Slide 1: Manufacturing & Digital Day Agenda
Manufacturing & Digital Day Agenda Introduction Stéphane Bancel, CEO 1:00–1:20 PM 20 min Norwood Overview Juan Andres, Chief Technical Operations & Quality Officer 1:20–1:50 PM 30 min Q&A Session 1:50–2:05 PM 15 min Competing in the Age of AI Karim R. Lakhani, Charles E. Wilson Professor of Business
Slide 2: Manufacturing & Digital Day
Manufacturing & Digital Day Stéphane Bancel March 04, 2020 © 2020 M oderna Therapeutics
Slide 3: Forward-looking statements
Forward-looking statements This presentation contains forward-looking statements within the meaning of the Private Securities Litigation Reform Act of 1995 , as amended including, but not limited to, statements concerning: the ability of digital integration to reduce labor requirements and cycle tim
Slide 4: mRNA is the software of life
mRNA is the software of life What if it could be turned into a drug? Central Dogma of Molecular Biology DNA messenger RNA Protein Hard Drive Software Hardware DNA stores our mRNA reads off the Proteins accomplish the genetic information DNA and instructs cells work in the body – structure, to make p
Slide 5: mRNA as a potential new class of medicines
mRNA as a potential new class of medicines 1. Large product opportunity 2. Higher probability of technical success 3. Accelerated research and development timelines 4. Greater capital efficiency over time vs. recombinant technology Slide 5 © 2020 M oderna Therapeutics
Slide 6: If mRNA could be a drug…it would be a platform
If mRNA could be a drug…it would be a platform If you develop a solution for one program, you have a template for many programs Formulation mRNA 5’ Region 3’ Region 1. Large product opportunity Coding Region (ORF) 2. Higher probability of technical success 3. Accelerated If mRNA works once, research
Slide 7: We have focused on managing risk since inception
We have focused on managing risk since inception With a focus on… 1 Technology risk 2 Biology risk 3 Execution risk 4 Financing risk Slide 7 © 2020 M oderna Therapeutics
Slide 8: To manage technology risk we took six modalities into the clinic in parallel
To manage technology risk we took six modalities into the clinic in parallel Personalized OX40L/ cancer IL-23/IL-36𝛾 CMV vaccine (triplet) vaccine VEGF-A (no LNP) KRAS Biology risk cancer OX40L Zika vaccine vaccine Fabry MMA PA H10/H7 Chikungunya influenza antibody vaccine Technology risk Modality…
Slide 9: Given the breadth of the portfolio we needed to build our own facility
Given the breadth of the portfolio we needed to build our own, fully integrated/digital facility to reduce execution risk mRNA 5’ Region 3’ Region Coding Region (ORF) Single platform economies of scale If mRNA works once, it should work over and over again for different coding regions → different pr
Slide 10: Progression towards a new class of medicines
Progression towards a new class of medicines Impact Stage of development Slide 10 © 2020 M oderna Therapeutics
Slide 11: Progression towards a new class of medicines
Progression towards a new class of medicines Impact Goal: Enter the clinic safely Develop mRNA science, delivery technology and Norwood facility planning manufacturing started early 2016; facility opened in July 2018 Scientific First-in-human foundation (Dec. 2015) Stage of development Slide 11 © 20
Slide 12: Progression towards a new class of medicines
Progression towards a new class of medicines Enabled by Norwood & Digital integration Impact Goal: Validate modalities through clinical data Develop mRNA Explore mRNA science, delivery technology across six technology and different modalities manufacturing Scientific First-in-human Clinical First Co
Slide 13: Progression towards a new class of medicines
Progression towards a new class of medicines Enabled by Norwood & Digital integration Impact Goal: File multiple BLA(s) and continue exploration Develop mRNA Explore mRNA Build Core modalities science, delivery technology across six technology and different modalities Investigate current manufacturi
Slide 14: Expanding core modalities with additional development candidates
