Explore cutting-edge advancements in bioproduct research and development with the Agile BioFoundry (ABF), a consortium of seven U.S. Department of Energy national laboratories. This webinar highlights innovative tools and processes designed to accelerate R&D and scale up biomanufacturing.
Key topics include:
High-throughput automation pipelines for strain engineering and production. Advanced process development for biomanufacturing using single-carbon and gaseous feedstocks. Merging synthetic biology with machine learning and robotics for bioproduct synthesis. Featuring insights from Berkeley Lab experts Hector Garcia Martin, Chris Petzold, and Eric Sundstrom, this session offers a deep dive into scalable biomanufacturing solutions and interactive discussions with industry leaders.
Learn how to collaborate with the Agile BioFoundry and explore licensable technologies like the Automated Recommendation Tool. Watch now!
Discovery series webinar today we'll be
exploring new ways to accelerate
bioproduct R&D developed at Berkeley lab
in addition to hosting we'll also share
insights on partnership opportunities
with the agile biofoundry and with
Berkeley lab briefly I'd like to provide
some context on what the agile
biofoundry
is if you want to go to the next slide
please Sean
thanks the agile biofoundry is a
consortia of seven Department of energy
National Laboratories focused on
developing tools and processes to enable
biom manufacturing of diverse products
with an emphasis on fuels and chemicals
we partner with academic and Industry
organizations of all sizes to leverage
ferally supported research to drisk
bioprocesses towards enabling
commercialization of bioproducts
something to keep in mind during our
presentation today is a current call
that we have looking for partners for
Tool development I'll explain more at
the end of the presentations we're also
very privileged today to have three
presenters who will each share their
expert perspectives on their
Technologies first up we'll hear from
Chris petal a staff scientist whose
Focus within ABF is developing Mass Spec
solutions to problems facing biofuel and
bioproducts production then we'll hear
from Hector Garcia Martin whose work
aims to make synthetic biology a more
predictive discipline by combining
machine learning approaches with
mechanistic models microfluidics and
genetic editing techniques last to to
present will be Eric sunstrom whose work
addresses critical barriers to process
scale up intensification and integ
integration for novel bio conversions
aiming to speed industrial deployment by
reducing process costs development
timelines and scale of risks before we
proceed I'd like to remind you that your
microphones will be muted throughout
this presentation and if you have any
questions you can type them in the chat
box at the bottom of your screen and
we'll address them during the Q&A
question or session at the end of the
webinar and now I'll hand off to Chris
thanks thanks Grace let me share and uh
we can get
started uh good morning everyone uh my
name is Chris petel and I will tell you
a little bit about the agel biof
foundaries um automated pipeline for
strain analysis and I'm uh like Grace
said I'm part of the test
team so uh Grace gave you a nice kind of
overview of what the ABF is uh focused
on and the developments that we do and
and uh within the uh ABF uh we do
additionally uh demonstration projects
uh to kind of show the tools that have
been developed the capabilities that the
ABF has and and really try and U Move
the bioeconomy forward and I'm going to
highlight a a couple of them today um
and focus mainly on one of them uh you
can see that they're separated um in two
two cases by organism and by the Target
molecule that's being um produced uh in
one case it's a the host is ponus petita
and the target molecule is muconate and
in the other cases Rose perum toroes um
is the host and it's focused on
producing lipid lipid
molecules now in each of these cases uh
there are different needs and and
different uh uh areas of focus uh uh for
these uh experiments and the
experimental design in in the sudam
Mona's case um how to screen pooled
genomic and genome scale libraries um
for muconate production is a key aspect
of uh What uh what the research is uh
moving for moving towards and in order
to do that we really want to be able to
connect the genotype with the phenotype
that we observe now uh some of the
examples are we have RB tnz libraries um
crisper eye knockdown libraries as well
as um Gene ere expression libraries that
can be screened and and aide and while
there are established methods for um for
analyzing these a lot of them require
either a growth-based selection based um
method such as Al um automated lab
Evolution or um screening Technologies
where there's already an existing
biosensor that enables droplet or or fax
analyses now in the case of of muconates
while there are biosensors and other
Technologies out there there there are a
lot of cases where you're just not uh
able to um use existing Technologies so
we need a fast non-growth based um
method of phenotyping and genotyping
these straints that that are being
created likewise the in a rotos peridium
case uh being able to screen uh large
numbers of transformant to to ensure
that the uh chromosomal modifications
are in the correct place as well as um
performing the correct function are
really key and uh rotus peridium has a
unique cell wall composition and a high
GC content which makes uh Colony PCR
difficult and and so developing work CL
that enable our Technologies and our
demonstration projects is is really kind
of where um where we're going with a lot
of the tools development at the
ABF now to give you an example of of the
APF strain analysis workflow uh it it
generally starts at a culture
transformation step where a DNA is is
transformed with a host of interest and
it's then uh subsequently plated on on
agar plates for initial Colony growth at
that point
uh the colonies need to be picked and in
in all these cases that I demonstrate
here they they're all commonly done
manually um in industrial labs and
academic labs and at at all levels but
being able to do that um in iut is is
really uh really the key the culture
growth uh step there on the right is is
actually a very uh highly variable uh
portion of of the the process where
