PowerPoint Presentation

Modeling Drug and Medical Device Innovation as Temporal Sequences using EventFlowNIH and the Science of Science and Innovation Policy:A Joint NIH-NSF WorkshopApril 7 8Bethesda MarylandC. Scott Dempwolf, PhDAssistant Research ProfessorUniversity of Maryland Morgan State Joint Center for Economic DevelopmentBen Shneiderman, PhDDistinguished University ProfessorUniversity of Maryland Institute for Advanced Computer Science (UMIACS)

(and a few networks)

1

Pennsylvania Innovation Networks 1990 2007Emergence of Philadelphia Biopharma cluster and Pittsburgh Nuclear ClusterModeled with Pajek & KING

2010

ME: Its cool, butHow do I make it useful?

BEN:You must use NodeXL

ME:Obiwan Shneiderman, you are my Jedi Master

3:002

Innovation

A process of transforming knowledge and scientific research into a new product in the marketplace.Think of that process as a sequence of related activities ResearchInventionProofCommercializationProductWith this intended outcome

This slide begins a short series of slides that starts with a broad definition of innovation, then establishes a framework for thinking about innovation as a temporal sequence of connected events. Each slide in the series adds a layer of information that transforms the abstract concept of innovation into a data-driven model.3

InnovationEach activity has inputs, outputs, associated documents and artifactsWith this intended outcome

Inputs and outputs are the basis for how we measure innovation now and are a familiar reference point. Associating individual activities with documents or artifacts sets up the linkages to data sources.4

InnovationEach activity involves people and organizations producing intermediate outcomesContributing to this intended innovation outcome

From the data sources we can identify people and organizations involved, which sets up the creation of network models. This slide also introduces the notion of intermediate outcomes from the innovation perspective things like patents and publications that may be final outcomes from the perspective of the people working on them. Innovation is not everyones goal.5

Innovation

The people and organizations from each activitycreate an activity network

This introduces the idea of an activity network the building block of innovation networks and ecosystems.6

Activities become sequences through shared people and organizations, citations, and other linkagesWith this intended outcome

Finally we layer on the data and linkages to show how the activities connect to each other to form temporal sequences.7

Innovation Ecosystems

Innovation networks with embedded knowledge & resources along with intermediaries comprise Innovation Ecosystems.Activity networks combine to form innovation networks.The Regenerative Medicine cluster (ecosystem) in Howard County, MDCombining two activities: NSF# 1551041 and todays presentationNSF# 1551041 activity network

This slide parallels the previous slide but from the activity network to innovation network to innovation ecosystem perspective.8

Innovation MetricsSome are based on organizations & resourcesNone are based on intended outcome

Some arebased on inputsSome arebased on outputsSome are based on talentSome are comparative indexes

Where innovation metrics come from now, and more importantly, where they do not. Total elapsed time to the end of this slide is 4:00.9

Modeling Innovation Sequences with EventFlow

We use newly developed EventFlow software to model innovation in drugs and medical devices from multiple datasets:RePORTER_PATENTS_C_ALLRePORTER_CLINICAL_STUDIES_C_ALLCTTI AACT DatabaseFDA Orange Book (drugs)Drugs@FDAPre-Market Approvals (PMA) (med devices)SBIR/STTR (pending)CrunchBase (pending)NSF (pending)

Supporting and core data sourcesNIH RePORTERPatentsViewUSASpendingSTARMETRICS

http://hcil.umd.edu/eventflow/

This slide shows the EventFlow screen with the overall data model we have constructed so far, and identifies the data sources we use.10

A Quick Tour of EventFlow

Each product (drug or medical device) is a record in EventFlow (34,331 records)

Event categories: Clinical Trials (commercialization activity)FDA Approval (proxy for product launch)Patents (invention)Research

Overview (Aggregation) Individual Timelines

Introduces the basics of EventFlow11

Product-Based Innovation Metrics

Temporal MetricsHow long does innovation take?How many activities are involved? What types? In what sequence? How long does each take? Are there gaps?Is the sequence pattern common or rare?