Expanding core modalities with additional development candidates Core Exploratory Personalized OX40L/ cancer IL-23/IL-36𝛾 vaccine (triplet) Biology risk VEGF-A (no LNP) KRAS cancer OX40L vaccine MMA PA CMV Chikungunya vaccine antibody Technology risk Systemic Localized Systemic Prophylactic…
Slide 15: Progression towards a new class of medicines
Progression towards a new class of medicines Enabled by Norwood & Digital integration Impact Goal: Deliver a large commercial portfolio of innovative medicines Develop mRNA Explore mRNA Build Core modalities Scale for commercial science, delivery technology across six technology and different modali
Slide 16: Manufacturing & Digital Day Agenda
Manufacturing & Digital Day Agenda Introduction Stéphane Bancel, CEO 1:00–1:20 PM 20 min Norwood Overview Juan Andres, Chief Technical Operations & Quality Officer 1:20–1:50 PM 30 min Q&A Session 1:50–2:05 PM 15 min Competing in the Age of AI Karim R. Lakhani, Charles E. Wilson Professor of Business
Slide 17: Norwood
Norwood Juan Andres March 4, 2020 © 2020 M oderna Therapeutics
Slide 18: Juan Andres, Chief Technical Operations and Quality Officer
Juan Andres, Chief Technical Operations and Quality Officer As Chief Technical Operations and Quality Officer, Juan Andres is responsible for the supply of product required for Moderna’s preclinical and clinical development programs, as well as scaling the CMC (chemistry, manufacturing and controls)
Slide 19: We are a platform…
NEW We are a platform… DNA mRNA mRNA + LNP Drug product Slide 19 © 2020 M oderna Therapeutics
Slide 20: NEW = Learning; PLATFORM = Many
NEW = Learning PLATFORM = Many Slide 20 © 2020 M oderna Therapeutics
Slide 21: We initially had more questions than answers
We initially had more questions than answers… purity, stability? reproduceable? potency? Quality scale? attributes? “platform” GMP manufacturing? scale? Slide 21 © 2020 M oderna Therapeutics
Slide 22: Platform research + technical development = create
Platform research + technical development = “create” + + Slide 22 © 2020 M oderna Therapeutics
Slide 23: We have solved many questions
CREATE We have solved many questions mRNA structure mRNA + Lipid encapsulation Purity Size of LNP How to measure? Stability LNP Stability Slide 23 © 2020 M oderna Therapeutics
Slide 24: Quality First
CREATE Quality First Slide 24 © 2020 Moderna Therapeutics
Slide 25: Once right… Speed Scale Cost
Once right… Speed Scale Cost ...were engineering problems! Slide 25 © 2020 M oderna Therapeutics
Slide 26: Before Norwood
Before Norwood Plasm id LNP Fill/Finish m RNA QC Slide 26 © 2020 M oderna Therapeutics
Slide 27: Plasmid mRNA LNP Fill/Finish QC
Plasmid mRNA LNP Fill/Finish QC Slide 27 © 2020 M oderna Therapeutics
Slide 28: Norwood is a clinical development site
Norwood is a clinical development site PRECLINICAL PERSONALIZE VACCINE UNIT CLINICAL Slide 28 © 2020 M oderna Therapeutics
Slide 29: Engines
3 Engines PRECLINICAL PERSONALIZED VACCINE UNIT CLINICAL Slide 29 © 2020 M oderna Therapeutics
Slide 30: Preclinical, Personalized Vaccine, Clinical batches and scale
PRECLINICAL PERSONALIZED VACCINE UNIT CLINICAL Batches 23,500 1 batch/1 patient; 90 to date 1 batch/many patients Scale Slide 30 © 2020 M oderna Therapeutics
Slide 31: Preclinical, Personalized Vaccine, Clinical batches and scale
PRECLINICAL PERSONALIZED VACCINE UNIT CLINICAL Batches 23,500 1 batch/1 patient; 90 to date 1 batch/many patients Scale 10mg >100mg 5g-75g+ Slide 31 © 2020 M oderna Therapeutics
Slide 32: What we know now…. Platform technology
What we know now…. Platform technology Similar Fast process Cell-free Less capital processes improvements investment Quality Speed Scale Cost Slide 32 © 2020 M oderna Therapeutics
Slide 33: 3 recent examples pave our future
3 recent examples pave our future PCV: Needle-to-Needle CMV SARS-CoV-2 Antigen UL131 gL UL128 gH UL130 gB Muscle cell Cytosol Slide 33 © 2020 M oderna Therapeutics
Slide 34: Case Study 1: PCV
Case Study 1: PCV In a matter of a few weeks Up to 200 patients for proof Coincides with Keytruda Scale “out” (not up) of concept (phase 2) dosing regiment Slide 34 © 2020 M oderna Therapeutics
Slide 35: Learning curve