growth in 96 well plates or bio
um or other bioreactor formats can often
dictate um be dictated by the
experimental need of the host or the
target molecule and like I was saying on
the previous slide being able to do
product extraction and assay them it can
be highly variable and is often a rate
limiting
step so what I'm going to talk about
today is is the an example study of the
PSE sudon petita for muconate production
wherein the the uh strains are uh
generated they're grown and we do the
the following uh analyses we do product
extraction and product analysis to see
how much mucan a is being produced uh we
do proteomic data acquisition to uh uh
essay what the uh proteome components
are within that uh within those strains
and then we also do genotyping so that
we can utilize that uh genotypic
knowledge uh for uh Downstream learn
activities so focusing on the uh The
Colony picking production culture step
this is uh typically the the initial uh
initial stage wherein uh colonies are
picked uh via a cupix automated uh
Colony picking system into 96 well
plates uh these 96 well plates are then
handled in step two uh where they're
grown overnight um uh to uh produce the
the initial cultures that we're going to
use uh for the um production growth
assays now in step three this is an area
that is can be highly variable depending
on your your needs now in this case I'm
demonstrating that the the nun plates
are being uh triplicate uh generated in
triplicate in Deep Well plates uh for
production growth but that can also
involve uh many pass uh multiple
passages into different media conditions
or different growth conditions depending
on the host and or the need and likewise
within that step we are sampling um uh
part of those initial cultures in into
uh 384 well plates for genotyping
purposes and then we also do uh stock in
case we want to return to those cultures
at at a later
time and lastly once the production is
um proceeded for the uh the appropriate
length of time uh we do uh sampling for
both proteomic samples of of cell
cultures as and in this case uh for
muconate we are sampling the supernon of
the cultures um for the uh for the rapid
fire assay and this rapid fire assay is
is one of the key aspects of being able
to to match the throughput that we see
uh from the uh Colony picking and and uh
strain generation stage in the actual
testing and Analysis
stage this is again is a a highly
variable uh step in the process
depending on your needs for your product
in this case muconate is secreted into
the media and so being able to just
simply uh sample the the supernon uh
dilute with the appropriate solvent um
enables the the actual assay to proceed
but in some cases if you're looking for
intracellular metabolites or you're
looking for uh lipid molecules where you
need to fully lace the cells um
different uh extraction protocols need
to be uh established and um we have a
variety of them in automated formats at
the
ABF likewise the one of the main
advantages of of the um mate assay is
the throughput uh now being able to
analyze uh targeted molecules on the
order of uh five four to five samples
per minute um is is very uh much higher
than traditional methods such as liquid
chromatography or gas gas chromatography
systems this enables us to uh compete
and or maybe not compete but at least
complement uh the throughput that can be
generated um at the transformation and
and and Library uh stage where we can
get thousands of samples analyzed per
day now we we've validated these um
rapid fire assays with with the more
traditional um uh methods to ensure that
the quantification that we get um from
these methods are accurate and um
representative of the the cultures that
we that we
see now in addition to just using the
this workflow as a screening tool where
you're picking the winners and and
moving on and and only um analyzing
those we want to enable um more uh
deeper levels of learning activities
where we can really um learn more about
the system and and and these these Tech
uh techniques you'll learn more from
Hector in in a few minutes uh but to Aid
that we have developed um High through
put uh automated proteomic sample
preparation protocols uh to take the
Cell Pellets and digest them all the way
down to peptides for um lcms analysis
and because uh the the traditional
proteomic assays are often timec
consuming and and uh work intensive
we've we've really focused on being able
to do this in high throughput where we
can analyze uh 200 proteomic samples a
day and be able to quantify over a
thousand uh proteins uh from each of
those samples this this provides a rid
data set for uh strain validation uh
just analysis of uh stress response and
and and other aspects of the of the host
as it's under it undergoing its
production essays and we imp we can
Implement that Implement that throughout
the time course of the growth we can
implement it as an endpoint and and it
really helps to inform what else is
going on and and feed in uh more
advanced machine learning Technologies
um that you'll hear about later
but it also we've developed automated
methods to provide uh analysis of the
proteomic samples and and uh in a high
throughput so that it's a timely uh uh
way of getting the data to the to the
users and the
collaborators But ultimately we we
really want to be able to identify the
genotype of of these strains and see um
uh What uh genomically is U matching up
to the phenotypes and we've developed
Colony PCR methods as well as alumin
Library preparation and my sequencing
methods in hyrup in in automated fashion
to really enable the ability to get
through the entire process in a in a
timeline that actually is actionable and
uh can move forward now in some cases
this involves just simply sequencing the
barcodes of uh of a tnse library or it
can be involve more more involved
sequencing for crisper ey or gene overe
expression libraries
so just to give you an example of our
application this um for an RBT and seek
Library we arrayed and phenotyped 16
plates colonies um of approximately 1500
colonies with um control strains on each
plate and as you can see on the right an
example of what we would expect to see
in terms of the muconate production
where there's a lot of production that
didn't change very much across the
different uh uh TN Library U strains and
then there's some winners and then
there's some losers and as you can see
