Ted will come by at supper time.
Alison: Android workshop.
Ted: Let's do it.
Brad: Portal where.
Ted: Room 118, building 23. Come in turn right. Every day, video taped. On website.
Open to public.
Merkle: Next speaker. Andrew Minor. UC Berkeley.
Teaches the course on nanomaterials.
NT CL3 Nanomaterials (Andrew Minor)
WhenTue, July 6, 11:30am \u2013 12:30pm
11:21 Thanks for having me.
Nanomaterials. Ralph gave you an intro.
If something is not clear, stop me.
Take perspective of material scientist.
Lab in Hurst library building -steel was studied.
Now diversified widely - biomaterials
Nano materials - what they are and how properties change with size.
Materials science -
What makes a material a nanomaterial.The size.
Some characteristic length of 10s to 100s of atoms.
Laws of physics don't change.
course on quantum effects teaches you they roll up into newtonian physics.
Broad/easy definition - nano materials exhibit size effects.
nanomaterials exibit size effects
\"size effects\" when the size of the material affects the properties
Key thing about being small
More surface area.
As you get smaller and smaller in your grain size, the percent at the volume increases and this makes the properties change.
- Example of a size effect is melting point
If you have a particle very small, melting point may be at room temperature
(See Buffat and Borel, Phys Rev A, 1976)
\"property changes the size\"
- Thin Films- nanomaterials common in technological applications
Sizes have real life implecations. Parts in electronic circuits are already nano-materials. They are thin film.
- Real industrial consequences of size effects
Traditional microchip in processors is a MOSFET,
usually the way they make things is they make things smaller. But what actually physically stop you to keep scaling things? When you get smaller and smaller, the properties change and thus different materials need to be used.
By the time you get down to 5 atoms of something, is it really the same as the origional material.
(See Muller. D. A., et al., Nature. 1999)
- Strength of Materials
Size is important.
Young's Modulus - Slope of elastic loading (GPa)
Single-walled carbon nanotube1,000+145,000,000+
Yield Stress - Actual Stress at which something breaks (MPa)
Pascal is a unit of stress and this is stress over an area.
- Most useful solids are crystalline
- Dislocations are missing rows of atoms which allow materials to deform mechanically w/out breaking.
- Explains why metals are ductile
Crystal is actually not crystalline, its amorphous
Dislocations are key to making metals ductile/brittle
- Early studies of size effects
First person to see size effects is Leonardo Da Vinci:
He took a very simple experiment, took an iron water and put a weight at the bottom to see how long the wire could be before it broke.
Half piece would break at half the weight as the long piece. The longer the wire the smaller the possibility of fatal flaw.
- Early tests on size effects (cont.)
Really small wires you can bend forever and they don't break.
- Ideal strength - the stress required to plastically deform a \"perfect\" (defect-free) infinite crystal = G/2pi
-Experimental strength = G/1000
- What happens when you don't have any dislocations?
Any material (except semiconductor silicon) has many dislocations in it
[ref Tensile Strength of Whiskers]
- Grain Boundaries
Imagine you have a cube of water in an ice tray, you can have the possibility for:
1) one grain boundary -> solidification starts at one place (same way as metals solidify)
2) 2+ grain boundaries -> solidification starts at multiple places
These grain boundaries are important for how a material acts
- Grain Boundaries and mechanical properties
yield strength ~ grain size ^ (-0.5), totally imperical no one knows why
-Metallurgical Techniques Borrowed from History-folding techniques to decrease grain size
Japanese sword making - folding the metal over and over and over and the grain size decreases
Properties change on grain size effect.
Stress vs Strain Diagram: larger grains break at lower stress, but are more ductile
- Focused Ion Beam Micromachining
\"Because of the sputtering capability, the FIB is used as a micro-machining tool, to modify or machine materials at the micro- and nanoscale. FIB micro machining has become a broad field of its own, but nano machining with FIB is a field that still needs developing. The common smallest beam size is 2.5-6 nm.\"
What's making materials smaller solely based on their size?
In Oregon Focused Ion Beam created
Ion Beam is like a sandblaster using ions. You can cut anything you want down to 7nm resolution. Semiconductor industry originally used this for mask repairing.
One of the inventors of this machine said that years after he'd been selling this to Intel for $1M a piece, they told him each machine they bought saved them $1M a day. That's how useful these were for editing masks.
Mike Uchic takes one material and just mills out different size pillars.
Same material, just different sizes, and then plots out the macroscopic yield strength.
From about 40 microns and up, you're going to find the same number every time- the number you'd find in wikipedia.
Log-log plot: double exponential strength increase as size goes down
People have done this for lots of different materials.
- Pillars - smaller is stronger
Copper, vanadium, magnesium, nickel - doesn't matter. Below 40 microns you get this hardening effect, purely based on size.
- Sample size vs defect density
Imagine a bulk sample of metal. There are lots and lots of defects (represented by T structures). If you start to pull or push this material, somehwere in the material a defect will move a couple atoms one way, and the material will start to deform. The bulk strength is where somewhere, one of those defects gives way. Now we make it smaller, so there are fewer defects. OK. Now we make it so much smaller that it's smaller than the characteristic length of the defect. Now it's a statistical thing - is there a defect anywhere in the material? If not, you have to create a defect to deform the material.