CASE STUDY 1: PCV Learning curve Quality Speed Scale Cost Success rate Cycle time Yield change Capacity change Slide 35 © 2020 M oderna Therapeutics
Slide 36: Learning
CASE STUDY 1: PCV Learning >90 UNIQUE PRODUCTS Slide 36 © 2020 M oderna Therapeutics
Slide 37: Increase learning…
CASE STUDY 1: PCV Increase learning… Phase 1 Success rate: Phase 2 Success rate: 69% 87% Quality Cost Number of Patient Batches Number of Patient Batches Slide 37 © 2020 M oderna Therapeutics
Slide 38: Multiple patient batches, same performance
CASE STUDY 1: PCV Multiple patient batches, same performance Quality Cost Lot 1 Lot 2 Lot 3 Lot 4 Lot 5 Lot 6 Lot 7 Lot 8 Lot 9 Lot 10 Lot 11 … Lot 90 Assay Results Slide 38 © 2020 M oderna Therapeutics
Slide 39: …Increase speed
CASE STUDY 1: PCV …Increase speed Speed 40% faster since 2018 Slide 39 © 2020 M oderna Therapeutics
Slide 40: Cost improvement
CASE STUDY 1: PCV Cost improvement Total $/Released Batch 65% 42% Cost 40% 2017/2018 2019 2020 Patients served 34 54 >90 Slide 40 © 2020 M oderna Therapeutics
Slide 41: Case Study 2: CMV
Case Study 2: CMV mRNA coding for: gH gL UL131 UL128 Pentamer UL130 gB gL UL131 gB UL128 Pentameric viral antigen gH UL130 gB Muscle cell Muscle cell Viral protein chains Cytosol Ribosome Cytosol Nucleus Encoded mRNAs Slide 41 © 2020 M oderna Therapeutics
Slide 42: CASE STUDY 2: CMV
CASE STUDY 2: CMV Slide 42 © 2020 M oderna Therapeutics
Slide 43: Create Make
CASE STUDY 2: CMV Quality Create Speed of technical development Make Slide 43 © 2020 M oderna Therapeutics
Slide 44: From few to many
CASE STUDY 2: CMV From few to many Speed Scale 5g 75g 150g Slide 44 © 2020 M oderna Therapeutics
Slide 45: Manufacturing cost reduction
CASE STUDY 2: CMV Manufacturing cost reduction Drug Substance $/mg 69% Cost 2019 Post 2020 Future Norwood Norwood goal (10g) (75g) (150g) Slide 45 © 2020 M oderna Therapeutics
Slide 46: CMV vials already produced for Phase 2 trial
CMV vials already produced for Phase 2 trial • Same lyophilized image intended for Phase 3 • Norwood site can produce Phase 3 and support commercial launch with fill/finish done at CMO • Over 10+ million doses/year could be made in Norwood Slide 46 © 2020 M oderna Therapeutics
Slide 47: CASE STUDY 3: SARS-CoV-2 Unplanned
CASE STUDY 3: SARS-CoV-2 Unplanned Slide 47 © 2020 M oderna Therapeutics
Slide 48: Convergence + Bold + Collaboration
CASE STUDY 3: SARS-CoV-2 Convergence + Bold + Collaboration Slide 48 © 2020 M oderna Therapeutics
Slide 49: Firing up on all engines…
CASE STUDY 3: SARS-CoV-2 Firing up on all engines… PRECLINICAL PERSONALIZED VACCINE UNIT CLINICAL Batches 23,500 1 batch/1 patient; 90 to date 1 batch/many patients Scale 10mg >100mg 5g-75g+ Slide 49 © 2020 M oderna Therapeutics
Slide 50: SARS-CoV-2
CASE STUDY 3: SARS-CoV-2 SARS-CoV-2 Digital sequence Speed engineering and Manufacture plasmid, DNA design mRNA, LNP Technical development and Selecting amino documentation 1st Shipment to acid sequence preparation the clinic • Jan 13 • Feb 7 • Feb 23 Day 1 Day 2-4 Day 5-17 Day 17-25 Day 41-42 Slide
Slide 51: CASE STUDY 3: SARS-CoV-2
CASE STUDY 3: SARS-CoV-2 Image Source: Success Kid meme. Know Your Meme, Literally Media, 26 Aug. 2007, knowyourmeme.com/memes/success-kid-i-hate-sandcastles Slide 51 © 2020 M oderna Therapeutics
Slide 52: Case studies
Case studies PCV: Needle-to-Needle CMV SARS-CoV-2 Antigen UL131 gL UL128 gH UL130 gB Muscle cell Cytosol Slide 52 © 2020 M oderna Therapeutics
Slide 53: Paving the future…
Paving the future… Quality Speed Scale Cost …we can do this! Slide 53 © 2020 M oderna Therapeutics
Slide 54: Q&A
Q&A Slide 54 © 2020 M oderna Therapeutics
Slide 55: Manufacturing & Digital Day Agenda
Manufacturing & Digital Day Agenda Introduction Stéphane Bancel, CEO 1:00–1:20 PM 20 min Norwood Overview Juan Andres, Chief Technical Operations & Quality Officer 1:20–1:50 PM 30 min Q&A Session 1:50–2:05 PM 15 min Competing in the Age of AI Karim R. Lakhani, Charles E. Wilson Professor of Business
Slide 56: Competing in the Age of AI
Competing in the Age of AI Marco Iansiti | miansiti@hbs.edu | @marcoiansiti Karim R. Lakhani | k@hbs.edu | @klakhani
Slide 57: Slide 57
Slide 58: Definitions of AI