from this um uh this data profile we can
see that in in this case there were
quite a few of the uh strains that
actually produced a sign significantly
more uh muconate uh under the conditions
that we screened and so being able to
feed this information into uh Advanced
learning uh methods is really a way that
the ABF um uh is providing workflows
across across the needs of of uh bio
bioengineering and the
bioeconomy so just to bring you back
back to uh the uh the workflow that uh I
started with uh there are multiple user
onramps for um for these methods with
the with the ABF uh users can uh bring
their own host um and bring their own
DNA uh for transformation and and
procedure through the process they can
bring transformed cells that are already
um ready for Colony picking or they can
bring uh seed cultures that are that are
um ready and prepared for um high
throughput production growth uh and and
lastly if uh uh collaborators want to U
bring uh existing extra extracts um that
are ready for um High throughput uh
analysis with the rapid fire or other uh
other techniques the ABF is is well uh
well suited to toh to work with you in
those
cases and uh with that I'll acknowledge
the people who helped out with this and
and uh let you uh listen hear from
Hector about what we can do with all
this data thanks
okay hello everyone and uh yeah so now
that um C and A can produce Lots
thousands of data points for you what do
you do with that and that's what I'm
trying to um show you
um
see can you see my screen
yes but it's not in the full view it's
in the the presenter View mode there we
go it's probably going to be better
let's see let me try that again
sry good
now did it work looks good good yeah
okay well um so yeah now that we have
the DF can produce thousands of data
points for you or tens of thousands what
do you do with that right uh the maybe
standard analysis using Excel sheets and
biological intuition has a limit and you
can't really deal with so much data so
I'll give you an example of one tool
that we develop the automa
recommendation tool that is able to
leverage that data to guide the
metabolic engineering or the synthetic
biology process effectively and this is
just one of several tools that uh we
have developed in my group as part of
the avf and you can go to qm. lbl.gov to
see many other uh tools and and examples
that can be useful to you and well the
the the reason that we do all of this is
that well synthetic biology can really
enable a circular bioeconomy and it's
produce some interesting products
already but it's too slow right if you
and the reason is that we don't have the
predictive power that is needed as as
the them Garnet said at some point our
ability to engineer predictable outcomes
from biological assistance reminds
infantile we can make the changes we
intend particularly with the new genetic
getting tools based on chisar and so on
but the effect on biology is gener
unexpected and and we think we can fix
that with machine learning with the
predictive power that machine learning
can provide but that uh in order to feed
the amount of um the high quality data
that machine learning needs to work
properly we need Automation and I'll
give you an example of of what that
means now um art is a machine learning
tool that can provide the predictive
power needed and what it does is just
takes an input and a response and this
could be um many things are this a
general tool that can be used for a
variety of different tasks could be a
inut protein concentration and then the
amount of bofu you're producing or the
promoter combination in your DNA and the
productivity or it could be the media
components concentration for for each of
the media components and the flul in
Tighter and then it takes all these data
it learns from that and it creates a
predictive model that's tells you well
this protein profile has 10% of chance
of producing 10 Mig per liter 20% chance
of producing 50 mli per liter and so on
so it gives you a probalistic model and
from that it gives you recommendations
so uh decrease this protein one to have
uh uh double this protein two to double
uh to to get an increased production and
uh and it gives you those uh uh those
recommendations and you can see some
examples in in a couple of Publications
the initial art paper and an application
of that for promo comination decision
now art has been designed for synthetic
biology needs first because it can work
with very few instances we and you can
see that in the paper we've been able to
obtain um good results or acceptable
results with as low as 27 instances
which is something very very rare in in
AI where people are are focused on the
um high end of of instances of of
training data it does provide a c
notification so you can tell whether you
can trust the recommendation of the
prediction U depending on how much
uncertainty is associated with that if
you have like a wide Pro distribution
you can't trust it too much if you have
a very narrow one then you can trust the
recommendations quite well and then
finally it doesn't Just Produce
predictions but recommendations it
really um um uh it really goes to the
Gap of between learn and design very
efficiently and it tells you well change
this protein or try this promotor
combination and so on so it does Prov
provide recommendations that are
actionable and you can use uh to to
really mind this gab and learn on design
um I will now give you an example
because I think it's best this is best
um U scen and appreciated through an
example in which we're optimizing media
we were trying to optimize or increase
the flab in Tighter and we have 15 media
components uh which we Chang 12
systematically over two orders of
magnitude to see if that made an impact
in Flav and from that again we got uh
several instances um around you know
you'll see around 15 per pertivity cycle
and we did around five six Cycles around
60 of those and we're able to predict
response and I create a predicted model
that said well that predicted that this
media design has a certain 10% chance of
producing 10 per liter and then we did
recommendations that said that you
decrease this component one and increase
this component two and so on in order to
provide enough data for R to to work
well we uh created a cetm pipeline that
allowed us to create new Med in a high
Manner and without Arrow which is
