User:Jonathan Cline/Notebook/Robotics/20091013

http://c.statcounter.com/5196544/0/c9629947/1/0.png

http://c.statcounter.com/5196544/0/c9629947/1/0.png

=Better, Faster Bio with Integrated Devices using Robotics=

Jonathan Cline

[mailto:jcline@ieee.org jcline@ieee.org]

2009

http://88proof.com/img/JCLINE-BUSINESS-CARD-320.png

= Robotics - Let's Do Something "Simple"=


 * Make biology easier to engineer
 * Make biology more reliable to engineer

by:


 * Methods: Improving the means and methods for bio/synthetic design;
 * Predictability: Improving repeatability and trust of experimental results and procedure;
 * Reusability: Engineering biological components and biological tools to be reusable across designs or across experiments.

for:


 * Increased productivity
 * Less busywork, more Design time
 * Less human interaction means less human error
 * Decreased financial cost
 * More experiments possible per day per person per lab

(Contrary to popular belief, lab techs are not "free")

= Robotics - Let's Do Something "Simple"=

Let's Automate a bio-protocol and remove the busy-work!
 * George Church Lab
 * "MAGE": A Machine That Speeds Up Evolution
 * "A genome-wide approach to genetic engineering greatly speeds the manufacture of bacteria for making drugs and biofuels."

Sounds great!

http://www.technologyreview.com/files/25028/evolution_machine_x220.jpg


 * Problem: Little re-usability?
 * "It's a great device for that experiment, and that experiment only"
 * It's not a general purpose bio-protocol processor.
 * Problem: Liquid Connectivity?
 * "It's self contained and can't connect to a larger protocol."


 * These are scalability and reuse problems.
 * Also known as Integration issues (industry slang).

= Robotics - Let's Do Something "Simple"=


 * Let's Automate a bio-protocol and remove the busy-work!
 * This time let's build an arbitrary liquid-handler that can move drops anywhere!
 * ¿AC Electrowetting Actuation of Droplets on a Digital Micro¿uidic Platform¿
 * ¿Rapid Prototyping in Copper Substrates for Digital Microfluidics¿
 * Problem: The droplets can only be moved by switching electricity by hand.
 * U-Toronto & UCLA: (A grad student's hand.)
 * Jonathan: I'll use electronics & software.

Jonathan's version
http://biosx.com/88proof/synthetic_biology/blog/wp-content/uploads/2009/06/p6033652w.jpg

http://biosx.com/88proof/synthetic_biology/blog/wp-content/uploads/2009/07/p7023926swc-prototype.jpg


 * Problem: Each millimeter of movement for the droplet requires an electrical switch. There are many millimeters to move, so hundreds of switches.
 * U-Toronto: "Even LabView makes it too complex."
 * The software represents a significant portion of the total system, and doesn't exist yet.
 * Problem: The chip only works about 1/3 of the time.
 * Well, the physics guys can fix that eventually.


 * These are scalability and reuse problems: control system issues and device interconnection issues.
 * The fabrication problem will be solved by any one of 10,000 postdocs.
 * Even when fabrication is solved, the control system problem still needs a solution.

= Robotics - Let's Do Something "Simple"=

Jonathan's version
Electrowetting using Copper PCB and solid state switches

http://88proof.com/biotech/img/jcline-ewod-2090711s.JPG


 * Only marginal success
 * Software is a big issue - for device control
 * Too difficult to debug both software & hardware at the same time
 * Improve software first (easier to debug); come back to hardware later
 * Develop software with working hardware

= Robotics - Let's Do Something "Simple"=


 * Let's Automate a bio-protocol and remove the busy-work!
 * This time let's use a big robot! It can do anything!
 * (at least, it costs enough, so it should be able to do anything.)

http://88proof.com/biotech/img/jcline-tecan-genesis-2000-1sw.jpg


 * Problem: It can really only to certain things that the vendor allows.
 * In fact, it can only move certain objects and perform a few operations.
 * And, some of those operations fail, too.
 * And, the vendor's software can't check for failure.
 * Failure requires human intervention, or can ruin the experiment.
 * Problem: Biologists have to Program it. Biologists do not program.

= Robotics - User/Software/Hardware Model=

http://88proof.com/biotech/img/Biolab-languages1.png

For Tecan, the commands break down even further, into High-level script commands (shown) and low level device operations unique to each device.