- In situ TEM mechanical probing at NCEM
\"Transmission electron microscopy (TEM) is a microscopy technique whereby a beam ofelectrons is transmitted through an ultra thin specimen, interacting with the specimen as it passes through.\"
National Center for Electron Microscopy
What's an electron microscope? It's very similar to a light microscope - in fact, optically, it's almost exactly equivalent. The reason it's useful is that a light microscope stops being useful at about a micron. Because the wavelength of light we see is about a micron, and you run into the Rayleigh limit. Electrons, though, their wavelength depends on the voltage of the electrons, and you can adjust that voltage and see all the way down to individual atoms.
- Electron Microscopes
You have a very sharp tip - tungsten usually - down to a few atoms, and you get a very high voltage, few hundred thousand volts, and it pulls off an electron that shoots down.
- Electron Optical Characterization: key tool for nanomaterials research
You can do things with electrons simultaneously that you can't do with light. You can do spectroscopy - distinguish carbon from nitrogen - you can image, and you can do diffraction and see the crystal structure. But in the end, the most important thing about electron microscopy is that you get spatial resolution down to 50 picometers.
- In situ mechanical testing
You have to put the whole laboratory inside a very small space, keep a vacuum, etc....why bother?
You want to figure out why something's deforming. Video showed of aluminium deforning.
[movie of aluminum being dented by diamond]
Grain boundaries can move at small scale.
Another reason to do something inside an electron microscope is to test something that's so small, you can't find it afterwards.
- In situ mechanical testing (2)
Cadmium[?] compression - another deformation video. Being able to monitor your test is really important, because something might have just slipped away instead of breaking. That's why we do it inside a microscope.
This is 100 nm - very small scale.
Being able to monitor your testing at a small scale is important.
- Cu nanocompression overview
This is a red blood cell, about the right size scale. We are basically doing tests at size scales that are within a cell. We're getting quanitiative information while also being able to see what's inside the test.
- Cu  Results
We're getting smaller, smaller and smaller. . . and we're running out of available dislocation sources (???)
- Cu  Single slip
The key is that you only have one type of dislocation.
Multiple slip and barreling.
- Mo pillar comparison w/ and w/o anneal
[movie of shear]
You do this at a stress of 16 GPa.
The ideal strength of the material. The material liquifies. Instataneous liquification across the plain.
- Understanding Mechanical Size Effects
Computational Modeling - (See Gouldstone, Acta Mat, 2007)
Merkle: small things stronger.
Large things without defect, stronger.
Large things built from many many small things should also approach theoretical limit.
Q: Tony: Recycling?
Lead free solder
Types of steels
Brad: Defects are place where there could be failure. Why not all edges?
Edge is very easy place to start defects. Also place get dislocation to zip right out.
Salim: Andrew will stick around.
Lunch. Soccer on screen.
Bio tech workshop this afternoon.
Questions about life tonight. 7:30 - will be wine.
Thanks for the day this morning.
Please take cans/wrappers with you as you leave the room.
=============== DURING-presentation notes
Color card definitions:
Green - Agree
Red - Disagree
Yellow - Slow down (content) / I don't understand
Grey - Speed up we know this stuff / Move on to another topic
hands moving apart (sideways) - speak slower, speak more clearly
Link to backchan.nl for questions:
9:00am\u201410:00am Morning Session I
10:15am\u201411:15am Morning Session II
11:30am\u201412:30pmMorning Session III
GSP10 Book List:
Either remove placebo effect people or know who they are.
Put some on the drug, osme not, do nested study.
Problem of research.
Interest in 23andME. Russ .. from Stanford.
Get all ph.gen studies. He was important advisor.
Design types of beneficial studies.
Big issue: you take drug, have good/bad reaction. More fundamental: how do you define disease.
Our definitions are 19th century. Symptoms.
Parkinson's - different versions.
LRRK2 - with certain variations - Sergi and mother both have gene.
Other people with disease - not associated.
Could be different drugs. Different treatment profiles if you understand the genetic profile.
Bryce: To have a disease is to have a genetic trait. Create unnecessary fear?
If portrayed correctly. It is a risk. Not absolute. Must be written in the right way. If I have this gene, my risk is increase by a certain amount.
Find people with gene, never got disease.
Man, age 70, whose father had Parkinsons.
What stops it? Lifestyle?
You have your genetic profile.
you are at liberty to fill out survey on line.
Punch - like.
Kind of fun. [ref]
Things you have been diagnosed with
Building out phenotype for you,
early association. Can you smell asparagus in your pee?
Two asparagus spears joke, \"Can you smell the humans in your pee?\"
150 studies going on.
So many people, over 50,000 now.
Hope more disease related publications.
Eriksson et al., \"Web-Based, Participant-Driven Studies Yield Novel Genetic Associations for Common Traits\"PLoS Genet 6(6) http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1000993
Having been at that for 4 years.
User interface had really evolved
Look at original - it was hillarious / antiquated.
Had mission now to develop concept in not-for-profit way.
Always get profit question.
In HC space, if for profit -> distrust.
Not customer base, but outside.
Still, lingering questions about model.
Brainstorm research foundation (www.brainstormrf.org/)
Autism, bipoloar, diseases of the mind. Challenging.
Need to focus.
Model where people can join the foundation.
Tools to track own brain health.
Are there markers for dementia?
Are there markers for decline, poor mental health? Make available for research.
Different brain exrecises. Do they work?
Physical exercise is better.
I'm a runner myself. Can change whole perspective.
Tremendous number of calls. They think I just give out money.
Can project manage things.
Foundation 2.0 - no overhead. All money goes to specific projects.