Definitions of AI “Machines that can think and act in a way that matches or surpasses human intelligence” (“Strong AI”) Any activity computers are able to perform that humans once performed” (“Weak AI”) Nils J. Nilsson, 2010; Stanford AI 100 report, 2016. 58
Slide 59: Example: Pig Counting
Example: Pig Counting Difficulties in manual pig counting: ❑ Pigs have similar appearance and move around quickly ❑ Large overlapping areas between pigs due to high breeding density ❑ Potential disease infection from frequent interaction between pigs and humans
Slide 60: Example: Radiation therapy targeting
Example: Radiation therapy targeting 60
Slide 61: Broadly deployed weak AI is changing the nature of firms
Broadly deployed “weak” AI is changing the nature of firms: The Ant Financial Story Users (MM) 1400 1200 1000 800 600 400 200 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Year
Slide 62: Broadly deployed weak AI is changing the nature of firms
Broadly deployed “weak” AI is changing the nature of firms: The Ant Financial Story Users (MM) Value ($B) ?? 1400 1200 1000 150 800 600 100 400 200 50 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Year
Slide 63: How firms create, capture and deliver value
To understand companies like Ant Financial we need to reflect on how firms create, capture and deliver value Business Model Operating Model Value Creation Value Capture Scale Scope Learning
Slide 64: Traditional Operating Performance Drivers
Traditional Operating Performance Drivers Operating Model Scale Scope Learning Volume Variety R&D • Platform • Cont. improvement Complexity IT Range IT IT • IT Customers Data Data IT IP generation Data Data Data
Slide 65: Traditional operating models can constrain the value of the firm
Traditional operating models can constrain the value of the firm Value Delivered Operational complexity Administrative costs Organizational inertia … Traditional Diminishing Returns N (scale, scope, demands on learning)
Slide 66: Firms like Ant use AI to drive digital scale, scope, and learning
Firms like Ant use AI to drive digital scale, scope, and learning PAYMENT ECOMMERCE Applications CONSUMERS CREDIT Platform APIs Software Components Libraries Data INSURANCE, INVESTMENTS Etc… At Ant traditional human centered processes are digitized to connect with market opportunities at near zero m
Slide 67: Ant's operating model is enabled by an AI Factory
Ant’s operating model is enabled by an “AI Factory” AI Factory Operating Model Core 67
Slide 68: Learning effects and network effects can combine to drive value
Learning effects and network effects can combine to drive value Impact of AI (scale, scope and learning)
Slide 69: From Disruption to Collision
From Disruption to “Collision”… What happens when a firm that has competed in a traditional Applications market starts to compete with a firm based on a digital operating APIs Software Components Libraries model? Data Platform • Value IT IT • IT IT IT • Data Data Data Data Data Traditional product b
Slide 70: The New Meta
The New Meta HARVARD|BUSINESS SCHOOL
Slide 71: Universality
Universality From media to financial services, our economy is increasingly dominated by digital market dynamics and digital capabilities AI/ML Use Cases Source: BCG Source: Keystone Strategy LLC
Slide 72: New Opportunities
New Opportunities AI can improve a virtually infinite variety of compelling use cases, cutting across every segment of our economy Customer Intelligence Market Diagnosis & Fraud Analysis & & Recommendations Intelligence Treatment Systems Investigation Customer-facing Retail Edge Surveillance Smart C
Slide 73: New Strategies….
New Strategies…. Not all “digital” firms are the same!!! Digital networks and AI are adding new (frequently subtle) dimensions to strategy and competition.
Slide 74: …driving the emergence of a new kind of firm.
…driving the emergence of a new kind of firm. Market capitalization per employee $10-30 M $3 M $2.5 M $1 M $250K $100K Business “Digital” Firm Financial Services Semiconductor Service Provider Manufacturing Retail Model: Design/IP (telco) (assembly) (traditional) Example:
Slide 75: …driving the emergence of a new kind of firm.
