Intelligent Media
Formulation Using
Machine Learning
Intelligent Media Formulation
Using Machine Learning
2/9/2021 // confidential
2
Webinar Series
Sept 2020 webinar: Better Media,
Better Outcomes (available to view on
Nucleus Biologics’ website)
Nov 2020 webinar: Faster Media,
Faster Outcomes (available to view on
Nucleus Biologics’ website)
Today: Intelligent Media Formulation
Using Machine Learning
2
Intelligent Media Formulation
Using Machine Learning
2/9/2021
Poll
2/9/2021
4
Who We Are
David Sheehan
Founder and CEO
Nucleus Biologics
Roddy O’Connor, Ph.D
Research Assistant
Professor
University of Pennsylvania
Alex Klarer
Head of CellTherapy
Development
BioCentriq
David Smith Ph.D
VPTechnical
Operations
Ori Biotech
Intelligent Media Formulation
Using Machine Learning
2/9/2021
5
2/9/2021
What Problem Are We Solving?
- Every day more data is published on how media components are inexorably
tied to the performance of cell therapy products
- We are in a rapidly growing industry and time to market is key for any therapy
developer
- Proprietary OffThe Shelf (POTS) media are the convenient option for scientists
but these formulations are decades old and are not disclosed to the scientist
making it impossible to optimize
- There are currently 425K papers onT Cells and 67K on MSCs
- No tool has ever existed that allows users to instantly create their own media
based on proven science and get samples quickly to test
Position Statement
NB-AIR™ is the only cloud-basedAI platform for media configuration
based on peer-reviewed articles and community research
Intelligent Media Formulation
Using Machine Learning
To recap…
2/9/2021
Intelligent Media Formulation
Using Machine Learning
7
Better Media, Better Outcomes
September 2020 Webinar Recap
• We are in a rapidly growing industry and
time to market is key for any therapy
provider
• Media and reagents are a source of
variability that will affect outcomes
• Individual components in your cell culture
medium can have a profound affect on your
therapeutic efficacy
• NB-Lux solves the long lead time dilemma of
custom media
• Creating your own formula gives you: ability
to optimize at ingredient level, have a
qualified second source, and create company
intellectual property that makes your
therapy more unique
• If you are not using optimized media, you
are losing a competitive advantage in your
therapy development
Intelligent Media Formulation
Using Machine Learning
8
• Underlying paradox ofT cell-based
immunotherapies
• antigen-specificT cells effectively infiltrate
tumors but reside there in a dysfunctional
state
• Mechanisms ofT cell dysfunction are complex
and multidimensional
• Antigenicity
• Immunogenicity
• Metabolic features of the tumor
environment
• Historical proprietary media formulations
contained high glucose levels
• Opportunity is here to optimize the medium
formulation to develop betterT cells
• A platform that offers insight into the qualities
we want to confer toT cells has therapeutic
potential
Central Challenge: How Do We Optimize T Cell
Quality Prior To Adoptive Transfer?
Aldinucci et al., Int. J. Mol Sci (2019)
Intelligent Media Formulation
Using Machine Learning
9
Glucose Exposure
New findings from Greg Delgoffe’s group show that exposure to high glucose
affects mitochondrial function
0.1 mM (tumor interstitia)
MitoΔΨ
5 mM (‘typical’serum)
11 mM (diabetic serum/RPMI)
25 mM (DMEM)
33 mM (X-VIVO)
40 mM (AIM-V)
40 mM (AIM-V) in G-rex
[glucose]
• Existing proprietary and
classical medias were designed
to keep cells alive and
proliferating, not model any in
vivo environment
• Hyperglycemic conditions
promote the same types of
mitochondrial stress observed in
TIL-equivalent glucose
• Can glucose be lowered in
vitro to prevent
mitochondrial stress and
enhance cell longevity?
Intelligent Media Formulation
Using Machine Learning
10
2/9/2021
Better Media, Better Outcomes
Media is a Critical Part ofYourTherapy Ecosystem
The dilemma of
media choice in
cell and gene
therapy
POTS
Unknown
ingredients, not
your IP, sole
sourced
Fast regulatory,
supposedly in-
stock
Custom media
6-month+ lead
time, difficult to
iterate
Own the media,
know every
ingredient
Intelligent Media Formulation
Using Machine Learning
Up until now, no perfect solution
11
Faster Media, Faster Outcomes
November 2020Webinar Recap
• NB-Lux™ development time is 22 weeks
the same as Proprietary OffThe Shelf
(POTS) media
• POTS is 33% more expensive than NB-
Lux custom media. At scale that equates
to $4 mil of cost for a therapy company
11
Intelligent Media Formulation
Using Machine Learning
• Creating custom media allows you
to study effects of single
components on your ecosystem and
ultimately allow you to create
defined formulations for your cells
So, how do you create
formulations?
