Introduction
4
SME market overview
7
1.1
Company numbers and types
8
1.2
Company staff numbers
9
1.3
The Service and Supply sector
9
1.4
The Core sector
11
Asset owner snapshot
13
2.1
14
Snapshot of five major asset areas
The SME point of view
17
3.1
Sector recognition of MDC
18
3.2
Broader government support for the sector
19
AI in drug discovery
4.4
Issues with valuation of AI
26
4.5
The UK’s position and SME advice for MDC and broader government
26
Complex cell models
29
5.1
CCMs user priorities
32
5.2
Pricing & regulators drive adoption
33
5.3
Innovation in CCMs
34
5.4
UK opportunities and calls for MDC action
34
5.5
Broader government recommendations
36
Acknowledgements
36
Appendix - Market review methodologies
37
7.1
Methodology to create the company database
38
7.2
Methodology to estimate the number of employees
39
7.3
Survey details
39
21
4.1
AI user needs
22
4.2
Automation and the value of well-prepared data
4.3
SMEs accessing pharma companies’ data 25
24
INTRODUCTION
State of the Discovery Nation 2019
This second State of the Discovery Nation report provides key insights from Medicines Discovery Catapult’s (MDC) market research, surveys, interviews, and experience from the drug discovery community.
MDC has two main purposes: 1) Identifying, industrialising and driving the adoption of technologies and methods that will improve productivity and predictability of medicines discovery. Two areas of Research and Development (R&D) selected to industrialise new tools are Humanised preclinical models with combinations of advanced bioanalytical tools and Artificial Intelligence (AI)/Machine Learning systems. 2) Small or medium-sized enterprises (SMEs) and translational academics require efficient access to UK infrastructure, both public and private, to support their R&D. MDC has therefore developed 3 platforms: • Virtual R&D: Brokered access to outsourced expertise and experimental support • Samples & Data: Brokered access to consented patient samples and clinical data sets • Syndicates: Managed consortia of medical research charities and other providers to affect portfolios of patient-driven discovery projects
Discovery Science & Technology
Informatics
BETTER DATA FASTER Figure 1
Samples & Data
Virtual R&D
Syndicates
ACCESS TO UK RESOURCES & EXPERTISE
State of the Discovery Nation 2019
Catapult interventions
The first State of the Discovery Nation report1 reviewed the key issues with medicines discovery productivity and identified key strengths in the UK’s science and talent base, as well as challenges in translation and commercialisation. This second report — based on new survey and interview data — focusses on two scientific areas of MDC focus: • AI for drug discovery • Complex (preclinical) cell models (CCMs) It also reviews the size and shape of the sector and captures the most recent views expressed by the community.
KEY INSIGHTS • 80% of SMEs in the community are Service and Supply companies, accounting for 90% of employment • 60% of companies have fewer than 5 staff; 80% have fewer than 20 staff • 70% of drug assets are in cancer, anti-infectives, or central nervous system (CNS) • Cancer is the strongest therapeutic area for UK companies • Companies working in anti-infectives face a challenging market • AI and cell and gene therapies were deemed the hottest areas of 2018 • Generally companies are calling for the government to prioritise direct financial support rather than infrastructure support • 90% of companies need AI; currently 75% of AI spend is on data access and curation • CCMs show much promise to reduce animal usage but need validation before large-scale use
This SME market review is intended to deepen the understanding of this vital sector, as well as adding insight on how MDC can continue to best support the community. MDC will continue to conduct research to listen to the sector and to shape its strategy in response to evidence.
