just as important, although
not as universally exploited, is the possibility of tightly integrating
data acquisition with automated controls.
Fig. 1 shows a DAQ system Brant Bergdoll, Senior Project Engineer
with V I Engineering in Indianapolis, Ind. built for Eli Lilly and
Company (Indianapolis, Ind.) to use in testing the reactions of
laboratory animals to experimental drugs intended to dilate or constrict
blood vessels. V I Engineering is a third-party integrator who specializes
in designing and building custom DAQ systems for clients in a wide
variety of industries--obviously including the pharmaceuticals industry.
|Fig. 1: This hemodynamics test system illustrates the basic
data acquisition elements: sensors to transduce physical quantities
to analog signals; signal conditioning electronics to prepare
those signals for conversion to digital data; and a DAQ card
to do the conversion and post the results to a host computer,
which sets down a permanent electronic record. Courtesy V I
Engineering, Indianapolis, Ind.
The measurement process begins with a set of pressure sensors inserted
through catheters into the heart and major blood vessels of the
test subject (whom we'll call "Bugs"). Specifically, there
is one sensor that monitors Bugs' left ventricular pressure and
another that measures his blood pressure. A third sensor monitors
Bugs' EKG through standard equipment. The EKG equipment returns
its waveform to the DAQ system through an analog-signal output.
The two blood-pressure sensors are fairly typical fluid-pressure
transducers, where the fluid pressure distorts a tiny membrane.
A strain gauge measures the membrane's deflection. Such transducers
are common, inexpensive and very, very fast and accurate. They do,
however, require a source of DC excitation for the strain gauge.
Bergdoll put in a regulated power supply built by Hewlett Packard
(now Agilent Technologies, Palo Alto, Calif.) to supply this DC
The analog signals from the sensors go directly to a module installed
in an SCXI chassis built by National Instruments in Austin, Tex.
SCXI is a standard instrumentation backplane format, just as PCI
is a standard format for computer backplanes.
A wide range of SCXI modules are available from a number of DAQ
instrumentation vendors. Generally, SCXI modules are used to condition
and multiplex analog signals. Conditioning consists of amplifying
signals to improve their noise immunity, supplying any needed excitation
power, and providing calibration corrections. Multiplexing consists
of scanning the input channels to present them, one at a time, to
the next component of the DAQ system.
That next component is commonly referred to as "the DAQ card."
It is an electronic circuit that receives the conditioned analog
signal from the SCXI chassis, converts it to digital information
and presents that information to the host computer.
Many DAQ cards have signal conditioning and muliplexing circuitry
built in. They are used for applications where the front-end signal
processing is less demanding than in this electrodynamictester application.
DAQ cards are available that can communicate through any port on
the computer. By far, the most common type of DAQ card fits into
a slot on the computer's PCIbus backplane. DAQ cards communicating
through USB and PCMCIA ports are also available and gaining popularity,
especially in installations where laptop computers are used to save
space on laboratory benchtops. Less popular, but still well represented,
are DAQ cards communicating through RS-232 (serial) ports and other
DAQ cards generally write the data directly into the computer's
random access memory (RAM) so that it is immediately available for
processing by the host computer. In most systems, it is then the
job of the computer's microprocessor (running a specially written
DAQ application program) to do any initial data processing, display
immediately needed real-time results, and make a permanent record
In the hemodynamics test system shown in Fig. 1,
the immediately required real-time results are Bugs' heart rate,
his left ventricular diastolic pressure and the time derivative
of his left ventricular pressure. The DAQ application program calculates
those immediately from the incoming data, and displays them on the
computer's monitor screen. At the same time, it records them along
with the raw data on the computer's hard drive. It can also upload
data and calculation results to a central database over Lilly's
V I Engineering also wrote some post-processing software, which
could run on the DAQ host computer or on an individual researcher's
office workstation (using data downloaded from the network). This
post-processing software is set up to calculate any of approximately
20 variables of interest, such as the area under the rising part
of the blood-pressure curve.
The input for a DAQ system can be just about anything that can
be quantified, and there is no effective limit on the number of
channels, either. Jeff Steele, National Instruments Area Sales Manager
in their Bridgewater, N.J. office tells of an application where
the customer was trying to screen a large number of compounds for
binding to a receptor.
"They want to do that quickly," he says. "The only
way to do that is by doing it in parallel."
