《Citeline：2024探索AI技术在制药领域的潜在价值报告（英文版）（32页）.pdf》由会员分享，可在线阅读，更多相关《Citeline：2024探索AI技术在制药领域的潜在价值报告（英文版）（32页）.pdf（32页珍藏版）》请在三个皮匠报告上搜索。

1、CLICK TO EDIT MASTER TITLE STYLEAI and ML in Pharma:Redefining the Forecasting LandscapeJuly 2024Todays PresentersDANIEL CHANCELLORVP Thought Leadership,EvaluateDAVID JAMESFounder,J+D ForecastingSTEFANO DRIUSSIHead of Software Engineering,J+D Forecasting20+years of experience supporting Pharmaceutic

2、al and Biotech companies with their forecasting needs.Experts in all pharmaceutical forecasting methodologies.Using innovative approaches to resolve forecasting challenges.Led by a senior level team of forecasting,market research,technical and data analysis professionals,with extensive experience.Ad

3、vanced technical capabilities,having developed over 1,000 forecast models and deployed our FC+software and FC365 forecasting platform in 70+countries.J+D ForecastingBy combining Evaluates world-class consensus forecasting and consulting expertise with J+D Forecastings specialised models,delivered th

4、rough cloud-based management and analytical solutions,clients can achieve a comprehensive understanding of the competitive landscape,seize important opportunities,and enhance the decision-making process.Evaluate,a Norstella CompanyAgendaGeneral OverviewApplication of AI in PharmaPractical uses of AI

5、 in ForecastingAI Revolution What to ExpectQ&AAI in Pharma Forecasting Challenges and OpportunitiesGeneral Overview“A.I.could be more profound than both fire and electricity”Sundar Pichai,CEO AlphabetDifferent areas of Artificial Intelligence.Artificial Narrow Intelligence(better than humans in ONE(

6、better than humans in ONE specific task)specific task)Artificial General Intelligence(capable as humans in every(capable as humans in every task)task)Artificial Super Intelligence(better than humans in(better than humans in every task)every task)TodayTomorrowAI in Pharma Forecasting Challenges and O

7、pportunitiesApplication of AI in PharmaApplication of AI in Pharma Globally valued at$905 million in 2021 ($9,241 million by 2030)50%of global healthcare companies plan to implement AI strategies(by 2025)AI-driven new drug development expected to grow 40%annually ($4bn in 2024)AI-driven new drug dev

8、elopment expected to grow 40%annually grow 40%annually ($4bn in 2024)Clinical TrialsDrug DiscoveryDiagnosticsAIAI Applications in Pharma AI and ML in Pharma:Redefining the Forecasting LandscapeDrug DiscoveryDrug DiscoveryClinical TrialsClinical TrialsDiagnosticsDiagnosticsIdentify new drug molecules

9、 that have so far eluded scientistsSynthetic data can bridge population data gapsIncrease the probability of success of NCEsLower R&D costs&reduced time to marketAI within Drug DiscoveryArtificial intelligence has many implications for research,drug discovery and development and trials:12Already mak

10、ing drug discovery faster and cheaper,with a number of multiple AI-designed drugs now being tested in humans:Time for drug discovery80%Cost of drug discovery70%Drugs in development50(Estimated predictions)https:/ within Drug Discovery:Forecasting ImplicationsEst.around 270 companies currently workin

11、g on AI-driven drug discovery.Lack of in-house AI expertise driving vendor partnerships:GSK has partnered with Cloud Pharmaceuticals and Insilico Medicine to utilize their AI platforms for target identification,drug design,and lead generation.Sanofi partnered with Atomwise to discover and synthesise

12、 drug compounds for five different targets,paying$20 million upfront for their innovation and AI capabilities.Impact:The future competitive environment will change as more drugs are discovered.An increase in partnership deals between industry and vendors.Reduction in time to market and associated co

13、sts will change financial thresholds for new drugs.45%adopted AI for drug developmentInhouse vs Vendors17%vs 83%33%already included within R&D plansJ+D surveyAI and ML in Pharma:Redefining the Forecasting LandscapeDrug DiscoveryDrug DiscoveryClinical TrialsClinical TrialsDiagnosticsDiagnosticsAI wit

14、hin Clinical Trials.AI can help to make clinical trials more efficient,more accurate,and more effective.Around 90%of clinical trials run significantly over time or over budget.86%of clinical trials fail to recruit enough patients within their target time frame.Between 25%to 40%of trials will fail be