2/9/2021
Intelligent Media Formulation
Using Machine Learning
It is time to give control back to
the scientists
2/9/2021
Intelligent Media Formulation
Using Machine Learning
14
Introducing Nucleus Biologics’ Artificial Intelligence
Research (NB-AIR)
AI Based Media Configuration
What it does
 Reduces formulation time by recommending media based on users Cell and Critical
Quality Attributes (CQAs)
 Identifies high value compounds and their affect on cell performance from published
papers
 Recommends multiple formulations to test based on concentrations from extracted data
analysis
 Enhances Cell Performance by allowing rapid ordering and testing of formulations
configured to identify key performance contributors In vitro
How it does that
1. Machine learning algorithm that searches PubMed articles for cell types, CQA and
compounds
2. Identifies conclusion of research paper and scores article and contribution to CQA
3. Uses machine learning to create recommended formulas that can be customized by
user
Intelligent Media Formulation
Using Machine Learning
2/9/2021
15
2/9/2021 // confidential
Walkthrough- Pick your cell
16
2/9/2021 // confidential
Choose Your CQAs
17
2/9/2021 // confidential
Pick Your Base + Components
18
2/9/2021 // confidential
Finalize your Formulations to Test
19
2/9/2021 // confidential
Export to NB-Lux™
20
2/9/2021
Pubmed Data Analyzed Using Machine
And Deep Learning Modules
Intelligent Media Formulation
Using Machine Learning
21
Components Score
2/9/2021
• To provide a relative value of the potential of each component
to impact your culture needs, we look at how many times we
find evidence of usage in your cell type of interest.
• We then weight the contribution of each paper by analyzing:
the impact factor of the journal, the number of citations the
paper has, and the number of years since it was published.
• This way, components that have been published several times
in papers that are highly cited have a higher score than
components mentioned once in a publication with no citations.
• While this approach doesn’t give us information on the
magnitude of effects, it gives us a good indication on the
possibility of finding components with an effect.
Intelligent Media Formulation
Using Machine Learning
22
2/9/2021
The NB Ecosystem
NB-AIR™
AI enabled platform that creates cell specific formulas derived from meta-analysis
of peer reviewed literature based on your specifiedCritical Quality Attributes
NB-Lux™
Cloud based portal that gives real time quotes and lead times for your media
formulations
krakatoa™ (coming soon)
A new paradigm that changes the environmental footprint of cell culture media
Intelligent Media Formulation
Using Machine Learning
23
NB-AIR
Q&A Session
23
Intelligent Media Formulation
Using Machine Learning
Missed a webinar? Check it
out on
www.nucleusbiologics.com
/education
2/9/2021 // confidential
2/9/2021 24
Thank you!
Connect with us
LinkedIn: www.linkedin.com/company/nucleus-biologics
Twitter: @nucleusbiologic
Intelligent Media Formulation
Using Machine Learning

Intelligent Media Formulation Using Machine Learning

  • 1.
    Intelligent Media Formulation Using MachineLearning Intelligent Media Formulation Using Machine Learning 2/9/2021 // confidential
  • 2.
    2 Webinar Series Sept 2020webinar: Better Media, Better Outcomes (available to view on Nucleus Biologics’ website) Nov 2020 webinar: Faster Media, Faster Outcomes (available to view on Nucleus Biologics’ website) Today: Intelligent Media Formulation Using Machine Learning 2 Intelligent Media Formulation Using Machine Learning 2/9/2021
  • 3.
  • 4.
    4 Who We Are DavidSheehan Founder and CEO Nucleus Biologics Roddy O’Connor, Ph.D Research Assistant Professor University of Pennsylvania Alex Klarer Head of CellTherapy Development BioCentriq David Smith Ph.D VPTechnical Operations Ori Biotech Intelligent Media Formulation Using Machine Learning 2/9/2021
  • 5.
    5 2/9/2021 What Problem AreWe Solving? - Every day more data is published on how media components are inexorably tied to the performance of cell therapy products - We are in a rapidly growing industry and time to market is key for any therapy developer - Proprietary OffThe Shelf (POTS) media are the convenient option for scientists but these formulations are decades old and are not disclosed to the scientist making it impossible to optimize - There are currently 425K papers onT Cells and 67K on MSCs - No tool has ever existed that allows users to instantly create their own media based on proven science and get samples quickly to test Position Statement NB-AIR™ is the only cloud-basedAI platform for media configuration based on peer-reviewed articles and community research Intelligent Media Formulation Using Machine Learning
  • 6.
    To recap… 2/9/2021 Intelligent MediaFormulation Using Machine Learning
  • 7.
    7 Better Media, BetterOutcomes September 2020 Webinar Recap • We are in a rapidly growing industry and time to market is key for any therapy provider • Media and reagents are a source of variability that will affect outcomes • Individual components in your cell culture medium can have a profound affect on your therapeutic efficacy • NB-Lux solves the long lead time dilemma of custom media • Creating your own formula gives you: ability to optimize at ingredient level, have a qualified second source, and create company intellectual property that makes your therapy more unique • If you are not using optimized media, you are losing a competitive advantage in your therapy development Intelligent Media Formulation Using Machine Learning
  • 8.