This slide shows the 26 drugs that we can trace at least partially from research to final approval. It also frames innovation activities from the perspective of products and lists some of the important questions we can ask and important temporal metrics we can derive from analysis with EventFlow (and CoCo).12

How long does innovation take? (drugs)

From: Patent application FDA approval

(26 products)

The distribution and statistics for our 26 drug sample. Measures from first patent application date to final approval date.13

How long does innovation take? (drugs)From: Patent application FDA approval (product launch)

(884 drugs in the FDA Orange Book)

The distribution and statistics for all drugs we have data on. Measures from first patent application date to final approval date.14

How long does innovation take? (med devices)

From: Start of clinical trials FDA approval

(1,225 medical devices)

A similar look at medical devices using clinical trials and final approval dates. Introduces the measurement of gaps.15

How long does innovation take? (med devices)

FDA Approval during Clinical TrialFDA Approval afterClinical Trial

Continues the previous slide, introducing the concepts of overlaps and of visualizing patterns in the overview panel of EventFlow.

There are two general sequence patterns in the data. The images on the previous slide showed sequence patterns for which clinical trials were completed, followed by a lag, followed by FDA approval. The other pattern, shown here on the right has FDA approvals overlapping the span of clinical trials.

16

Illinois Battery Cluster 2010 2014Modeled with NodeXL

BridgeBroader applications of temporal metrics:the Illinois Battery ClusterInnovation Ecosystemsresearch componentIndustry componentBridging component

We prepared network models fir the Illinois Science & Technology Roadmap. One of those was for the Battery Cluster. From an academic perspective, one of the things we noticed was that the activities tended to be organized into two main components research and industry with a small group of activities that seemed to span these two components. Well call this third component the bridge show here in collapsed form as simply a gray band.

17

BridgeThe Innovation Ecosystem and the Valley of DeathA network representationof the valley of death

Then we recognized similarities between the network graph and our graphic of the so-called valley of death. We realized that we might actually be looking at a network representation of that valley of death. If that was the case, what would we expect to find in the bridge? 1) corporate sponsored research; 2) SBIR / STTRs; 3) intermediaries like accelerators and incubators; and 4) public-private partnerships like federal labs. When we opened up the bridge for a closer look, that is exactly what we found.18

Emerging Theory & Research

Bridge

Whats in the Bridge?Working HypothesisRegions with denser, more connected bridging components will be characterized by faster innovation sequences and more innovation sequences leading to new products.

Measured using new temporal metrics

This allows us to frame a new working hypothesis about innovation ecosystems and the differences in innovation outcomes between different regions. And the new temporal metrics we are developing with NSF should allow us to test that hypothesis soon. So for us the value of University Centers is this diversity of partnerships around the tasks of developing new methods and metrics; developing new tools and solving practical economic development problems; and synthesizing those activities into new knowledge and understanding about the nature of innovation and its impact on economic growth.19

Stem cell products groupCommercialization supportAccelerationAttract complementary firmsDelivery devices groups,ECM groupFacilitate collaborationNiche market developmentAttract complementary firmsRegenerative Medicine & Nutraceuticals groupsDevelop KeystonesPromote local sourcing Industry partnershipsFDI / Business expansionAttraction - supply chain University partnershipsUniversity groups (JHU, UMCP, UMB)Leads for licensing (green ties)Key labs (dense subgroups)Opportunities for faculty spin outsAccelerate student startupsCorporate PartnershipsTargeted Economic Development StrategiesAt the Cluster Level

Regenerative Medicine Cluster Howard County, MDInnovation-Led Economic Development