…driving the emergence of a new kind of firm. Market capitalization per employee $10-30 M $3 M $2.5 M $1 M $250K $100K Business “Digital” Firm Financial Services Semiconductor Service Provider Manufacturing Retail Model: Design/IP (telco) (assembly) (traditional) Example:
Slide 76: Competing in the Age of AI
a au qT] | COMPETING 5 oil : [aA “IN THE |) in A= Peeve Strategy and Leadership When Algorithms and Networks Run the World miley, MARCO IANSITI KARIM R. LAKHANI
Slide 77: Competing in the Age of AI
Competing in the Age of AI Marco Iansiti | miansiti@hbs.edu | @marcoiansiti Karim R. Lakhani | k@hbs.edu | @klakhani
Slide 78: Moderna's Digital Strategy
Moderna’s Digital Strategy Marcello Damiani March 4, 2020 © 2020 M oderna Therapeutics
Slide 79: Marcello Damiani, Chief Digital and Operational Excellence Officer
Marcello Damiani, Chief Digital and Operational Excellence Officer As Chief Digital and Operational Excellence Officer, Marcello Damiani brings a wealth of experience from both Tech companies and Life Science companies and a true passion for making Digital a reality. He has 20 years experience worki
Slide 80: Data: The most valuable resource
Data: The most valuable resource Source: The w orld's most valuable resource is no longer oil, but data w ww.economist.com/leaders/2017/05/06/the-w orlds-most-valuable-resource-is-no-longer-oil-but-data Slide 80 © 2020 M oderna Therapeutics
Slide 81: Digital a key enabler for advanced analytics & AI
Digital a key enabler for advanced analytics & AI AI/Machine learning Unified data sources Digital transformation Requires large, structured, well- Diverse internal and external data Create a data-centric organization to annotated datasets sources linked and made accessible build foundation for AI S
Slide 82: Moderna 10-year vision
Moderna 10-year vision We are a leading mRNA platform and continue to invest like it’s the beginning of the journey We and our partners are discovering and developing an unprecedented breadth of meaningful investigational vaccines and therapeutics We integrate automation and digital technology into
Slide 83: Moderna 10-year vision
Moderna 10-year vision We are a leading mRNA platform and continue to invest like it’s the beginning of the journey We and our partners are discovering and developing an unprecedented breadth of meaningful investigational vaccines and therapeutics We integrate automation and digital technology into
Slide 84: Why build a digital biotech?
Why build a digital biotech? Ribosome Proteins mRNA mRNA technology Platform approach Business strategy • Instruct cells to produce protein • Modalities: combine • Massively parallel R&D • Instructions are code — Chemistry • Ecosystem model • The very essence of our — Bioinformatics • Progress concu
Slide 85: Moderna's digitization building blocks
Moderna’s digitization building blocks DIGITAL ARTIFICIAL INTELLIGENCE ANALYTICS AUTOMATION & ROBOTICS INTERNET OF THINGS (IOT) INTEGRATION CLOUD Slide 85 © 2020 M oderna Therapeutics
Slide 86: Building a digital biotech
Building a digital biotech Quality Speed Scale Cost Reduce errors Accelerate learning Support Increasing demand Reduce waste Increased probability of success Increased agility to change Sustain and repeat processes Improve performance Slide 86 © 2020 M oderna Therapeutics
Slide 87: The digital ecosystem @ Moderna
The digital ecosystem @ Moderna People Processes Technology Digital culture Value stream design Flexible User experience Lean/6σ principles Future proof User adoption Compliance Innovative Slide 87 © 2020 M oderna Therapeutics
Slide 88: Digital integration @ Moderna
Digital integration @ Moderna Technical Development More Better data algorithms Research Quality More Better trials, mRNA experiments medicines Clinical Clinical Manufacturing Development Slide 88 © 2020 M oderna Therapeutics
Slide 89: Digital integration @ Moderna: Research
Digital integration @ Moderna Advancing ideas to development candidates Research Slide 89 © 2020 M oderna Therapeutics
Slide 90: Research digitization
Research digitization Design mRNA • Intuitive design tools for building mRNA sequences • Workflow tools for codon and nucleotide options • AI-algorithms for back translation and sequence optimization to design the best mRNA Scientists have designed over 20,000 unique mRNA constructs Slide 90 © 2
Slide 91: Research digitization
Research digitization Order mRNA Design mRNA • Intuitive, streamlined • Intuitive design interface for rapid tools and for building mRNA accurate ordering sequences • Runs several automated • Workflow quality checks and tools forAI optimization algorithms codon and nucleotide • options Automate
Slide 92: Research digitization
Research digitization Order mRNA Design Production mRNA and • Intuitive, streamlined QC of Preclinical • Intuitive interface design for rapid tools and mRNA for building mRNA accurate ordering sequences • Integrated Runs severalwith over 30 automated • robotics quality Workflow checks and platfor
Slide 93: Research digitization
Research digitization Inventory and Registry of mRNA • Full chain-of-custody and audit trails for materials to ensure quality of data using samples • Registration of every unique entity from its first conception • Downstream applications point to unique registry IDs to tie together all Stores ~1M
Slide 94: Research digitization
Research digitization Study Design Inventory and and Management Registry of mRNA • Complex study design • Full toolschain-of-custody for capturing and audit trails for detailed metadata of materials to ensure all studies needed for better analysis quality of data using samples • Integration with