important we had a biom a liquid Handler
that given a media composition was able
to put it uh together uh then with did
the fermentation of vctor with the where
we suppli the the the the organism
antibiotic inducer and so on WE asset
production uh through fluoresence then
we uh gathered this data into the
experiment data Deo over here and then
art pulled the data from the experiment
data Depo um made a recommendation from
New Media and then it passed it to a
Juber notebook that took those contion
recommended and then it are where of the
robot um Handler steps that it has to do
to create uh those those media um uh
recommendations what we did see in the
first time boot Rider we had initial
dbtl Zer to check your variability and
then two dbtl one and two to um with
around 15 um instances each
um um and then we train art with these
data and we saw that the predictions
versus the observations were not incredi
good incred good they had an R square of
32 but there was at least no random and
what was interesting is that in the next
dbtl 3 cycle we were able to use the
recommendations to increase production
by around 148% which was uh quite
unexpected we didn't expect media to
have such an impact in production at the
very end um we did this three times and
we obtained equalent results we did it
this is the original one where we did uh
five Cy 5 Cycles in total then we did it
again with exactly the same um um the
same um um conditions and in this case
we have to wait until dbtl 5 to get an
improvement in faulin it was around 170%
in Tighter and then we try do it with
the yield and we got an increase of
around 300 uh of in the in the third
dbtl cycle so this this seems to work
and uh quite work quite well and and and
in a repeatable fashion with of course
you know some stochasticity but what was
interesting is they provide really
surprising results in the sense that
sodium chloride was the most important
media component to explain production so
this is the sh value that means that
this is the impact in the model that
means that it's if it has a high value
if sodi CL is high it impacts the
production positively and it has low
value it impacts the the production
negatively and what was really
surprising is that sod Chlor was the
most important of that and uh in fact uh
the optimization took us to the limit of
almost to the limit of whaton P can can
really withstand so that's just an
example of the surprising results and
the optimization and that that machine
learning can bring to your project art
is being used in a variety of projects I
said it's a general tool that can be
applied to different problems we're
using it to uh give targets for Cris par
to down regulate to increase sustainable
Aviation production and these are each
of the Cycles when you can see the open
tighter increase significantly we're
using it for Designing poly synthesis
and creating what you think are the
first quantitative predictions of the
production of poly synthesis and and
we're also using um to optimize um elect
operation conditions in a in a microfil
driving lab and we're also finally uh
using to predict the scale app how to go
from Bor to Amber in a predictable
fashion but these are things that we're
working on right now and I hope to
publish very soon if you're interested
in art uh there is a website uh with um
tutorials and explanation of art and and
tutorials and even a publication that
explains how to use it and um it you're
interested in in other tools Beyond art
we you can actually check uh you can
also check q. lbl.gov and get more
information and uh that's it thank you
very much for your attention
great is that working for
everyone looks good great uh so I'm Eric
sunstrom I'm a Staff scientist at
Berkeley lab um I lead the fermentation
and Recovery team uh at abpdu the
advanced biofuels and bioproducts
process development unit uh we are a
very uh complimentary capability with
the AO biofoundry uh so our specialty is
really process development
uh and process scale up and I think kind
of a nice way to frame it is to think
about the
concept uh
of Frozen
here uh the concept of uh the the scale
up Valley of Death so many of you have
probably heard of this uh essentially
there's a lot of uh government and
University activity at low technology
Readiness levels typically uh the types
of projects that would be funded by NSF
or uh early stage Department of energy
funding uh and then there's of course a
lot of interest in the private sector in
adopting and commercializing these new
technologies uh but there's this sort of
Gap in investment where um it doesn't
it's not necessarily as interesting to
the basic researchers uh this requires
expensive capabilities uh things that
may not be off the shelf um but the
Technologies haven't yet been
sufficiently der risked for uh the
private sector to be putting money in
and so abpdu was really founded to uh
step into that Gap and help catalyze uh
that drisking step to enable
Technologies to move to the private
sector uh and I think we've been quite
successful in that regard over the years
we've been around I think about 12 years
now um abpdu has had 85 industrial
Partnerships uh over that time primarily
these are small companies looking to
access our capabilities but that can
also include larger companies that have
uh specific needs that aren't met by
their facilities or that are looking to
access other resources in the National
Lab
system we also work extensively with
academic and National Laboratory
Partnerships including uh many consortia
such as the a biofoundry uh the key
capabilities at abpdu are primarily
around bioprocess scale up process
integration process
intensification uh so we have high
throughput fermentation platforms like
the amber 250 system which is a fully
AUD autom at uh High throughput
fermentation screening platform our
capabilities go up to 300 L pilot or
pre-pilot uh which can be very useful
for assessing scalability and also for
generating larger quantities of product
and then we also have a gas fermentation
capability uh this is a newer capability
uh and that's what I'm going to be
focusing on
today the reason we've invested in this
is uh that there is a whole host of feed
stocks uh that are really uh ripe for
Bio conversion uh and there's a bit of a
gap in terms of capability that's out
there to access those feed stocks uh so