Drill-down to Company ProfilesClick to follow link

Nascent / emerging Growth stageInfrastructure for maturing cluster~Labs

As a practical matter innovation ecosystems and regional innovation clusters are the same thing. Here we show one ecosystem / cluster for regenerative medicine in Howard County Maryland, comprised of those aggregated activity networks. This small, emerging cluster does not show up in traditional cluster analysis because 1) the activity is too recent and 2) the activity is not organized according to existing NAICS codes. Thus this analysis was valuable to Howard County Economic Development. Each group includes people and organizations that are connected based on what they are working on together. The graph is organized with the largest, most connected group in the upper left and the smallest, least connected group in the lower right. It turns out that this layout is useful in helping to organize and target different types of economic development strategies to specific companies and groups so that the overall cluster strategy appropriately targets limited resources for effective economic development. The interactive network tool allows economic developers to zoom in and explore different parts of the cluster in detail. Users can also click on certain nodes to get more detailed information. This interactive network model was build using NodeXL developed in part by the same computer scientists who created EventFlow.20

Howard County, Maryland - Full Innovation Network

Universities (JHU, UMCP, UMB, UMBC+)Follow-up leads for licensing or other engagements (green ties)Identify key labs (dense subgroups) and evaluate for expansion / enhancementIdentify opportunities for faculty spin outsIdentify / accelerate potential student startups that can be seeded in this clusterBuild long-term sponsored research relationships with keystone companiesMain Innovation ClustersRegenerative MedicineTelecom / networks / cyberDefense / Security / SBIRNutraceuticalsResearch & DevelopmentEntrepreneurial Acceleration OpportunitiesCommercialization, acceleration, entrepreneurial support for early stage companies located in the countyAssistance with market Connections to capital & cluster keystonesBusiness Attraction OpportunitiesFocus on early stage companies with innovation cluster growth potential; companies are located outside of the county but have a HoCo connectionDevelop relationships and help them plan for move to HoCo for next growth stageConnections to capitalKeystonesIdentify & cultivate keystones in each innovation clusterIdentify & cultivate capital networks around each innovation clusterBusiness Expansion & FDI OpportunitiesFocus BRE on growth stage & mature companies in innovation clusters. Develop keystones in the process.Engage MD DOC in developing FDI.Engage foreign-owned companies in innovation clusters to expand their presence in the cluster through FDI.Workforce DevelopmentDevelop industry partnerships (EARN) around innovation clustersWork with universities & community colleges on talent pipeline

Federal Strategy

pending

Strategy Gradient

The group-in-a-box layout organizes groups from largest to smallest. This also corresponds to a strategy gradient for economic development.Research ComponentEntrepreneurial strategiesAttraction strategiesResearch & Tech Transfer strategiesRetention, Expansion & Workforce strategies

Industry ComponentBridging Component (partial)

14:30 15:30 But here is how the visualization organizes the information.21

A few Data Issues & NeedsData cleaning & disambiguationData matching across datasetsRePORTER, Clinical Trials, FDA, SBIRMatching on full project numbers (not core)SBIR More complete dates; Access to bibliographies for citation linkagesFDA, Clinical Trials Basic information at the front-endFDA ability to roll up drug families i.e. Adderall 10mg, 15mg, 20mg

Upcoming EventsApril 13, Wednesday 10am at NIH Porter Building 35A, Room 610,NIH Main Campus, Bethesda, MD

Interactive Visual Discovery in Event Analytics:Electronic Health Records Ben Shneiderman

datascience.nih.gov/community/datascience-at-nih/frontiersMay 26, Thursday at University of Maryland Human-Computer Interaction Lab EventFlow Workshop

hcil.umd.edu/eventflow/ hcil.umd.edu/eventflow-workshop-2016/`

Implications for Universities: visualizing labs and research partnershipsIdentify key labs (dense subgroups) and evaluate for expansion / enhancement

Identify opportunities for faculty spin outs

Identify / accelerate potential student startups that can be seeded in emerging clusters

Link to Lab and researcher pages (click to follow)University of Maryland, College ParkResearch labs, research partnerships, and individual researchers

15:30 17:3024