Slide 95: Research digitization
Research digitization Study InventoryDesign and and Management Registry of mRNA Automated • Complex study design •Data Full Analysis toolschain-of-custody for capturing and audit trails for • detailed metadata Over 2 dozen of materials all todata ensure studies needed automated for better qualitya
Slide 96: Digital integration @ Moderna: Technical Development
Digital integration @ Moderna Transforming development candidates into medicines Technical Development Slide 96 © 2020 M oderna Therapeutics
Slide 97: Tech dev digitization
Tech dev digitization Recipe Editor Development / / Hub Process Templating Experiment Execution •• New product Structured recipes capture of can be built and process parameters, deployed materials in and just a few hours analytical results •• Support IntegratedModerna’s sample core platformand submi
Slide 98: Tech dev digitization
Tech dev digitization Development Hub / Experiment Execution • Structured capture of process parameters, materials and analytical results • Integrated sample submissions and results, for high- throughput assay processing • Integrated powerful data analytics, perform statistical analysis of process a
Slide 99: Digital integration @ Moderna: Manufacturing
Digital integration @ Moderna Manufacturing investigational medicines Clinical Manufacturing Slide 99 © 2020 M oderna Therapeutics
Slide 100: Manufacturing digitization
Manufacturing digitization Materials • The operators are digitally instructed to which raw materials to use through a real-time integration with SAP • Materials are barcoded, scanned, mixed and charged to ensure accurate inventory controls • The transformation of products is digitally tracked from r
Slide 101: Manufacturing digitization
Manufacturing digitization Materials Equipment • The operators are digitally • instructed Operators to which areraw digitally materials to use through a instructed real-time to which integration with SAP equipment to use • Materials are barcoded, • scanned, Readiness mixed and charged of the to ensu
Slide 102: Manufacturing digitization
Manufacturing digitization Materials Analytics EquipmentDriving Operations • The operators are digitally • instructed Operators to which areraw digitally • Ability materialstotoanalyze instructed use through a to which real-time integration with SAP personnel and materials equipment flows to drivet
Slide 103: Digital integration @ Moderna: Quality
Digital integration @ Moderna Testing and releasing medicines for intended use Quality Slide 103 © 2020 M oderna Therapeutics
Slide 104: Quality digitization
Quality digitization Product & Raw Material Testing • Operators are digitally instructed to what samples to take • Real-time integration with the LIMS sample plans • Sample labels are printed and applied on the shop-floor • All information required for testing is passed digitally to the LIMS system
Slide 105: Quality digitization
Quality digitization Product & Raw Material Testing Release Laboratory Data • Operators are digitally •Integrity QC test results are instructed to what samples passed electronically to take •• with QC release labs Real-time haveconditions anwith integration integrated the LIMS sample dataplans • Bat
Slide 106: Quality digitization
Quality digitization Product & Raw Material Testing •Laboratory Data Operators are digitally instructed to what samples Integrity to take •• QC labsintegration Real-time have anwith App access Gateway integrated the data LIMS sample plans Non compliant • integrity solution Sample labels are printed
Slide 107: Quality digitization
Quality digitization Product & Raw Material Testing Release Environmental •Laboratory Data Operators are digitally •Monitoring QC test results are instructed to what samples Integrity passed electronically to take • Self-contained •• with QC release labs Real-time testing haveconditions carts anwith
Slide 108: Digital integration @ Moderna: Clinical Development
Digital integration @ Moderna Advancing new medicines to patients Clinical Development Slide 108 © 2020 M oderna Therapeutics
Slide 109: Clinical development digitization
Clinical development digitization Clinical Trial Dashboard • Roles and responsibilities of the clinical trial team • High level summary of key operational milestones • Summary text informs team of important updates Slide 109 © 2020 M oderna Therapeutics
Slide 110: Clinical development digitization
Clinical development digitization Clinical Trial Enrollment Dashboard Forecasting •• Roles and subject Track actual responsibilities enrollment relative of to the the clinical plan trial team • Real-time detection • High level summary of enrollment issues of key operational • Enrollment milestones t
Slide 111: Clinical development digitization
Clinical development digitization Milestone Clinical Trial Enrollment Tracking Dashboard (Target vs. Actuals) Forecasting •• Plan Roles Trackand andmonitor actual key subject responsibilities milestones across all enrollment relative of Moderna to the the clinical planstudies trial team •• Easily Re
Slide 112: Clinical development digitization
Clinical development digitization Milestone Clinical Trial Enrollment Tracking Dashboard (Target vs. Forecasting Site Visit Actuals) Monitoring •• Plan Roles Trackand and Visibility ofmonitor actual site key subject responsibilities milestones across trainingall enrollment relative monitoring, of Mo
Slide 113: Thank you!