this would include things like uh
electrochemical feed stocks uh
particularly green hydrogen which is
really an upand cominging technology
that's seeing a lot of investment and a
rapid decrease in cost uh especially
coupled with renewable energy which is
similarly seeing a major decrease in
cost uh it also includes carbon capture
and CO2
utilization uh anerobic digestion uh
which is a uh pretty robust source of
renewable methane gas biomass
gasification which is a uh similarly a
way of converting biomass into these C1
feed stocks uh and then also a variety
of uh Industrial Waste gases like flare
natural gas uh steel mill off gas
there's a lot of C1 feed stock out there
uh that can be converted biologically
and there's a lot of groups interested
in this space so we have groups looking
at production of very lowcost uh things
such as uh Foods uh medium item medium
cost items such as polymers and
biochemicals and uh we even have groups
looking in the food space we've worked
with a group looking at making chocolate
for instance uh from these
electrochemical feed stocks so it really
uh covers a broad space and at abpdu
when we did some industry Outreach in
the space and talked to a lot of the
practitioners uh there was a pretty uh
clear consensus that uh that there is a
gap as far as scale up uh that is even
worse for gastr fermentation than it is
for uh typical sugar based fermentation
processes uh the reasons for this would
be that that this uh the space is fairly
nent so there's a lack of effective scal
down capability uh that can mimic what
you will see at Large Scale uh so a
serum bottle is kind of uh the gold
standard for what people are doing in
the lab bench and it really does not
replicate what they see at Large Scale
in terms of gas partial pressure total
gas supply um you know a biomass density
that you can achieve pH control all of
the different controls you can put on
the bioreactor are just not the same uh
there are barriers to operating these in
a safe fashion both at the bench and at
Large Scale um particularly flammability
comes up a lot but also gas toxicity if
you're working with carbon monoxide uh
and then these systems often work Under
Pressure which creates additional
barriers to um to getting on that
equipment and there's really no
established uh best practices there's
very few off-the-shelf pieces of
equipment so our goal is to kind of step
into that space um and provide a
resource that's accessible to people who
are just entering and want to be able to
exercise their process under industrial
industrially relevant
conditions so we've invested in the
capability you see here um the first to
come online was our 8X 500 mil uh
bioreactor block these are HL
bioreactors that are individually
pressure controlled uh and fully
stainless steel and enclosed in a fum
hood and so what that allows us to do is
operate with essentially any mix of
gases uh in a safe and secure fashion
and also to uh put pressure on the
system that can mimic pressurized
systems or even hydrostatic pressure at
depth that you'd see in a very large
bioreactor our more recent capability to
come online is the 20 L bioreactor uh we
just recently installed our uh gas
supply system for that we have a
hydrogen generator that's coming online
so that we can uh move away from gas
cylinders and finite gas supplies and
move towards uh being able to run with
large volumes of gas at scale uh we can
run with any combination here of
hydrogen oxygen CO2 Co methane ethane
nitrogen uh we'll have to do safety on
boarding for those processes but we've
got them all on boarded and tested out
and we can do back pressure control up
to five bar uh which can drive very high
gas transfer and also subject strains to
conditions that you're likely to see
during scale up
uh we also have off gas analysis set up
so we have an agilant micro GC and we
have a a thermprocess mass spec so that
we can close mass balance around these
perations so just to give a flavor of
some of the different projects where
we've applied these um first one I'll
talk about is uh industry collaboration
that we had with uh with checkerspot
this was run through biomade um along
with the UC Davis and the idea here was
actually not gas fermentation which made
it a really nice commissioning exercise
for us it was uh to look at pressurized
fermentation and how pressure affects
scale up of bioprocesses um especially
the cheer spot heterotrophic
bioprocesses for production of fats and
oils so what we did here is that we ran
a series of fermentations in the small
scale bioreactors at low pressure at
high pressure and then we ran comparable
conditions in the 300 liter bioreactor
uh again varying the back pressure and
also the feeding parameters to mimic
what you'll see at small and large scale
and we were able to um resolve some
initial scale up issues and get that
variation down below 10% between uh the
small and large scale bioreactors which
helped them resolve a uh a hurdle
towards scale up to 15,000 liter and
Commercial deployment uh and uh just in
January of this year they actually ended
up releasing uh the product that we were
working on here this is a a new infant
formula so they're tailoring fats uh to
match those that are found in human
breast milk
uh We've also been working with a number
of other companies um Cersei bioscience
industrial microbes uh Windfield bio is
our our latest partner here and uh we're
excited to continue to to work with more
companies in this
space um also talk about some of our
more basic uh fundamental research so
here we're supporting two projects from
the energy biosciences Institute uh on
the left here we're working with um
thermophilic uh ARA uh acido
thermophilic Ara and we're looking at C1
conversion in these organisms uh to see
if they can be good hosts for production
of thermotolerant enzymes for instance
uh or other kinds of thermochemical uh
processes uh and we're able to co- feed
formate and hydrogen uh and Achieve
biomass densities and and formate
conversion that otherwise would be
inaccessible uh on the right we're
working with uh cervus Neer which is
kind of the workforce host for
conversion of CO2 and hydrogen and what
we're looking at here is the CO2 uptake
pathway uh and whether we can drive CO2
conversion from Air so um could this
organism essentially serve as a carbon
concentration mechanism to utilize CO2