They achieve high parallelism in their test stream by tagging the
bound receptor molecules with a fluorescent die, then presenting
them to a machine-vision camera in a 96-well mitrotiter plate. Wells
containing compounds bound to the receptors show fluorescence, and
those containing compounds that do not bind stay dark.
They use a standard microtiter-plate handler to present the plates
to the camera under a UV light. From there on, the hardware is pretty
standard for a machine-vision application. The camera picks up the
image, which it transfers to a frame-grabber board plugged into
the host computer's PCIbus backplane. The frame grabber's job is
to select one image out of the video stream coming from the camera
(which it does when the robot signals that the plate is in position)
and convert it to an array of numbers.
Looking at the system from a DAQ perspective, the camera stands
in the position of the sensor. The frame grabber stands in for the
DAQ card. The acquired image goes directly to the computer's RAM
just like any other array of numbers representing acquired data.
This application converges into a clear DAQ application in the
software. The software knows a priori what parts of the image correspond
to each well in the plate. It then simply isolates the numbers representing
the light levels seen in those pixels, and sums them to get a fluorescence
level for each well. The 96 numbers representing the fluorescence
from each of the 96 wells are then the output the system reports.
To write the program for this application, Bergdoll had to blend
pure DAQ functions with image-analysis functions. That process was
made easier by electing to write the program, which is shown in
Fig. 2, using a graphical programming environment optimized for
DAQ applications (NI's Lab-VIEW) that also contains a library of
image-analysis functions (Imaq Vision). The application could also
have been written in a more conventional programming language, such
as Visual BASIC or C.
|Fig. 2: Programming environments developed specifically
for DAQ applications make software development easier. Courtesy
National Instruments, Austin, Tex.
Linking the Islands
These custom applications are still DAQ islands of information.
The real challenge comes when drug researchers want to combine disparate
analytical techniques into a test-process flow. Such process-automation
challenges have long been the norm on the production end of the
pharmaceuticals business. With the expansion of massive drug-candidate
screening efforts, they have reached all the way back to the discovery
stage as well.
"Combinatorial chemistry, for instance, produces libraries
of 100,000+ compounds," says Michael Swartz, Pharmaceutical
Marketing Manager at Waters Corporation in Milford, Mass. "Scientists
now have to look at each individual compound and test it against
a particular activity for therapeutic use. They do that using a
whole arsenal of analytical instruments, such as NMR, mass spectroscopy
and chromatography. No one instrument vendor has control or data
acquisition and manipulation for all of those instruments in one
Advanced drug-candidate-analysis techniques have taken on a hyphenated
alphabet-soup surrealism that puts Depression-era Federal agencies
to shame. Typical method designations include LC-MS-MS (liquid chromatography
followed by mass spectrometry followed by more mass spectrometry),
MUX-LC-MS-MS (multiplexed LC-MS-MS), LC-ICP-MS & LC-API-MS (LC
followed by inductively coupled plasma dissociation of target molecules,
followed by MS integrated with atmospheric pressure ionization--which
adds charge while leaving the target molecules otherwise intact--followed
by MS) and a host of other, similarly mind-wrenching, combinations.
The elements (LC, MS, ICP, etc.) are all sophisticated analytical
steps run automatically from predefined methods. They have to be
coordinated physically, temporally and electronically. On top of
that, the software analysis systems have to account for the effects
of all the twists and turns in the convoluted analysis procedures.
Finally, it all has to happen rapidly, reliably and repeatably--and
without driving everyone, from the system integrators to the researchers
wrestling with the results, crazy.
This tall order has driven instrument makers to become adept system
integrators for their customers. The integration has to operate
on three levels: physical flow of materials under test through the
system, flow of electronic signals coordinating the different elements
of the system as well as carrying test results back to the host,
and flow of information through the software.
Micromass, a UK-based division of Waters, offers a system called
ProteomeWorks, which combines sample preparation, instrument cleanup,
chromatography, several types of mass spectroscopy, protein sequencing
and cloning. The system's purpose is to start with a set of separated
proteins and automatically process them through to final protein
identification--and provide an intelligible report.
Since, as Swartz pointed out, no one instrument company has all
of the pieces of the puzzle, they have to make alliances with each
other to fill in the gaps.
A laboratory automation company like Camile Products of Indianapolis,
Ind. is in a slightly different situation. They have a history of
providing the glue to bind together complementary components from
unrelated companies into custom laboratory systems. For them, its
only a case of selecting the appropriate bricks to set in their
Camile's CLARK (Camile Laboratory Automated Reactor Kit) is an
example of the software needed to bind the various functions together.