Alex Pedilla, California State Senator, SB482 Personal Genome Tests http://info.sen.ca.gov/cgi-bin/postquery?bill_number=sb_482&sess=CUR&house=B&site=sen
\"[This 2009 bill would have provided Californians the consumer and privacy protections necessary when they seek to obtain information about their unique genetic makeup and hereditary predispositions from companies offering these services. HELD IN COMMITTEE]\"
Old/antiquated. Paternalism needs tobe weeded out.
Community outreach. What we are proposing in a new bill.
Room for change.
Meeting around that topic. [ref]
Q\"Bryce: Agree on legistlation.
Getting big enough chohrt. P Obtaining consent under the structures of the Declaration of Helsinki \"Ethical Principles for Medical Research Involving Human Subjects\" http://ohsr.od.nih.gov/guidelines/helsinki.html)
Other countries may not have the capacity to process that much data.
Paradigm of informed consent can stand?
Early on, had form. Very straightforward to opt into research or not.
You are joing 23andMe and we do conduct research.
Not clear cut enough.
Commercial ERB (ethics review board)
No ideas what survey we would put out to people.
First meeting, early '07. My daughter has a genetic disorder.
We've got ERB approval now.
New consent form launched last week.
ERBs are trying
Rolling consent is fascinating.
(see 2006 interview genetic ethics: http://bioethics.georgetown.edu/pcbe/transcripts/nov06/session4.html)
People can change their minds.
Coming up with ways for people to change their minds.
What would be models for that: enought privacy and yet contribute to research.
Q: You sais anon is mythn. Yet de-identified - hard to connect to person.
Rolling consent means wyou want to get back in touch.
Conflict - maintain consent .. practical conflict?
At least 23 very closed model. Not open to public domain.
Even in company, we store personal information - credit card - separate from genetic data. Only scientists have genetic.
IRB ? approval
i think we have decided to just take this model for now.
ee how it goes.
News feed created uproar. Will shift with time.
Q: Julielynn Wong/ Canada/US: Modern age of social networking can give response bias (but large numbers). How are stat dealing with tat?
Clean up dataset by analysis.
Do you have blue eyes yes/no.
If conflict, their data may be ignored. Are they giving what we think is an accurate answer.
Cohort changes every day. P value changes. Plot is interesting.
Can you sprint? Frame of reference. I kicked butt in 100 yard dash.
My dad. Baseball player. Didn't have even one copy of the gene.
You almost have to have that gene to compete in the olympics.
Curvy/wavy hair - image comparison is helpful.
Certainly room for error.
Is larger less clean better than pristine?
Wikipedia vs encyclopedia
Tools - without preconceived notions.
Q: Derek/Canada: Can you talk about how Brainstorm foundation is going to include external data sources. Integrating with George Church's 1000 genomes project https://secure.wikimedia.org/wikipedia/en/wiki/1000_Genomes_Project
Capability of uplaoding
DNA nexus model [https://dnanexus.com/]
Works across platforms. Make it universal.
Then I don't have to raise money to get them sequenced. REally helpful.
Q: Raymond: Loved that data on chip is open.
Almost anyone else who has done things commercially has not done that.
A lot of debate?
Leting people see data?
Right off the bat. We wanted our own data.
They paid for it, they should have it.
onus on us to make interface place where they want to keep their data.
Make interface interesting and usable.
More coming out.
Updates all the time. New rports on weekly basis.
People think - log in, see what they see, So not true.
Want to make suer everyone is aware of that.
Chiara: You don't get data at lab. I pay you to know what is my genetic info. but if I am a researcher, I cannot get. It is not really open source.
Your data beling to you personally. If you download data and give to researher.
Bryce: Can researcher come to you? Ask you to send request for consent?
Main thing is if customers are OK with that.
As you go into the community.
Over 4000 different conversations, back to 2006.
If everybody saidk Hey, we like this guy down at UCLA, let's send him our data, you could do that.
BrycE: 300 person dataset. If people are mallicious or ... giving data not your own.
More rare diseases. If interesting from fresearch perspective ..
Surveys article on Wired cover this month (http://www.wired.com/magazine/2010/06/ff_sergeys_search/all/1)
How quickly we could come to similar conclusions.
Guy had answer in 10 minutes.
What questions would you like to pose? Proactive. Keeps data very secure.
10:09 Break [23 on pad]
10:17 Lori Giver (http://www.codexis.com/)
- The Evolution of Directed Evolution
First things first. Evolution in the laboratory for industrial purposes.
- Corporate Phylogeny
More general talk for this audience. How one person can have impact on the world. Zaffaroni - I'm a local girl, born in 1968. Very different area. Biotech grew up here organically. Zaffaroni founded ALZA. In blue shows how companies were acquired.
- Alejandro Zaffaroni quote 1
One of my colleagues had this great quote. Be exploring new areas, fail safely. Found another quote:
- Alejandro Zaffaroni quote 2
(AZ: Biotech Hall of Fame: http://www.chemheritage.org/exhibits/biotech/zaffaroni.html)
In looknig for people to run new compnaies. Broad perspectve.
- Codexis Technology
We are focused on optimizing bio catalysts.
Almost all talk about individual enzyemes.
Why / how / how changed in 20 years / where 5 and 10 years forom now
Not a pharma co.
Active campaign in biofuels.
Exciting new projects I won't be able to talk about now for carbon capture.
- Why Biocatalysis?
Why use rather than chemical process. Variety of reasons:
Not extreme temperatures.
If you can get reactoin done for you, will make the desired compound, not as an enantiomer[?]
Rarely something they can throw in, see how it works.