directly from air and uh this is an
experiment U it did it worked we were
able to see growth here um just from
just from hydrogen and air uh and it
also kind of emphasized the value of
having this system because you cannot do
this experiment in a bottle uh you need
that continuous supply of hydrogen and
air uh we need this scal down system in
order to actually determine uh whether
these bugs can hit that metric because
if you try and do this in a bottle
they'll immediately run out of carbon
dioxide another project we've been
working on here this is uh conversion of
CO2 and electricity to Fuels this is a u
bioenergy Technologies office project
where we're looking at conversion of CO2
to carbon monoxide uh and then aerobic
fermentation of carbon monoxide to
produce prod fuels and chemicals which
uh allows production of longchain fuels
and chemicals uh that are essentially uh
ready to go as sustainable aviation fuel
uh so the the typical path uh we've seen
in this space has been to ferment S gas
to Shorter chain intermediates and then
assemble those catalytically into longer
chain fuels we've shown here that it's
possible to aerobically convert carbon
monoxide directly at very high rates and
to generate very high um densities of
biomass via this throughout so uh we
were able to get 27 G per liter dry cell
weight from uh just Co and hydrogen feed
stock which is 200 times higher than
what had been reported in the literature
uh without access to this kind of uh
feedback based control on the gas supply
pressurization uh continuous gas supply
all the things that uh are enabled by uh
this kind of uh equipment investment uh
we were able to show uh pretty rapid
rates of uh Co conversion in that system
uh the last one I'll touch on is uh
electrochemical integration so we've
been working a fair bit as well on uh
different configurations where we could
directly integrate uh generation of
electrochemical gaseous feed stocks with
conversion uh that simplifies the
process and also allows for direct
dispersion of the gases into the system
uh so this is an example where we
actually did water splitting
electrolysis uh in a fermentation vessel
and use that to generate a um expression
of a heterologous product actually U the
same molecule that Hector was just
talking about flaviolin uh so we can
produce it in this tank from uh just
electrons and CO2 in a single vessel uh
and grow a uh genetically engineered
strain of uh kuber Vitus necor to
express production of that product uh
which would allow you to produce these
things from electricity essentially
anywhere you have access to uh a CO2
source so this is just kind of a flavor
of the things we can do we uh the the
whole purpose of these capabilities is
to support collaboration so uh if any of
this kind of jogs some potential
application in your mind I'm very happy
to uh discuss
further all right thank you Eric and
thank you everyone for your
attention just going to go ahead and
share my screen
okay all right um so before we move on
to the Q&A session and hear a little bit
more from our presenters um go ahead and
keep adding those questions to the chat
but I'll take a few moments here to
discuss how businesses and other
organizations can partner with Berkeley
lab and with the agile biof Foundry I
hope I'll be able to provide you an
opportunity to understand how our
strategic Partnerships and Tech
technology transfer work within the
framework of a national lab and how they
can be leveraged for commercial success
so my role at Berkeley lab and with the
AEL biofoundry is really to support our
scientists in getting their Technologies
into the marketplace and to Foster these
Partnerships and collaborative
research so at Berkeley lab there's
several ways to kind of Leverage these
capabilities and work with us um firstly
our intellectual property office is
responsible for patent and software
licenses uh this office also handles
agreements such as indas and mtas that
are typically needed prior to you taking
on the license of our IP uh we provide
all matters relating to IP including IP
management with other institutions and
Labs um and IP also works closely with
our strategic partnership office on IP
terms uh and they handle the various
agreements such as user agreements
strategic partnership projects or
cooperative research and development
agreements so what are those this is a
doe chart that spells out the specific
terms regarding IP and data rights you
get under these different types of
agreements so first up here is ACR a
Cooperative research and development
agreement ACR anticipates both Partners
work together under an ownership Falls
and inventorship type model so typically
these relationships will last several
years and the scope will have tasks for
both parties uh you can work onsite
alongside our scientist Center of and
and the IP terms here the lab will take
title to our subject inventions but the
partner has an option to negotiate up to
an exclusive license in a pre-negotiated
field of views data can also be create a
protected information held for 5 years
and um our background IP is provided to
the participant at no cost for the
duration of the
project second type of agreement here is
a strategic partnership project or SP
used to be known as work for others if
you recognize that and this is a classic
sponsored research project where an
external partner funds our scientists to
do a specific scope of work on their
behalf at essentially the same cost it
costs us to do the work so it's full
cost recovery if the sponsor is a US
company using non-federal funds the
subject inventions can belong to the
partner this chart also shows user
agreements at the bottom here and we do
use these at Berkeley lab but they're
not very agile biofoundry relevant so
I'm happy to answer questions about user
agreement opportunities and where else
at Berkeley lab you might see these pop
up um if there's anything that comes up
feel free to drop that in the
chat okay so specific to Agile
biofoundry um both at Berkeley lab and
utilizing those other seven Labs
capabilities and we have some
partnership information of ailable at
the site on this QR code and I also saw
that Sean dropped the link in the chat
so you can follow that and at this link
you'll be able to view a create a
template so that's the Cooperative
research and development agreement and