Thank you! Slide 113 © 2020 M oderna Therapeutics
Slide 114: Advanced Analytics and AI @ Moderna
Advanced Analytics and AI @ Moderna Dave Johnson, PhD March 4, 2020 © 2020 M oderna Therapeutics
Slide 115: Dave Johnson, PhD
Dave Johnson, PhD Dave Johnson is Head of Informatics, Data Science, and AI at Moderna, where he delivers innovative Digital and AI solutions to support Research and Clinical Development. Mr. Johnson holds a PhD in Information Physics and has over 15 years of experience in software engineering and d
Slide 116: Advanced analytics and AI enables various parts of our platform
Advanced analytics and AI Quality Scalability enables various parts of our platform Cost efficiency Speed Algorithms In Production • Long Sequence Deconstruction Used in SARS-CoV-2 Vaccine • mRNA Sequence Optimization • PCV Supply Chain Simulation • Vaccine design pipeline for the Personalized Cance
Slide 117: Advanced analytics and AI enables various parts of our platform
Advanced analytics and AI Quality Scalability enables various parts of our platform Cost efficiency Speed Algorithms In Production • Long Sequence Deconstruction • mRNA Sequence Optimization • PCV Supply Chain Simulation Used in Personalized Cancer Vaccine • Vaccine design pipeline for the Personali
Slide 118: Advanced analytics and AI enables various parts of our platform
Advanced analytics and AI Quality Scalability enables various parts of our platform Cost efficiency Speed Algorithms In Production Algorithms In Development / Testing Numerous Computational Techniques • Long Sequence Deconstruction • Improved mRNA sequence optimization • Rule based optimization • mR
Slide 119: AI use cases
AI use cases Sanger Sequencing Analysis CryoEM Image Analysis ̶ Automated analysis pipeline ̶ Neural network image segmentation ̶ Poly-A tail neural network QC ̶ Detection/classification of particle types ̶ Bayesian inference algorithm ̶ Nanometer scale LNP morphology Slide 119 © 2020 M oderna Thera
Slide 120: Sanger sequencing
Sanger sequencing Input DNA • Used to QC our DNA templates and final research grade mRNA • A single nucleotide change can result in a nonfunctioning protein Slide 120 © 2020 M oderna Therapeutics
Slide 121: Sanger data analysis
Sanger data analysis TARGET • Analysis requires a complex pipeline of algorithms that does: CONSENSUS 1. Peak detection 2. Base calling 3. Contig building 4. Consensus calling READ 5. Discrepancy identification • We licensed an off-the-shelf pipeline READ of algorithms • Standard analysis tool for 2
Slide 122: Automated Sanger analysis
Automated Sanger analysis Sequencing Vendor Spot Compute Instance Instrument Spot Compute Instance Instrument Raw Data Spot Compute Moderna Cloud Instance Sanger Storage App Instrument … Spot Compute Instance Slide 122 © 2020 M oderna Therapeutics
Slide 123: Automated Sanger analysis
Automated Sanger analysis Sequencing Vendor Instrument Instrument Raw Data Moderna Cloud Sanger Storage App Instrument … Slide 123 © 2020 M oderna Therapeutics
Slide 124: Visualization tool
Visualization tool Slide 124 © 2020 M oderna Therapeutics
Slide 125: Analyses and data growth
Analyses and data growth 3,500,000 3,000,000 2,500,000 Number of Reads 2,000,000 We’ve analyzed over 1,500,000 3 million Sanger reads of raw data 1,000,000 500,000 0 Dec-16 Jun-17 Dec-17 Jun-18 Dec-18 Jun-19 Dec-19 Slide 125 © 2020 M oderna Therapeutics
Slide 126: Poly-A tails
Poly-A tails • Our mRNA typically ends with a poly-A 100 tail AAAAA~100 • We build the poly-A 100 tail into our DNA templates Slide 126 © 2020 M oderna Therapeutics
Slide 127: Sanger analysis of poly-A tails
Sanger analysis of poly-A tails Good Tail • The off-the-shelf algorithm cannot distinguish these • Need to be reviewed manually by trained operators Bad Tail Slide 127 © 2020 M oderna Therapeutics
Slide 128: Human calling is expensive and inconsistent
Human calling is expensive and inconsistent Operator 1 — Operator 2 — 64.4% Operator 3 — 55.1% 67.6% Operator 4 — 58.5% 71.0% 84.4% Operator 5 — 73.7% 72.4% 62.0% 69.1% • 10 operators manually created over 20k Operator 6 — 69.7% 64.9% 77.6% 73.5% 68.9% labels of pass/fail tail chromatograms Operator
Slide 129: Human calling is expensive and inconsistent