- Sources of Biocatalysts
Look in natural diversity. Lot of activities inearly 90's
People out there digging interesting organisms out of hot spring.
Extreme Ph, high temp, low temp.
If looking at extremes, know what is extreme here on earth
Went to school in Indiana [ref]
to work on that.
Second approach: build enzyme yourself.
We don't know enough about how they fold and function.
Protein Eng - evolve to what you want it to be.
- Protein Engineering
Most don't do what you want. Traits need to be improved.
High fructose corn syrup use.
Dfficult - 300 amino acids. Basically a lot of ways to modify.
How is it possible to screen millions and millions of variants.
Lot of ways to bring it down to a managable problem.
Ways it may not be doing what you want it to do.
Try to make it do what you want.
We just don't understand enough.
Screen for those you want. Keep building on that.
- Approaches to Protein Engineering
A statistical model that correlates sequence with function.
WHAT THEY DO NOW: They sequence variety from the process and code for that. For every gene we have they will know how active it is or its relativity to its parent.
What is the mutations impact on the gene status??
- ProSAR Model: Alignment View
Color coded ranking of correlation coefficients
All you can do is try to replicate the key parameters as best you can.
- Directed Evolution Technologies
OK, after ProSAR analysis, we've changed how we do our process.
We are thinking about mutations all the time.
Which are most beneficial.
Designing new libraries to keep building on that.
If I take this mutation and stick it into that gene, it may help, but not sure.
Always designing libraries, not trying to make one perfect gene.
- Evolution of Directed Evolution
This changed our view of how recombination works and how to make it work more effectively. Which are beneficial mutations even if they are in genes whcih are less fit than we started with.
- Modern Biocatalysis Paradigm
Changed to new paradigm.
Design around available - > design process you want and we'll build biocatalyst.
- Going Green Keeps Getting Easier
This compound, C5, they were sourcing, put into product.
Designed 3 enzyme process.
Convert to alcohol, then ...
Won award. Big publish for us.
- Chemical Cyanation of ECHB
Pfizer doing that with process from Mitsubishi
You can make a whole lot of things you don't want.
Volume fractionalization needed. Better to make ti more economical.
Q: Talk about Going Green slide more.
Went through really fast. Really important.
i'm leaving out that, at the time, big blockbuster drug.
Mfg of drug - if changed, require
- Enablement of Biocatalytic Cyanation
- Climbing Mount Improbable*
- Diversity Generation
- Automated Parallel SOEing (APS)*
- APS Formats
- APS Workflow Accelerates Enzyme Evolution
- Project Example
- Sitagliptin: A potent and selective DPP-IV inhibitor
- Merck\u2019s Current, 2nd Generation Process
- The Aspirational Process
- Problem and Approach
- Establishing Activity on the Pro-Sitagliptin Ketone
When you are trying to create biofuels, try to break down cellulose.
Chop up, get slurry. Most sugars not available.Working on whole suite of enzymes.
Feed organisms to create second generation fuel molecules.
Whole area scaling to work on genome level.
Kind of same level as gene shuffling.Using omex? technologies.
Lot of Codexis and Merck collaboration.
11:13 5 minutes of quick questions.
Q: Tony: How far are we from doing eugenics for humans?
I think we are doing genome evolution at microbial scale.
Yeast. Starts to rival human genome size.
Which ones are beneficial mutations. You could do them on any type of genome. Whether you would or not is another question.
We have been able to expand out.
Right now, it is not infeasible.
Take microbe which makes ethanol.
Train a yeast to use zylose. A few changes have to occur.
Just train it.
Do full genome sequencing and produce variants.
Moving on to humans, it is just more sequence.
Talk before - will be affordable soo.
Q: Sasha: Most of execution is done in controlled settings. Biofuels.
Unintended consequences, environmental space. Mutations, reactions you didn't expect.
These 5000 l reactors ARE reactors.
Can you grow algae. It needs to be OUT.
They need light.
Generally, right now, there is no release of recombinant organisms.
Not in the projects we do or the projects we plan to do.
Huge opportuniteis there, but engineer so they can't compete with wild.
have to be hampered. Keep them from exploding out wildly.
Q: Chiara: If I understood well, when you test on plate, why don't you usea 3D system. If not, why not?
We did a it of work awhile ago. Really interesting. 430 O plates stacked.
Visual indication of which working
Right now, we are not needing to push number beyond rountine.
Haven't been pushing teh 3D approach.
Depends on screen and what you are trying to look for.
Q: Erez/Israel: Thank you for very interesting talk. invented something similar to PROSAR 8 years ago. Interesting to see it again.
Revisco. Part of photosynthesis. Passive enzymes. Project with Rio Tinto. Can you improve photosynthetic organisms. Very challenging.
A tetramer of octomers. Almost every residue - very easy to cripple that enzyme. Hard to improve it.
Might be worth another look now.
How they might use - photobioreators or what they would do.
Q: Screening. Specifically which is going in silico.
We don't do any in silico for enzyme function.
Others try that. Go into lab and synth top variants.
Even those suggest, interesting starting points for directed evolution.
Lot of in silico screening - didn't need to do that.
Raymond: Thank you Lori.
Couple things as takeaways:
directed evolution - don't need to really know how it works to improve it.