15、cause they cannot meet their goals.Patient recruitment and screeningAI can help to reduce the time and cost of clinical trials,and it can also help to ensure that the right patients are enrolled in the right trials.Risk assessmentAI can help to ensure the safety of patients,and it can also help to i

16、dentify patients who are at risk of experiencing adverse events.Decision supportAI can help investigators to make informed decisions about the design,conduct,and interpretation of clinical trials.Data collection and analysisAI can help to identify patterns and trends in the data,plus identify patien

17、ts who are most likely to benefit from a particular treatment.Predictive modellingAI can help develop predictive models to identify trials that are most likely to be successful,and those that are most likely to fail.Benefits of AI application:Success probabilityTime to launchKey finance metricsAI wi

18、thin Clinical Trials:Forecasting Implications.Impact:Decreasing drug development time and cost will require adjustments to forecast assumptions around probability of success,time to launch and key NPV assumptions.Reduction in associated costs50%Reduction in time to market50%(Estimated predictions)Es

19、t.Cost SavingsEst.Time SavingsJ+D surveyPotential to revolutionise the way clinical trials are conducted-estimated 50%reduction in time to market and associated costs.Improved,faster recruitment for clinical trials,reducing overall trial length and potential costs.Creating a more effective use of R&

20、D budget.https:/ and ML in Pharma:Redefining the Forecasting LandscapeDrug DiscoveryDrug DiscoveryClinical TrialsClinical TrialsDiagnosticsDiagnosticsAI within DiagnosticsHas the potential to make healthcare more accessible and affordable plus enhancing efficiency and accuracy of diagnostics.Challen

21、ges AI can address:Identifying at risk populations for early intervention.Diagnosis and decisions about treatment plans.Personalised treatment based on patients genetic makeup,resulting in better patient outcomes.Identifying Lung CancerIn a study of more than 42,000 low-dose computed tomography scan

22、s(LDCT),AI performed as well or better than six radiologists in its ability to detect lung cancer tumours.AI Detecting Heart DiseaseAI test providing higher diagnostic accuracy,reduces the need for unnecessary invasive angiograms by 83%and reduces healthcare system costs by 26%.Next-Generation Seque

23、ncingAccelerated genomic data analysis by 10-fold helping advance biomarker discovery,accelerate drug development and develop new diagnostic tools.AI in practice:Credit:NVIDIAAI within Diagnostics:Forecasting ImplicationsAI techniques are already being used to diagnose numerous diseases.How far,how

24、fast and how effective could these be?Ability to detect and diagnose rare diseases that until now,have been extremely difficult.Impact:Earlier diagnosis of diseases has implications on patient outcomes,therefore change patient distribution across lines of treatment or severity of disease.A significa

25、nt impact on diseases such as oncology which the forecaster will need to reflect in their forecast models.Improved first line treatment resulting in less focus on next line drug development.Earlier diagnosisImproved outcomesLower progression ratesImportance of 1L treatmentsIncrease target population

26、Impact ModelKEY WORDKEY WORDThere are many variations of passages of lorem ipsum available but the majority slightly believableKEY WORDKEY WORDThere are many variations of passages of lorem ipsum available but the majority slightly believableBiomarker testing and positive identification rate has gon

27、e from 55%to 85%over the course of 5 years.AssumptionsProgression to metastatic(Stage III to Stage IV)has reduced from 80%to 20%.Challenges of Adopting AIQuality and quantity of data.As for any machine learning model to work efficiently,a comprehensive training data set is critical.Data Challenges W

28、ith such huge spend,there needs to be some demonstrable revenue on this investment that overcomes the hype around AI.Business Value Considering how new the concept of AI is,finding people with the necessary knowledge and skills is a considerable challenge.Skills Challenges The industry typically rel

29、ies on transparent and user-input type modelling when conducting forecasting exercises.AI/ML models may reduce buy-in.Transparency and accepting quantitative results AIData ConsiderationsImportant to embed privacy considerations-including anonymising data,minimizing data collection and applying data

30、 protection measuresNeed to prioritise transparency and user consent to ensure individuals understand the data collection and processing activities associated with AI systemsMonitoring and compliance to ensure organizations adapt to evolving privacy requirements and address any potential privacy ris