this template is the form that we use
for all of our projects that utilize doe
funds so this this form is
non-negotiable if you do have that Doe
funding and the point of this is to aim
is to aim to speed Contracting so we can
actually get to work faster um I think
we all know Contracting can be a little
bit of a barrier and when you're you're
working with multiple National Labs this
is a great way to get that done quickly
so you can go ahead on that link and
review that form if that's interesting
to you and then I'd like to highlight
one opportunity that I mentioned at the
start that we have available right now
um that's relevant I think to what you
heard today so this is our strategic
core industry Partnerships or skips as
we call them in these projects our
industry Partners will provide um up to
20 or at least 20% cost share in
addition to Industry perspective and
expertise in collaboration uh so those
tools you heard about today are
available for these Partnerships as are
several others and you can read about
all the available opportunities on that
partnership site as well as a a skip for
these core for biochemicals and biofuels
that I mentioned at the start of today's
presentation you'll also find on this
work with us site a list of our
licensable ABF Technologies both from
Berkeley lab and from our partner Labs
including those that you heard about
today and those links will take you to
more information um and you can also
just email me and I can help you through
identifying the IP you're interested in
h all the way through executing a
license agreement or direct you through
our partnership
opportunities so I'll in here I've
already seen some great questions in the
chat and we can start addressing them
shortly um and please continue to submit
those queries as we proceed and we'll
aim to cover as many as possible um if
you have more questions something comes
up later here's our contact information
for all the presenters and then at the
bottom of the screen again you you'll
see the link to the ABF website as well
as our link to IPO at Berkeley lab where
you'll see all of Berkeley lab's
licensable
Technologies
okay so now Oops I did not mean to stop
screen sharing also that back up but
it's Q&A time um so I'm going to read
through a couple
questions let's see what has been
answered already in the chat all right I
think it'll be for Eric up first it
looks like this question to Chris got
answered so Eric how do you clean the
bio reactor after one specific
experiment uh yeah good question so our
bioreactors uh essentially anything at
the 10 lit scale and Below we is
autoclavable um so we can fully
autoclave everything anything at the uh
20 L scale and above is uh steam in
place so all of those have integrated
steam in place systems that can fully
sterilize so we haven't ever actually
seen any kind of issues with uh
contamination from one run to the next
for
instance thanks
Eric okay also for Eric is it possible
to have Mycelia production in such
bioreactors yes uh so we have grown in
our sort of heterotropic bioreactor
systems we have grown filamentous fungi
uh and we haven't seen any specific
issues there but you'd have to you know
make sure that everything's tested out
so that there's no uh issues with Shear
Force especially uh we have not done it
in the gas fermentation bio reactors I
don't see any reason why it wouldn't
work it's just a question of what people
bring to us so happy to discuss
further okay one more for you while
you're on um oh maybe this is a followup
even if it contains solids is it
autoclavable I assume that's to you Eric
followup to that question you just
answered uh would I guess would depend
on the the application um
so as long as it can be uh there's no
problem autoclaving solids I guess the
question is whether the solids how the
solids would respond to autoclaving so
we do have some feed stocks like
hydrates that are challenging in that
way where where um you could get
caramelization or something if you were
to actually autoclave it those are
trickier to deal with you can uh
sterilize them other ways or uh if not
sterilize them you can pasteurize them
for instance uh so you have to get
creative
sometimes okay one more on your slide
about water electrolysis in the
fermentation broth Do You observe
poisoning of your electrodes due to the
presence of several biomolecules in the
fermentation broth yeah yeah that's a
great question so um I wouldn't
necessarily say poisoning but we do have
to um when we do these integration
experiments we do have to change our
broth to eliminate any ions that are
going to create issues um so that would
include things like chloride that may
result in chlorine formation or may
create side reactions that create
toxicity we did actually see in that
particular experiment that the uh that
the electrolyzer was actually active in
chemically altering the product uh so
that was something that was
unexpected um but potentially down the
line you could see it and I do think you
do have to play close attention if
you're going to integrate those unit
operations to uh very long uh time
scales to make sure that your media your
product everything is
compatible a follow up on that one are
there proteins in your broth that lead
to
poisoning um no I
don't know if proteins specifically I
mean um but over time I think it's
potentially that there could be um if
there's cell Isis for instance we can
get proteins in the broth and and you
could get more complexity in the broth
we didn't see anything like that in the
time skills that we're operating uh I do
think if you ran them for very long time
skills you'd start to see these kinds of
failure
modes okay any employment of single-use
technology bioreactors or other Process
Equipment wave style
bioreactors yeah so the the Amber 250
system is single use um and it has been
used for things such as cell culture uh
where where single use is typically
deployed we haven't we don't have any
single use capability above that scale
um typically those are more common in um
biopharmaceutical applications uh it is
something we've thought about we haven't
really moved into cell culture that's
also something we've thought about I
think wave style Maybe more popular in
those applications where it's where it's
low Shear so in in general there's a
pretty robust uh contract research
environment for biopharmaceuticals and