DNA synth still a blocking point
Improve through automation.
helpful review article(s):
Tracewell and Arnold 2009 \"Directed enzyme evolution: climbing fitness peaks one amino acid at a time\" http://www.che.caltech.edu/groups/fha/cara.pdf \"Evolution has created numerous specialized enzymes that function in living cells to catalyze the chemical reactions of life. Their specificity is tuned so that they generally do not tread on each other\u2019s toes. But that does not mean their specificity is absolute: a recurring observation has been that many enzymes have weak activity on non-native substrates and that directed evolution can amplify these weak activities.\"
Up next: Nanotechnology
NT CL4 DNA Origami (Paul Rothemund)
11:32 Merkle: Everyone back for next talk.
Brilliant, fascinatingand delightful.
You will not want to miss this speaker.
Folks in the lobby - come back, come back ....
Honor to have Paul Rothemund (properly pronounced Rote-a-mund)
Grabbed cover with smiley face.
Labs said, \"You can make a smiley face. You can make anything\"
Those in the field were kicking themselves because they said \"If only I'd realized you could make smiley faces!\"
Giving talks all over about the inspiration of how he came up with this idea
It's a great honor to say we have Paul here to explain what's going on.
- Beyond Watson and Crick: programming DNA self-assembly towards integration with nanoelectronics
This event is about thinking seriously about the longer term. A kind of event we should have more of, but not a new idea. You should read \"Future Shock\" by Alvin Toffler.
A lot of the lessons of that book are applicable today, so I'd encourage you to think about them.
The Singularity is a science fiction concept. I'll indulge the science fiction desires for a second.
- Before I die ... I'd like to see humans:
I'd like to be able to program biology with the same facility with which we program computers.
We'd like to be able to create whole organisms to specification.
I'd like us to find ET life.
I'd like us to create AI.
I'd like us to find ET *intelligence*.
I'd like us to create an equitable social operating system, allowing every person to live up to their full potential (i.e. world peace).
Oh, and a pony.
What if anything does nanotech have to do with any of that? And do our predictions for the time course have any bearing on reality? Are we calibrated for that?
I work on things that go towards the first two -programming biology and creating life.
A comment about programming vs Programming. Little-p programming is like a VCR or a chemist specifying a particular structure. Some finite set of possibilities. Specifying when your TiVo is going to turn on, that sort of thing. Capital-P Programming is the computer science sense - anything you can do with a computer - Turing-complete programs.
- \"Synthetic biologists\" \"Molecular Programmers\"
A whole bunch of people are synthetic biologists, and a bunch of people that don't know what they should be called but synthetic biologists let us hang out with them. \"Molecular Programmers.\" Mostly biomolecule-oriented. I want to give the sense that a lot of people are working on this stuff. I'm on the molecular side - changing the actual substrate for the things we're building.
The field I'm part of is DNA technology and DNA computing.
- For 30 years, Ned Seeman and his descendants
Seeman succeed in making enormous complex DNA polyhedra. That's one subfield of DNA nanotech: how do you encode in molecules strucutre.
My part of the world is function and dynamics with DNA.
- (p)rogramming Function and Dynamics - 2000, physicists
Bernie Yurke made a little DNA tweezer - 50 piconewtons of force - pretty respectable on the atomic scale.
Then others made it into walkers and little Rube Goldberg machines.
- (P)rogramming molecules to perform computation - 1994, computer scientists
The third strand is programming molecules to perform computation - DNA computing. Was heralded as a competitor to electronic computing. What happened? Many ways to compute with DNA, but none were competitive with electronic computing. For all the methods anybody could dream up, it didn't seem we could come within 4 orders of magnitude of electronic computers.
But at a molecular level, a little computation goes a long way. Cells have to decide what kind of cell to be, and that's transacted by a little molecular computer inside your cells. Cell divison - an elaborate choreographed dance - regulated by networks, possibly modelable as circuits - are mediating that as well. Our interest now is to do just a little computation in DNA and use that for other little molecular machines.
Q: How do inputs come? How are they processed?
Set of sensors sense food. Transducers bring that into cell.
Then there's a complex cascade of proteins that [...] motors on the end of the bacteria. In this way bacteria can move up chemical gradients, getting closer to a food source. It costs energy to make the machinery to digest certain kinds of food. If there's a new kind of sugar around you need to switch your metabolism. So you have sensors for those sugars, and a regulatory network that interacts with the genome, turning genes on and off depending on what's around.
Decisions made at molecular level that are just decisions. Not like a chess program.
It's a relatively short computation. Doesn't have the depth of a chess program or something like that.
- All three - structure, function and computation - are necessary for the creation of systems with the complexity of living systems.
So we're taking a crack at all three. The reason we use DNA is because, while we have a field of protein engineering, we understand DNA a lot better and can do a lot more with it. It's an approachable substrate to actually do engineering in.
- It's .. skip this
- Catalog of molecular parts (Goodsell, http://www.pdb.org/pdb/home/home.do)
There's a photosensor in cells. There's a light-harvesting complex. There are switches. Signals come in, there's computation that's performed, and it's decided whether to turn genes off.
There are girders - protein filaments that are quite stiff - and motors, proteins that walk along these protein filaments.
- DNA bricks - Symmetry sets the stage
DNA strands bind and wrap around each other to make this little brick.
They're single strands - not double helix.
Brick has four sequences on its edges. By giving each edge a different sequence, it's possible to program this object - in the little p sense - to make interesting structure.
In one case we programmed them to make crystalline tubes that are similar to those proteins in biology.
- [diagram] TTACG
- [photo of DNA bricks]
We could use two different ones to make walls with diagonal or perpendicular stripes...
The point is, we can begin, with DNA, to recapitulate some of the forms we see in biology. These DNA tubes are about as long as protein filaments, and almost as stiff. Nobody could make a de novo protein that would be anything like that. This is the power we see with DNA nanotech.