It provides a graphical representation of the physical system with
displays of instrument-control parameters and real-time-updated
|Fig. 3: Vendors of laboratory automation systems are used
to integrating islands of automation into larger systems. Courtesy
Camile Products, Indianapolis, Ind.
Camile's system consists of both hardware and software. The hardware
is a set of boards that fit into a CLARK chassis, just as the signal
conditioning modules fit into Bergdoll's SCXI chassis. The CLARK
modules, however, perform the analog-to-digital conversion of the
DAQ card and the digital-to-analog conversion needed to provide
control signals to the analysis equipment. These boards communicate
to the analysis equipment via analog voltages and to the computer
via an RS-422 link. RS-422 is the multidrop version of the venerable
RS-232 serial port available on virtually every personal computer.
Additional system components might bypass the CLARK chassis by communicating
directly with the host computer via the RS-422 link their own digital
"We recently did a project where we were bringing in data
points from a particle analyzer and intefacing that with our Camile
TG software," Connor reports. "In a typical batch reactor
application there are not a lot of data points--maybe for 18-20
I/O points for the total system. When, you hook up a particle analyzer,
it starts collecting thousands of data points per second."
That makes a big difference, but it is a quantitative difference,
not qualitative. You can still use the same software and hardware
technology, just more of the same channels.
Pharmaceuticals firms now want to automate their whole laboratories.
That ratchets the system-integration problem to a qualitatively
Physically, it increases size of the problem by at least an order
of magnitude. It also expands the volume of information that has
to be moved and stored. Most significantly, the information output
from the laboratory flows directly to the company's enterprise-wide
information system. That moves it from the "instrument automation"
world into the "information technology" (IT) world.
"The things that are mainstream in the DAQ world," Agilent's
Winter points out, "are the exception in the IT world. Something
as common as IEEE-488 in the DAQ world is the exception for the
IT people. It gets difficult because the lab people have to justify
using it. It is not 'mainstream.' It is different from the office
"That was okay as long as the DAQ systems were islands of
information in the environment. Now, the pharmaceuticals companies
have to relate all of these islands of information into their ocean
of information. We are talking about networking all of these DAQ
systems so that they report all of their data into a central data
Agilent decided to standardize on networked data systems quite
some time ago. That means offering analysis systems based on IT-standard
interfaces, such as an Ethernet based network connection using TCP-IP,
which is just beginning to be done in the DAQ world. Being able
to support industry standard protocols with the instrumentation
itself avoids having the DAQ system become an island of information
in the first place.
These networked data systems can control multitechnique instrumentation,
like LC-MS, GC-MS, UV/VIS spectroscopy, capillary electrophoresis,
etc. There are even general-purpose interfaces that would allow
you to capture digital output from just about any other device that
exists. You can hook all of these up to the networked data system.
The networked data system will control the instrument, pick up the
data, interpret the signals and then spit out numeric results upon
which some kind of decision can be made. The software needed for
one instrument is called an instrument driver. The instrument driver
includes software modules for instrument setup, control and data
|Fig. 4: Networked data systems ease the system integration
problem by leveraging standard client/server network technology.
Courtesy Agilent Technologies, Waldbronn, Germany
The instrument setup modules include a graphical user interface
(GUI) that gives you access to the parameters that are relevant
on that instrument. The GUI for a diode array detector, for example,
would set parameters like wavelength and bandwidth. For units that
have more than one lamp, it would determine which lamp is to be
used. The GUI might also set a stop time, and whether it should
run an automatic balance routine at the beginning of the run.
Since the instrument drivers are modules within the same family,
you can plug them together to fit exactly the system that you have.
The important bit is that they can all be connected to a central
data repository, such as an Oracle database, that can pull together
all of the data that is being measured in the analytical lab. That
is useful not just for archival purposes and for backup, but for
also for correlation--putting pieces of data together that have
been measured over time on different instruments by different people.
"When I started in this business," Swartz recalls, "we
had our strip-chart recorder hooked up. It drew a red or a blue
or a black line on paper, and that was it.
"Now, we use workstations. We refer to them as clients on
a large network. A lot of our major customers have client-server
networks with point-of-use PCs that log onto servers and the applications
that are running are centrally managed."