we haven't seen it as really a core part
of our doe Mission but it's definitely
something that we're open to it's
something we've thought
about okay thanks
Eric so question I'm not sure which of
our presenters might be able to answer
this um but a question about
biodegradable batteries so so this
company is a fairly new and looking to
raise funds and get their battery safety
tested is this something that the ABF
can assist with I'm not sure if any of
our presenters want to take that or
Nathan even if you want to jump in and
give some
comments I'm happy to chime in Grace
specifically what did you want me to to
speak to sorry that's fine um so the
question question is from a company
making biodegradable batteries um and
they're looking to get their batteries
safety tested I'm not sure if that's
something that the ABF might be able to
assist with or not I don't think that's
going to be something that the ad
biofoundry is directly going to be set
up um to give you some support on but I
think in the the chat Sean did put in a
link to lab partnering um services so
that might be a good good place to start
um and the other just the the general
websites for the National Labs including
Berkeley lab might be another set of
resources um but the Agro biofoundry
specifically probably isn't set up for
doing battery safety testing yeah so you
might check out this ipo. lbl.gov link
as well and and give a look see if
there's anything there that might be
relevant okay a couple more have come in
um I think this will be for you Eric
again is it economically feasible to use
a single-use
bioreactor uh yeah so I guess the answer
to that is it depends um for for uh
process development certainly I think
the uh the labor savings at small scale
uh are higher than the cost of the
bioreactors in a production setting I do
think we've seen the pharmaceutical or
the the biopharmaceutical industry move
towards single use and plastic
bioreactors because they have very very
stringent sterility requirements uh so
if your product is very high value and
produced it relative ly smaller
bioreactor
scales uh I think it's possible but for
any kind of fuels or chemicals or
industrial type products um likely not
unless somebody comes out with some kind
of radically new design that would lower
the
cost okay and then also for gas
fermentation can you describe your
method for determining mass balance in a
little more detail yeah that's a great
question so for gas fermentation we have
separate mass flow controllers for the
flammables for the oxygen for the carbon
source and then for an inert gas uh so
we know the rates that we're putting
these gases in we also have a flow meter
on the back end so we know how much gas
is coming out uh we can also kind of
calibrate that Upstream of the run to
make sure that our flow meters are are
matching what we think they're putting
out uh then on the very back end we also
have for our smaller scale system an
agilant micro GC which has a manifold
that can cycle between the eight
bioreactors so you can continuously
measure uh every couple minutes what's
coming out the back end we can combine
that with the flow meter and we can also
combine that with the innert gas feed so
sometimes We'll add a little bit of
argon or nitrogen as a tracer uh and we
can use the ratio of the other gases in
the off gas to that iner gas uh to make
sure that we're aligned uh on the the
gas feed r so it is it's more
complicated than uh you would think to
get good mass balance but we have a good
handle on it
now okay so one might actually be for
Sean here when is the bioenergy liftoff
report coming out I'm not sure if
anybody knows thanks gra yeah I was
actually looking at that that question
right now in the chat and um to be
determined but I will make sure that uh
I provide a link to the lift off report
on this uh site and if you'd like to
find out more about uh the lift off
reports and when they're released you
can um subscribe I'll put the link into
uh the chat right now so that you stay
up to date on all the latest
reports okay and Hector maybe you'll
want to take this one how can generative
AI like Chad GPT improve the outcomes of
ABF
efforts uh yeah good question so there's
a variety of player where things like
generative AI could be of interest one
obvious one is literator reviews see
what's been there in the past has been
used you know like example in media
components for example or or a gene
knockout or things that have worked in
the past another place where it could
work out is by using um by creating DNA
uh large language models that's
something that uh we have been
collaborating with people from naron and
other places to create large CL models
that really utilize all the 7 540
terabytes that uh of of DNA sequencing
that GGI provides per year to uh make
sure that we can use that data to to
with phenotypic data to to uh create
foundational models that help us predict
things better um so those are the two
things that I can think of right now um
I think that you know of all the AI
things out there generative models are
probably least useful for us because
there isn't a lot of data we don't have
the density of data to make their predic
capabilities really useful right uh for
that what we need to do is collect more
of that data in a more systematic way
and then make it available to many of
these models but but again you know we
don't have my feeling when the time that
we would use this kind of approaches
there is not enough density of uh of uh
scientific data to create really uh
really useful uh uh recommendations but
we we keep working on that and we can
create bigger databases that can that
can be leveraged by llms you know such
as jbd or CL or any of these to really
try to uh to do things that that to
suggest things that the or do
recommendations or do predictions that
can help the the the
restructure all right thanks Hector it
looks like that is going to wrap up our
Q&A session so if we didn't get to your
question although I think we hit all of
them um we can follow up with you by
email within the next uh week or so and
again go ahead and um jot down these
email addresses or or go to our site if
you you have any other questions um and
thank you again for your participation
in today's webinar your engagement
interest are greatly appreciated and we
look forward to welcoming you to Future
sessions