We can also make pretty much any wallpaper pattern you want - can design a little DNA brick for any pattern. But we're not just interested in repeating patterns - only good for very crystalline memory. I worked with Erik Winfree at Caltech to do Programming - embed a computer program in those DNA bricks.
- Algorithmic crystals that can simluate cellular automata
- [zoomed view] Pascal's triange modulo 2
4 different types of bricks in here. As they grow, they begin to compute. What they're actually computing is the form of a fractal - it's got errors, but - it's the form of a Sierpinski gasket. They're performing a binary computation and leaving this pattern in their wake. It's the binomial coefficients - a somewhat useful pattern. Much more complex than the periodic patterns I talked about before. As these little bricks self-assemble, rather than just making a boring wallpaper, you could use it to lay out a complex circuit. Won't go into how we're doing that, but that's one model.
The thing is that it turns out that this algorithmic self-assembly from Erik is useful for building very large objects that have some structure, but not useful for arbitrary structure at a small length-scale. You are grown in an algorithmic way. Genome - 3 bln base pairs - a process called development grows it into you, and the genome specifies you vs an elephant. You are much bigger than the genome that is used to build you. The genome gets decompressed into a very large structure. But at the smallest length scale, you want something with some random structure - maybe any pattern that you want to make - not suitable for algorithmic.
- So we can
- [long strand]
- [short stands]
My approach to arbitrary patterns is what I call DNA Origami. This is a long single-strand of DNA. A very small piece of a 7000 base-pair strand. You can imagine it stretching all thw ay around the room and meeting itself on the other corner of the screen. We use short strands of DNA to fold that very long strand of DNA. We call the short strands staples. The left half of the staple binds the long strand in one place, and the right half binds the long strand in a different place.
- [long strand bound by short strand]
So it enforces a constraint between them and binds those places together. The net action of about 200 of those short strands fold the long strand into some pattern - for instance a rectangle.
- [cartoon animation of molecular self-assembly of what it must be like.]
You can see as it goes on some double-helices forming...not complete double-helices, but they're held together by what are called crossovers. Where a strand is held by one part of the long strand but then jumps to another helix. In this way we make something that approximates the shape we want, but all the individual pieces are mostly making double-helices, which is what they want to do anyway.
We actually don't know that much about the detailed trajectory that this process takes over hours or days - they're probably binding and unbinding all the time.
- [animation of long scaffold shape - arbitrary shape - smiley face]
I wanted to show that this really could make arbitrary patterns - structures that weren't just convex - had holes, and other things. So I decided to make this smiley face. To give you an idea of the lengthscale. This is 100 nm across. 1/70th the width of a red blood cell. Very small. How do you see it? With an AFM (atomic force microscope).
AFM using a very small micromachine needle, dragged back and forth across a surface. (So we have to somehow get the DNA from solution onto that surface.) As the needle hits the surface we can measure the deflection. So we have a method - could have been done in the 50s - very low-tech, amazingly - for seeing what hapens.
- [100 smiley faces floating around]
How do get onto the surface? You wash the solution over mica, which is very flat. The DNA origami stick to it, and then you can take an AFM image of that surface.
- [arrangement of smiley faces]
Look at the image, and you see smiley faces. They aren't perfect, but 72% are what you wanted. You can actually see the path that the scaffold takes and you can see the little unfolded bit at the bottom.
- A catalog (again) [with smiley face overlay]
To compare to other things that we're trying to recapitulate in biology, we're actually beginning to engineer and make arbitrary shapes at that level. We cannot make things with the fine resolution of proteins. We can't begin to approximate that. But we're only about 10 times larger.
- a, 12.5mM MgCl2 [with triangular shapes.
If you make a structure that's more rigid than the smiley, you don't have this problem with the structure getting busted, and the yield is more like 95%. It's really a defect of the weak smiley face that we don't get a higher yield.
- image - looks like map - China! Lulu Qian
Even put Taiwan on there.
Completely replicable. Just made what she wanted.
- crosses, girders
Since people have done all kinds of things - 3D structures- William Shih. Rigid structural elements, crosses, girders of all kinds. They take about 2 weeks to fold, but you can bring all these helices together in this tiny space and make amazing things.
(see CAD Nano http://cadnano.org/)
- 3D structures - machine walking along protein track
The major thing that it's going to do is custom instruments for biology. Up here is a 3d shape that makes a little track. He's intending to use these as tracks for proteins. The goal is to use these as probes for what happens in actual biology. There are lots of questions about protein walkers. And one way to answer them is to make tracks with different steps. If there's a bigger distance, maybe it would walk faster. Or maybe it would fall off altogether. We can't engineer the tracks as proteins, but we can engineer DNA tracks. We can put protein walkers on them and watch them.
It turns out the protein engineering actually is good enough for this. But it gives you the feeling that this is the kind of thing we're going to do.
- Custom instruments for biology
Another example - a Japanese group -Endo and Sugiyama - they made a DNA picture frame. Two wires going across it are DNA double helices. One is short and one is long. The long is loose, the short is taut. And when you put this red enzyme that puts a mark on it, it will modify the loose one but not the taut one. Another enzyme that cuts, won't cut at the mark, but will cut the taut one. We've made this custom jig that allows us to probe the activity of enzymes, in a single-molecule setting. We can really begin to interact with proteins in a much richer way - answer complicated and subtle questions. One of the fantasies is that this will be the first thing biologists think of - that when they want to study what a protein's doing, they'll make a DNA origami custom instrument.