= Robotics - User/Software/Hardware Model=

http://88proof.com/biotech/img/Biolab-interconnect-model.png


 * The current usage for devices is monolithic: each device is programmed separately
 * Each device has it's own programming method
 * Difficult to re-use software written for one device, on another device
 * Biologist has to work harder; each experiment has unique elements

= Robotics - User/Software/Hardware Model=

http://88proof.com/biotech/img/Biolab-interconnect-model2.png


 * Bio-protocol application can be re-used
 * Robotics software framework abstracts "hardware operations" from real devices or network devices

Device data is stored in the database

 * Each device has operational data and environmental data
 * One-time setup that is YAML and sharable
 * Standardization of environment is one of the most important aspects of automation

= Robotics - Data Format=

Data storage formats are very important:
 * Make the data usable today
 * Make the data editable today
 * Make the data survive into the future; no obtuse or ridiculously hard to learn format
 * Make the data sharable

Watch out -
 * Computer scientists love inventing data formats
 * Leading to the problem of... too much complexity
 * Vendors love inventing data formats
 * Leading to the problem of... patent / proprietary lockup
 * Open formats are important for innovation and scalability over long term

=Robotics - Data Format=
 * Need to represent robotic environment and bio-protocol actions
 * Computer Scientist response: "Oh, but we can simplify that with XML"

  351 Valerie Corning pZero 3550 2007-10-04T11:05:38.5170000+02:00 20 20   406</ID> <DateDispensed>10/4/2007 12:45:40 PM</DateDispensed> <Wells> <Well WellNumber="15"> <WellVolume>75</WellVolume> <WaterVolume>37.5</WaterVolume> <Drops> <Drop DropNumber="3" ProteinFormulation="Erbeta+Org2" ProteinVolume= "0.25" WellVolume="0.25" /> <Drop DropNumber="2" ProteinFormulation="BufferC" ProteinVolume="0.2 5" WellVolume="0.25" /> <Drop DropNumber="1" ProteinFormulation="Erbeta+Org1" ProteinVolume= "0.25" WellVolume="0.25" />

Just say No to XML

 * Much too difficult to edit this
 * Much too difficult to read this
 * Much too difficult to learn this
 * Now needs extra files to describe the custom format
 * XML, HTML, anything that ugly, is a Bad Idea

=Robotics - Data Format - Device data=


 * Each device has control commands (a dozen or several hundred).
 * Storing that data in an easily readable and editable format is very important for software re-use.
 * YAML Format is Human readable, Human editable and Computer Readable
 * YAML allows references to prior definitions and either simple or complex assignments

%YAML 1.1 --- # Fialab-Microsia address: syringe: A     valve: C      peristaltic: D      external: B  valve: send: # delay after cmds 100ms - 1 sec NP_SET: desc: set number of physical ports opcode: NP             args: - 1                 - numport:4-12 redundancy: 2 # send cmd twice delay: 100 # ms             recv: ok: ~ err: ~ NP_GET: desc: get number of physical ports opcode: NP             args: - 0             redundancy: 2 # send cmd twice delay: 100 # ms             recv: ok: ~ err: ~ peristaltic: send: # delay after cmds 100ms - 1 sec SET_SPEED: desc: set pump speed opcode: G             args: - 1                 - speed:0-100 delay: 100 recv: ok: ~ err: ~ SET_DIRECTION: desc: set pump direction, direction:1=counterclockwise or 2=clockwise opcode: W             args: - 1                 - direction:1-2:default=1 delay: 5000 recv: ok: ~ err: ~ syringe: send: # delay after cmds 100ms - 1 sec INIT_ALL: desc: init all (both) pumps opcode: _Z0R args: - 0             delay: 100 recv: ok: ~ err: ~ SET_INPUT: desc: position valve in for given pump opcode: /$1IR args: - 1                 - pumpnum:1-2:default=1 delay: 100 recv: ok: ~ err: ~
 * Made to be readable and editable
 * Made to be scalable
 * Contains all definitions of a device's operation
 * YAML is Best current practice for readable data format

=Robotics - Data Format - Environmental data=

%YAML 1.1 --- version: 2009-09-04 tecan: genesis: points: roma0: magnet-hover: '14056,1850,980,1800' magnet-place: '14056,1850,687,1800' sampletray-hover: '14057,2828,980,1800' sampletray-place: '14057,2828,582,1800' shaker-hover: '1780,3569,1535,1800' shaker-put: '1780,3569,865,1800' shaker-take: '1780,3569,865,1800' shakerlock-1: '1762,1177,1535,900' shakerlock-2: '1762,1177,815,900' shakerlock-3: '1191,1177,808,900' shakerlock-4: '1762,1177,815,900' shakerlock-5: '1762,1177,1535,900' shakerlock-6: '1780,3569,1535,1800' shakerlock-hover: '1780,3569,1535,1800' HOME1: '11165,2525,980,1800' ...
 * Each device has environment data associated with operating it.
 * Points in space
 * Containers
 * Solids, liquids, gases
 * Storing that data in an easily readable and editable format is very important for software re-use.
 * YAML Format is Human readable, Human editable and Computer Readable