31、ks that may arise from the use of AIPublicly available third-party datasetsNot copyrighted.Purchased third-party datasetsPublicly available third-party datasetsCopyrighted.What about using third party datasets to feed AI systems?https:/ico.org.uk/media/for-organisations/documents/4022261/how-to-use-

32、ai-and-personal-data.pdfWhat about data privacy/patient consent?Note:several IP cases being reviewed by the Courts which will help to shape future direction.AI in Pharma Forecasting Challenges and Opportunities.Practical uses of AI in ForecastingThere are a lot of unknowns:What could future applicat

33、ions look like and whats happening now?“Everyone is telling me about what they can do but nobody is showing me anything.Business Need:Peak Share Prediction for New Product Launch.Peak Share prediction in the pharmaceutical markets is generally based on three main drivers that are mutually exclusive

34、and exhaustive:Accessing the appropriate data sets and collecting relevant dataDataLack of skilled resources/time constraintsResourceAnalysing the data in an appropriate and efficient manner,which includes removing any redundant attributes OR not including unconsidered attributesAnalysisValidate bas

35、ed on any historical dataValidationCompeting clinical profilesE.g.Efficacy,Safety,Dosage flexibilityMarket structureE.g.Order of Entry,Pricing competitiveness,Generic/Biosimilars vs.BrandCompany ProfileE.g.Marketing strength,Therapy reputation ChallengesAI within Forecasting:Peak Share Prediction fo

36、r New Product LaunchesReimbursement statusIndicationGeographic areaTarget patientsTherapeutic characteristicsPlus othersDefine New Product CharacteristicsBased on a varying sources/databases AI will do the work of the user to precisely identify the most appropriate candidatesClinical Profile DataMar

37、ket StructureDataCompany ProfileDataOutputCommercial Info:Drug development info,articles(ex-SCRIP,IN VIVO),regulatory,MedTech,etcClinical Info:services that provide information on clinical trials(e.g.,Trialtrove,Sitetrove,Pharmaprojects)Other sources(e.g.Synthetic Data)Business Need:Market Research

38、for Pharma IndustryResourceTime constraints as market research usually requires a substantial amount of time to design,execute and analyse dataBudget costsCapabilityLack of skilled resources able to combine market research and pharmaceutical forecastingRelevanceDesigning market research appropriate

39、for pharma forecasting modelsEstimating preference shares for new products Estimating preference shares for new products when entering the marketwhen entering the marketAnalysing key metrics such as unmet Analysing key metrics such as unmet need and key driver analysis need and key driver analysis S

40、upporting secondary data Supporting secondary data validationvalidationHow market research is applied within pharma forecastingValidationAbility to update/track resultsChallengesAI within Forecasting:Market Research for Pharma IndustryDefine New Product ProfileIndicationGeographic areaTarget patient

41、sEtc.Clinical Profile DataMarket Research Profiles/StructureQuestionnaireKey competitorsStandardised clinical attributes and levelsDefine StudyMkt Research data/textNLP Design Questionnaire Run StudyAnalyse/Embed into FC Model/OutputsCollect and validate dataBusiness Need:Effective Forecasting Proce

42、sses.TimeInformation is usually required to be updated and re-assessed every forecasting cycleCapabilityInformation is disseminated based on the forecasters skill and experience,which may vary considerably DataCollecting business intelligence and analysing the information may be time consuming and a

43、nalysed in a non-systematic mannerVariabilityTime and skill levels may impact the ability to effectively provide all information requiredReportingEffectively reporting forecast outputs with key drivers in a transparent way is time consuming and may require ad hoc changesKey pain points during foreca

44、sting cycles is access to actionable support material when inputting into forecast models,and the ability to visually present the outcomes to senior management in a timely and effective manner.AI within Forecasting:ProcessesData source:RegulatoryData source:ClinicalData source:FinancialData source:M

45、arketingData source:Primary researchData source:Secondary dataAI as a Forecast Support AgentAI in Pharma Forecasting Challenges and OpportunitiesAI Revolution What to ExpectThe AI Revolution what to expect.When bringing this advanced technology to your business:Incorporate AI into your business stra

46、tegy instead of treating it as a side project Build strong AI skills inhouse and/or partner with dedicated tech vendors Remember the technology is still evolving,so remain open to new learning and possibilitiesMore accurate and data driven forecastsFaster and more efficient forecastingImproved decision makingMore personalised forecastingThank https:/