    8 • Underlying paradoxofT cell-based immunotherapies • antigen-specificT cells effectively infiltrate tumors but reside there in a dysfunctional state • Mechanisms ofT cell dysfunction are complex and multidimensional • Antigenicity • Immunogenicity • Metabolic features of the tumor environment • Historical proprietary media formulations contained high glucose levels • Opportunity is here to optimize the medium formulation to develop betterT cells • A platform that offers insight into the qualities we want to confer toT cells has therapeutic potential Central Challenge: How Do We Optimize T Cell Quality Prior To Adoptive Transfer? Aldinucci et al., Int. J. Mol Sci (2019) Intelligent Media Formulation Using Machine Learning
  • 9.
    9 Glucose Exposure New findingsfrom Greg Delgoffe’s group show that exposure to high glucose affects mitochondrial function 0.1 mM (tumor interstitia) MitoΔΨ 5 mM (‘typical’serum) 11 mM (diabetic serum/RPMI) 25 mM (DMEM) 33 mM (X-VIVO) 40 mM (AIM-V) 40 mM (AIM-V) in G-rex [glucose] • Existing proprietary and classical medias were designed to keep cells alive and proliferating, not model any in vivo environment • Hyperglycemic conditions promote the same types of mitochondrial stress observed in TIL-equivalent glucose • Can glucose be lowered in vitro to prevent mitochondrial stress and enhance cell longevity? Intelligent Media Formulation Using Machine Learning
  • 10.
    10 2/9/2021 Better Media, BetterOutcomes Media is a Critical Part ofYourTherapy Ecosystem The dilemma of media choice in cell and gene therapy POTS Unknown ingredients, not your IP, sole sourced Fast regulatory, supposedly in- stock Custom media 6-month+ lead time, difficult to iterate Own the media, know every ingredient Intelligent Media Formulation Using Machine Learning Up until now, no perfect solution
  • 11.
    11 Faster Media, FasterOutcomes November 2020Webinar Recap • NB-Lux™ development time is 22 weeks the same as Proprietary OffThe Shelf (POTS) media • POTS is 33% more expensive than NB- Lux custom media. At scale that equates to $4 mil of cost for a therapy company 11 Intelligent Media Formulation Using Machine Learning • Creating custom media allows you to study effects of single components on your ecosystem and ultimately allow you to create defined formulations for your cells
  • 12.
    So, how doyou create formulations? 2/9/2021 Intelligent Media Formulation Using Machine Learning
  • 13.
    It is timeto give control back to the scientists 2/9/2021 Intelligent Media Formulation Using Machine Learning
  • 14.
    14 Introducing Nucleus Biologics’Artificial Intelligence Research (NB-AIR) AI Based Media Configuration What it does  Reduces formulation time by recommending media based on users Cell and Critical Quality Attributes (CQAs)  Identifies high value compounds and their affect on cell performance from published papers  Recommends multiple formulations to test based on concentrations from extracted data analysis  Enhances Cell Performance by allowing rapid ordering and testing of formulations configured to identify key performance contributors In vitro How it does that 1. Machine learning algorithm that searches PubMed articles for cell types, CQA and compounds 2. Identifies conclusion of research paper and scores article and contribution to CQA 3. Uses machine learning to create recommended formulas that can be customized by user Intelligent Media Formulation Using Machine Learning 2/9/2021
  • 15.
  • 16.
  • 17.
    17 2/9/2021 // confidential PickYour Base + Components
  • 18.
    18 2/9/2021 // confidential Finalizeyour Formulations to Test
  • 19.
  • 20.
    20 2/9/2021 Pubmed Data AnalyzedUsing Machine And Deep Learning Modules Intelligent Media Formulation Using Machine Learning
  • 21.
    21 Components Score 2/9/2021 • Toprovide a relative value of the potential of each component to impact your culture needs, we look at how many times we find evidence of usage in your cell type of interest. • We then weight the contribution of each paper by analyzing: the impact factor of the journal, the number of citations the paper has, and the number of years since it was published. • This way, components that have been published several times in papers that are highly cited have a higher score than components mentioned once in a publication with no citations. • While this approach doesn’t give us information on the magnitude of effects, it gives us a good indication on the possibility of finding components with an effect. Intelligent Media Formulation Using Machine Learning
  • 22.
    22 2/9/2021 The NB Ecosystem NB-AIR™ AIenabled platform that creates cell specific formulas derived from meta-analysis of peer reviewed literature based on your specifiedCritical Quality Attributes NB-Lux™ Cloud based portal that gives real time quotes and lead times for your media formulations krakatoa™ (coming soon) A new paradigm that changes the environmental footprint of cell culture media Intelligent Media Formulation Using Machine Learning
  • 23.
    23 NB-AIR Q&A Session 23 Intelligent MediaFormulation Using Machine Learning Missed a webinar? Check it out on www.nucleusbiologics.com /education 2/9/2021 // confidential
  • 24.
    2/9/2021 24 Thank you! Connectwith us LinkedIn: www.linkedin.com/company/nucleus-biologics Twitter: @nucleusbiologic Intelligent Media Formulation Using Machine Learning

Editor's Notes