=Robotics - What's Running Now=


 * General purpose software to control FIAlab syringe pump and 10-way valve
 * General purpose software to control part of the Tecan arm
 * Network operation of the Tecan; run a bio-protocol from your desk
 * Get the current operational status in real time for current experiment
 * Solid foundation; much more to be done
 * Lab Automation mailing lists have already responded with high interest
 * 2 releases already made to the public internet software archive for Perl (CPAN)

The end user writes a simple Perl application to control all devices and robotics.
 * Perl is the #1 bioinformatics language.

sub Main { $hw = Robotics::Tecan->new(         connection => 'network,Robotics::Tecan::Genesis,genesis0',          token => 'M1',          serveraddr => 'heavybio.dyndns.org:8088',          password => $ENV{'TECANPASSWORD'});

$hw->attach("o"); $_ = $hw->status; exit -2 if !/IDLE/i;

# Load worktable $hw->configure("client-traymove1test.yaml");

my @path = (         "shakerlock-hover",           "shakerlock-1",          "shakerlock-2",          "shakerlock-3",          "shakerlock-4",          "shakerlock-5",          "shakerlock-hover"          ); checkok $hw->move_path("roma0", @path);

checkok $hw->move("roma0", "shaker-take"); checkok $hw->grip("roma0"); checkok $hw->move("roma0", "shaker-hover");

checkok $hw->move("roma0", "sampletray-hover"); checkok $hw->move("roma0", "sampletray-place"); checkok $hw->grip("roma0", 'o', 120); checkok $hw->move("roma0", "sampletray-hover"); checkok $hw->park("roma0");

checkok $hw->park("liha");

=Robotics - What's Running in "Upcoming Weeks"=


 * General purpose software to control FIAlab
 * General purpose software to control Tecan arm & Tecan liquid handler & Tecan peripherals
 * Network operation of all of the above; run a bio-protocol entire flow from your desk
 * And get the data results back from MATLAB in real time
 * Advanced error handling for Tecan bio-protocol problems: automatic re-trying
 * Complete suggested bio-protocol as initial high-throughput run

Medium Term Goal


 * Prove system across multiple & varied bio-protocols
 * Swap out the existing devices for upgraded or custom versions.
 * Use simple language to control the bio-protocol
 * Improve maintainability of both software applications & hardware designs
 * Allow remote users to access local hardware (as a lab service)
 * A new Robot Scientist is born

Long Term Goal


 * Run a complete bio-protocol using the English language
 * Clarification: limited bio-latin-english

=Robotics - Benefits=


 * Write scripts into the framework to build up scale and reuse; rather than typical stand-alone scripts which are 1-time throw-away
 * Allows many devices to be controlled from the same user program ("hardware integration")
 * Allows device operations not supported by the vendor
 * This includes much better error handling & re-trying
 * Allows abstraction of the devices: swap out one device for another
 * Removes vendor lock-in, creating more competitive forces to drive innovation among various devices
 * Plug in new custom devices, can quickly operate with same user program
 * Pipe data to/from MATLAB, the web, the wiki, .csv, others ("data integration")
 * Network operation
 * The controller PC will miss fewer commands since vendor application does not take CPU time
 * The user PC doesn't have to be Windows to run a device (many benefits there)
 * The user can access device status and device output from anywhere (lab or home)
 * Built-in network security
 * The user can program complex algorithms using multiple devices, creating a control system with feedback to optimize a protocol or make arbitrary decisions
 * The "Robot Scientist" was claimed to have identified new targets "on it's own" - using data feedback and prediction; smart algorithms should be possible

=Robotics - Benefits=

Example devices insertable into the Robotics 'flow'


 * Inkjet piezo-heads - pL or nL droplets, gradients, etc
 * Alternative substrates vs. well plates - CDs, other?
 * Millifluidics (Peter)
 * Physical handling (repetitive tasks) - stuff with motors for plate fetch & store
 * other?

=Robotics - Protolexer=

Let's Do Something "Simple" : Revisited

Just Use English.