There is a CAD software for these things that you can download and play with http://cadnano.org/intro.html
- DNA boxes
- Lund, et al, 2010 (Stojanovic, Winfree, Walter, Yan)
Referencing a couple papers at the end - DNA strands hanging off origami, protein with flourescent markers, and arms that can cut the dna strands, and they can have proteins walking along the DNA origami, a hybrid protein-DNA walker. A synthetic system for studying artificial molecular walkers. People have even made walkers that pick up cargo and put it into a little package. We can talk more about that offline.
- Each staple goes to unique position
Let's talk more about DNA's patterning ability. You have a bunch of staples that will end up in a particular position.
So you can make patterns. You can make a snowflake pattern or the americas, and you can spell things - you might think it's just artwork, but artwork is exactly what we need for technology: photolithography for microelectronics is a way of taking artwork and scaling it down.
Wires and switches are printed in metals and semiconductors. If we have a better way to do that at a smaller scale we could make better faster computers. Fantasy of putting functional components on the DNA.
- each component
- Two dimensional organization of carbon nanotubes on DNA origami
We've started to make halting steps in that direction - nanotube transistors this way? Put nanotubes on DNA sequences that are complementary to bits of DNA origami. Should organize the tubes into a cross, which for the right kind of nanotubes is all you need for a transistor. We did that and you can use a special tool to write electrodes to it. You can connect up to it, and you can measure that you can switch. So we can make a transistor this way. [Aside: When we did this work, we were having a problem with aggregation, and this DNA ribbon made them more stable for unclear reasons. Potentially not necessary for this system.]
- The Design ...
- Benchmarking [slides not shown, skipped]
- Two dimensional organization of carbon nanotubes on DNA origami (with AFM image)
However, the problem is you make these things in solution and they go all over the place. The great challenge is that you have to put these on a surface to wire them up, and it's like taking a deck of cards and throwing them on the floor. You have to find them! It's an incredible challenge. And they cross-link, aggregate, all kind of things.
- Two dimensional organization of carbon nanotubes on DNA origami (Hunt and Peck)
These electrodes were written down to the nano-scale - had to be done tediously by grad students. They made these beautiful crosses but it's impossible to integrate them.
- One solution to placement
Another strain of research we have is putting DNA origami where you want on surfaces. I did that with a group at IBM Almaden. You can use standard lithographic techniques to make sticky patches, so when we pour the DNA across they'll go exactly where we want.
- IBM group
- origami on surface
We can convince ourselves that not only do we get one origami for each hole in the surface, but we get 95% of them in line and where we want.
- combining placement (registration) with positioning of devices
The overall picture is that where we are in the end, we're going to make specific shapes that will bind to specific sticky patches. Then later we'll add the components that will assemble on top of the DNA origami. So we won't have problems with the nanotubes causing the origami to aggregate. We won't have to worry about drawing these crazy roads down to the devices with microscopy - we could build it in a very standard way.
- IBM Hung /Cha. Now at UCSD
People have extended this process, and people at UCSD have already put gold nanoparticles on these triangles. Beautiful orientation. We're actually beginning to organize little nanodevices this way using DNA origami.
- [three shapes - rectangle, smiley, triangle]
I would ask - what makes DNA helices come together. Ralph? Ralph: hydrogen bonds. That's wrong. Not that Ralph is wrong, but it turns out to be wrong. In the textbooks they tell you the major reason. it's not hydrogren bonds. The facility with which he said hydrogren bonds - I wanted to bring this out, a lack of understanding on the part of professional scientists that we don't really understand these molecules well enough to engineer with them. We've known that hydrogen bonds weren't major for hybridization for decades, but people don't keep that in their working memory.
- [three shapes - rectangle, smiley, triangle with photos]
The ends of the helices in the triangles are hidden in the corners. The ends are strick for each other, and that's the primary force that causes them to hybridize. 75% of the forces that bring DNA together are the lateral forces between the bases.
- Base Stacking
People argue about exactly what the nature of the interaction is - dipole, van der Waals - but we can measure it. The edge of a rectangle will stick very strongly to another DNA helix. The hydrogen bonding makes the specificity. If they can't hydrogen bond, it won't be able to stack.
- a b c d
It turns out, we've learned how to master this force. If you have a bunch of helicies they can click together.
- [lots of skipped slides]
- Triangles, one broken
Here is a DNA origami angle. It has a specific shape on it. The only thing that's holding them together are these stacking interactions.
DNA origami are not something we can simulate. We got to them by trial and error and by refining our intuitive model of what DNA can do. By adding stacking interactions we refine it even more. We can hopefully even mke machines parts. Things that slide. Maybe look a little like Merkle and Freitas machines. But we can't, from first principles, get to a predictive situation. No program can predict whether a long strand and a bunch of staples will fold.
We can build things we don't understand.
The best molecular modeling guys in the world can't help me do my job better. Instead, I'm helping them calibrate their models. That's probably going to be the situation for 10 or 20 years. The simulations are not capturing the fundamental interactions. Not to say anything bad about simulation, but that's how far behind the curve we are.
- Growth of design complexity
We have our own Moore's law - 50 labs pursuing this now, size of structures doubles exponentially.
- Before I die (again)
\"I'm going to die, probably within 1.5 billion seconds.\"
Before I die I'd like to see all the stuff [from list earlier]. But I won't. We'll have molecular manufacturing. Smalley is wrong. But we won't use DNA origami. Anything you see or think of today will be replaced dozens of times in the next decades. If I draw a picture of gears, or Ralph does - we're not going to build any of that in the future. We'll have to completely reimagine as we learn more. That's one of my main messages. We can't predict any of this - even if you only use the things you understand best, you're still going to get bitten in the butt when you go in the lab.