Human calling is expensive and inconsistent Operator 1 — 92.2% Operator 2 — 64.4% 78.3% Operator 3 — 55.1% 67.6% 91.9% Operator 4 — 58.5% 71.0% 84.4% 88.9% Operator 5 — 73.7% 72.4% 62.0% 69.1% 96.6% • 10 operators manually created over 20k Operator 6 — 69.7% 64.9% 77.6% 73.5% 68.9% 87.5% labels of p
Slide 130: True-positive rate
Poly-A CNN model Human Labelers 1 Lenient operators Poor performers Strict operators True-positive rate 0.0 0.2 0.4 0.6 0.8 1.0 False-positive rate Slide 130 © 2020 M oderna Therapeutics
Slide 131: True-positive rate
Poly-A CNN model Human Labelers 1 Balanced Algorithm • We built a convolutional True-positive rate neural network (CNN) to predict call of entire team • Running in production for over 1 year, saving countless hours of manual review 0.0 0.2 0.4 0.6 0.8 1.0 False-positive rate Slide 131 © 2020 M odern
Slide 132: Other issues with current algorithm
Other issues with current algorithm Mixed Base • We’ve discovered many other failure modes that are not being handled correctly by the off- the-shelf algorithm: — Mixed bases — Contamination — Mispriming — Failed contigs 30% of population has — Peak drift a mutation of C → G — Etc. Slide 132 © 2020
Slide 133: A new algorithm from the ground up
A new algorithm from the ground up Peak Detection • Using Bayesian MAP analysis, model peaks as superposition of gaussians • Based on statistics of huge number of + + = C+G historical reads (73.5% C, 26.5% G) Base Calling • Use Bayesian Hypothesis testing, calculate probabilities of all possible bas
Slide 134: AI use cases
AI use cases Sanger Sequencing Analysis CryoEM Image Analysis ̶ Automated analysis pipeline ̶ Neural network image segmentation ̶ Poly-A tail neural network QC ̶ Detection/classification of particle types ̶ Bayesian inference algorithm ̶ Nanometer scale LNP morphology Slide 134 © 2020 M oderna Thera
Slide 135: CryoEMs are images of our lipid nanoparticles
CryoEMs are images of our lipid nanoparticles Components Assembled LNPs CryoEM Images • Understand how process/components impact assembly • Understand how LNP structure impacts in vivo performance • Have generated 500k+ images across numerous conditions Slide 135 © 2020 M oderna Therapeutics
Slide 136: Neural network for image segmentation
Neural network for image segmentation • The first step is Convolutional Encoder-Decoder to segment the image into individual LNP particles • We trained a state-of-the-art convolutional encoder-decoder Conv + Batch Normalization + ReLU Pooling Upsampling Softmax neural network Slide 136 © 2020 M oder
Slide 137: Neural network for image segmentation
Neural network for image segmentation • The first step is Convolutional Encoder-Decoder to segment the image into individual LNP particles • We trained a state-of-the-art convolutional encoder-decoder Conv + Batch Normalization + ReLU Pooling Upsampling Softmax neural network Slide 137 © 2020 M oder
Slide 138: Automatic detection and classification of particle types
Automatic detection and classification of particle types Bleb Non-bleb Frequency • Pixel-value based random forest model achieves excellent 20 40 60 80 100 performance Major axis length (nm) • Saves countless hours of manual Model Accuracy Precision analysis Logistic Regression 0.87 0.75 Random Fore
Slide 139: Measurement of nanometer scale LNP morphology
Measurement of nanometer scale LNP morphology Extract radial profiles, align edges Average within particles Average across particles Average Lamellar formulation profiles width Pixel Intensity Pixel Intensity Distance (nm) Distance (nm) Slide 139 © 2020 M oderna Therapeutics
Slide 140: Moderna's AI factory
Moderna’s AI factory Structured Data Data Capture & Integration Algorithm Rapid Research & Production Development AI Model Deployment Design Cloud & DevOps Infrastructure Quality Speed Scale Cost Slide 140 © 2020 M oderna Therapeutics
Slide 141: Q&A
Q&A Slide 141 © 2020 M oderna Therapeutics
Slide 142: mRNA as a potential new class of medicines
mRNA as a potential new class of medicines 1. Large product opportunity 2. Higher probability of technical success 3. Accelerated research and development timelines 4. Greater capital efficiency over time vs. recombinant technology Slide 142 © 2020 M oderna Therapeutics
Slide 143: Thank you!
Thank you! Slide 143 © 2020 M oderna Therapeutics