 * Feed the protocol directly to the computer.
 * Computers are smart enough.
 * The software knows what devices are attached or available on the network.
 * Integrate the devices together into a long chain of bio-operations.
 * Complain if bio-protocol requires device that is not available (Dependency checking.)
 * Bio-protocols have fairly standardized formats and standardized language.
 * Or can be, with human editing of the English, and a human quickly verifying the "compiled" result before robotics operation.

If the computer can't understand a bio-protocol, then the bio-protocol is ambiguous and should be re-written anyway

http://biosx.com/88proof/synthetic_biology/blog/wp-content/uploads/2009/06/protolexer1.png

=Robotics - Data Format - Bio-protocols=

%YAML 1.1 --- protocol: Mate-Paired Library Preparation for Sequencing methods: - &standard-purify purify with column: - &cp1 Add 3 volumes of Buffer QG and 1 volume of isopropyl alcohol to the sheared DNA. If the color of the mixture is orange or violet, add 10uL of 3M sodium acetate, pH5.5 and mix. The color turns yellow. The pH required for efficient adsorption of the DNA to the membrane is <= 7.5. - &cp2 Apply 750uL of sheared DNA in Buffer QG to the column(s). The maximum amount of DNA that can be applied to a QIAquick column is 10ug. Use more columns if necessary. - &cp3 Let the column(s) stand for 2 minutes at room temperature. - &cp4 Centrifuge the column(s) at >= 10,000g (13,000 rpm) for 1 minute, then discard the flow-through. - &cp5 Repeat steps 2 and 4 until the entire sample has been loaded onto the column(s). Place the QIAquick column(s) back into the same collection tube(s). - &cp6 Add 750uL of Buffer PE to wash the column(s). - &cp7 Centrifuge the column(s) at >= 10,000g (13,000 rpm) for 2 minutes, then discard the flow-through. Repeat to remove residual wash buffer. - &cp8 Air-dry the column(s) for 2 minutes to evaporate any residual alcohol. Transfer the column(s) to clean 1.5-mL LoBind tube(s). - &cp9 Add 30uL of Buffer EB to the column(s) to elute the DNA and let the column(s) stand for 2minutes. - &cp10 Centrifuge the column(s) at >= 10,000g (13,000 rpm) for 1 minute. - &cp11 Repeat steps 9 and 10. - &cp12 If necessary, pool the eluted DNA. - &bead-purify purify with magbeads: - &bp1 Add 100uL of DNA to 95uL of magbeads. - &bp2 Vortex at 1,000RPM for 1 minute. - &bp3 Incubate on magnets for 300 seconds at room temperature to allow DNA to                bind to beads and beads to settle. - &bp4 Remove supernatant while beads are magnetized. - &bp5 Elute while beads are magnetized using 100uL of EtOH. Pause for 90 seconds during each wash to allow beads to settle. Allow EtOH to evaporate until beads are dry and cracks are visible in the bead surface. - &bp6 Resuspend with 15 uL Buffer xx to resuspend beads. - &bp7 Vortex at 1,000RPM for 20 seconds. - &bp8 Incubate for 200 seconds on magnets at room temperature. - &bp9 Save the eluted DNA.
 * Easy to edit (it's text with indenting)
 * Easy to read (no crazy formatting words)
 * Very easy to share
 * Standardized format
 * Self-contained, no extra files needed
 * Mostly easy to learn
 * Easy for computers to read, process, write, share
 * Defined in YAML

=Robotics - Competing Method: BioStream (MIT)=

"Towards a High-Level Programming Language for Standardizing and Automating Biology Protocols"


 * "Abstraction Layers for Scalable Microfluidic Biocomputers", William Thies, John Paul Urbanski , Todd Thorsen , and Saman Amarasinghe, Computer Science and Artificial Intelligence Laboratory, Hatsopoulos Microfluids Laboratory, Massachusetts Institute of Technology


 * Defines "high level" (like Java/C++) language for describing protocols
 * Can take some standardized English protocols as input
 * Can automatically generate the "high level language" for the computer
 * Can output a human-readable clean English bio-protocol
 * This is a "Validated-clean" version of original bio-protocol

=Robotics - Competing Method: BioStream (MIT)=

BioStream "clean English output" for DNA extraction from tissue protocol''' http://88proof.com/biotech/img/Biostream-DNA%20extraction%20from%20tissue%20protocol-English-Output.png

=Robotics - Competing Method: BioStream (MIT)=

"Towards a High-Level Programming Language for Standardizing and Automating Biology Protocols"

BioStream "high level language" for DNA extraction from tissue protocol'''

http://88proof.com/biotech/img/Biostream-DNA%20extraction%20from%20tissue%20protocol-Source-Output.png