- Before I die (3)
We have all these exponential trends, but....I saw this talk that said protein folding would be solved by faster computers. But it's not. We can only do small perturbations. De novo protein designs - only two ever, and one was a fraud. All these things will come to pass, but maybe our grandchildren will see them. We want to figure out how to accelerate. The way to do that is to maintain civilization, give to science, get your hands dirty in the lab. DON'T GIVE BUILDINGS. Endow faculty, graduate students, specify they have fellowships for 5-6 years. There are a number of structural things we can do. I don't think I'm just blind because I'm in the trenches. We just don't understand things well enough. But I want to see it happen.
- Some difficulties
Don't go watch world cup right now.
- What we can do to accelerate progress
Wish we were going 10x faster.
Bryce: Population genomics.
To what extent, if that field grows, is it competing?
David R: Speed to decode DNA underestimated due to things OUTSIDE field. Maybe new tools would help you accelerate?
Absolutely yes. Every few years someone comes out of woodwork and gives us something new. Even that is not enough. Sequencing, ... people were blind? Actually turned out to be easier problem.
Those are singular things. Typically they have to do with bits. \"Easy\"
Some of those underestimating things weren't engineers.
Can differentiate between easy and hard.
Merkle: Anything to do with bits is easy.
David H: DNA - 4 blocks. What about other non-organic building blocks?
Funny. Feynman prize winner. Does great stuff, but he didn't get tenure.
Thing is - a holy grail. Novel chemistries to compete with DNA. 20-30-40 year challenge to come up with polymer.
DNA and RNA are not interesting electronically.
People make little hetero polymers.
It would take an effort on the scale of what we've already had with DNA.
People want to do that.
Chiara: With DNA origimami - build circuit.
Encode - ACGT and compute with that. DNA computing field:
Haven't figured out way for it to scale yet.
Second distinct thing. Organize things like carbon nanotubes. Gold for optics. Organize a more conventional computer.
[12:42] more questions poll
Nice dolphins. Then dolphins together. Then university changed logo.
3D trojan horse.
Reason I did map of Americas. Wanted to do something more universal (rather than just a flag, say).
Em: Reminder - workshop this afternoon not in 583C, but outside by building 20.
=============== DURING-presentation notes
Can't stand the colors of Etherpad? Turn them off with Pad Options.
(Note: this changes the color for EVERYONE.)
Color card definitions:
Green - Agree
Red - Disagree
Yellow - Slow down (content) / I don't understand
Grey - Speed up we know this stuff / Move on to another topic
hands moving apart (sideways) - speak slower, speak more clearly
Link to backchan.nl for questions:
http://singularity.backchan.nl/meetings/view/142 First speaker
http://singularity.backchan.nl/meetings/view/143 Second speaker
http://singularity.backchan.nl/meetings/view/144 Third speaker
GSP10 Book List:
PLE CL Introduction to Policy Law and Ethics
Kathryn will review later looking for any [ref] markers--She and TFs will fill these in with references, but all are welcome (encouraged) to add references / put in as many details as available.
Try backchan.nl here: http://singularity.backchan.nl/meetings/view/124
Computers let you do a million times for the same cost of doing something once.
Good useful things have become scalable.
Problem is, bad things can scale as well.
In old days, things didn't scale that way.
History of Law - been attempt to stop evil from scaling.
We are leaving laptops on desks because only a few theives around.
Most people mostly want to live honestly.
No need for guards.
Number of bad actors are small.
With scalability - one bad actor can be multiplied by a million times.
One spammer ...
We can also overreact. Go nuts. Try to lock down society because of fear.
Other ways to scale:
Corporation - sometimes good, sometimes bad vision.
Computers allow putting bad intent out in world
the great firewall of China - only possible in part due to scalability
Everything is copyable.
I forgot to do something. I talk very quickly.
Don't feel rude if you want to hold up a yellow card.
I want you to understand this material.
Our challenge is to find some way to fight it but NOT give up our values.
Risk of giving up, due to fear, of giving up advantages.
Inside other tracks, they will give you some material.
Things that can go wrong, what people will do, how to protect freedom.
One thing at EFF - intellectural property. We will give you 3 hours on this later in program.
Changes in copyability have caused rebirth in music and movie industries.
Book industry is heading into that now.
Design is going to be copyable. 3D printers - can copy shapes. Nanotechnology - nano forges - at very fine structural level. Starship Enterprise level within your lifetimes.
Imagine a Napster for Cars. Download this Corvette ... it's cool.
Issues that tore up music industry. Still selling more product.
Avatar - brought in 2 bn [ref]
Information/design can become abundant.
Also concerned about economy of creativity.
People who make money.
People who do expensive creative works.
Find that, without locking down technology.
If you accept possibility of human minds being copyable, ethics is fascinating.
Society definitely not ready for this.
What will be economically worthwhile?
What will be protected?
Via secrecy? Harder and harder in a world where nanobots can do complete reverse engineering.
Materials - some raw materials may be particularly valuable.
Q: List - replication / credibilty? Personal services?
By location, certain places will be cool.
As Lex Luther says: beachfront property might be valuable. Dubai had idea to create more beachfront property.
There will cool places.
Perhaps AIs will do services - even if you can get any physical object you will want to be waited on hand and foot.
We have an affinity for originials - as apart from an absolutely nano-level copy.