https://md.catapult.org.uk/resources/report-state-of-the-discovery-nation-2018/ ‘Service’ is used in this document to mean companies delivering services for companies developing medical assets, such as Advisory (e.g., Market Analysis/Information Consultants/Communications/Specialist Consultants with the Regulatory Expertise [organisations]); Outsourcing (e.g., Clinical Research Organisation + Contract Manufacturing / Research Organisations); and Materials Supply companies 3 ‘Core’ is used in this document to mean companies actively researching and developing their own human therapeutic products 1 2
SME MARKET OVERVIEW KEY MESSAGES • There are 1,500 UK medicines discovery SME companies split 80% Service2 and Supply; 20% Core3 companies • UK medicines discovery SME companies employ 21,000 people, with 8,000 in research-outsourcing companies • Advisory companies are the most common; 80% of these employ fewer than 5 people • Core companies employ only 2,500 staff; fewer than 5% have over 50 staff • Over half of sector staff are in larger service companies with over 50 staff • Therapies other than small molecules now represent over half of all Core companies • 70% of assets are in oncology, anti-infectives, or CNS
State of the Discovery Nation 2019
SME market overview
The UK science base should support 50 more biotech companies at early clinical stage
300 Core companies, 1,200 Service and Supply
1600 Number of companies
1400 1200 1000 800
Service and Supply Other Materials Supply Outsource
Service and Supply sector
600 400 200
Figure 2
0
Advisory Core
https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/707072/strength-and-opportunity-2017-bioscience-technology.pdf Estimates have been rounded to reflect the volatility in company numbers 6 A vision for the UK life science sector in 2025, BIA 4 5
SME market overview
State of the Discovery Nation 2019
SME employment is mainly outsourcing and materials supply 100 90
Service and Supply Other
80 Percentage %
70 60
Materials Supply
50 40 30 20 10 0
• Advisory — 33% • Outsourcing — 29% • Material Supply — 22%
Outsource Advisory Core
SME market overview
State of the Discovery Nation 2019
Total employment: Core vs Service and Supply 19,000 18,000 17,000
15,000
80% 8
16,000
14,000
Company staff numbers range: Core vs Service and Supply
12,000 11,000 10,000
100
9,000
90
Percentage of employees
Number of employees
13,000
8,000 7,000 6,000 5,000 4,000 3,000 2,000
Figure 4
20-49
80
10-19
70
5-9
60 50 40 30 20
1,000
10
0,000
0
Service Core and Supply
100-249 50-99
0-4
Core
Service and Supply
SME market overview
State of the Discovery Nation 2019
Small molecules are no longer the majority therapeutic modality in UK SMEs medicines discovery
Increasing diversity in therapeutic modalities
100
Percentage therapeutic modalities
90 80
Cell and gene
70
Vaccines Therapeutic proteins
60
Antibodies
50 40 30 20 10 0
Small molecules
MDC Core
Core plus CGT
Figure 5
The majority of SMEs in the Core sector employ 4 or fewer people
7
NB, this number is taken directly from the OLS data and not reviewed further as these companies are in MDC’s scope
SME market overview
State of the Discovery Nation 2019
Unlike the Service and Supply Sector, there are very few Core companies with over 20 staff, due to the tendency for smaller UK companies to be acquired after key milestones are met. This would suggest that Core companies focus on scientific project progress rather than organisational growth, whereas Service and Supply companies are more likely to need additional people to deliver more services to maintain long-term productivity. It also reinforces BIA and MDC’s concerns that the UK market for growth capital is not allowing UK SMEs to build in size locally. Cancer, Anti-infectives and CNS make 70% of the assets in development Further analysis using information from Global Data suggests that UK SMEs are currently developing over 930 drug assets. The most common therapy area was Oncology (349 drugs), followed by Infectious Diseases (182) and then CNS (115) (Figure 6). This is different from the global large pharma company late-stage assets mix identified by Deloitte8, which also has oncology as the largest group but has fewer infectious disease and more respiratory and immunology assets.
21% Other
20% Infectious Disease
12%
Central Nervous System
38% Oncology
5%
Immunology
4%
Gastrointestinal
Core Biopharma - Drugs by Therapy Area Figure 6
8
Source: Global Data
Unlocking R&D productivity. Measuring the return from pharmaceutical innovation 2018. Deloitte Centre for Health Solutions, 2018
ASSET OWNER SNAPSHOT KEY MESSAGES • Cancer cluster is strong and supported with world leading science, funding, and coordination • Anti-infectives cluster is large but faces commercialisation challenges • CNS is heterogenous, with Dementia showing particular potential • Respiratory and musculoskeletal disease have relatively few preclinical companies
Asset owner snapshot
State of the Discovery Nation 2019
Snapshot of five major asset areas Deeper analysis of Core SMEs was performed across five prioritised therapy areas: Oncology, including Immuno-Oncology; Nearly half of the SMEs Anti-infectives; CNS, including Alzheimer’s Disease and Pain; in this subset are Musculoskeletal disorders; and Respiratory9. More detailed focussed on oncology analysis of a subset of these SMEs, which were confirmed as active across one of the five therapy areas, are defined here as ‘Asset Owners’. Nearly half of the SMEs are focussed on oncology, with an equal balance between preclinical companies and those with assets in the clinic (Figure 7). A strong understanding of the biology of cancer, and the ability to connect clinical unmet need with basic science, provides a solid basis for translation, company formation, and growth. This is in line with evidence from the State of the Discovery Nation 2018 report10, where 100 UK experts rated the strength of UK basic science and translational medicine in oncology top of the 12 therapy areas examined. Clinical Research-only
55%
Oncology 48%
45% 54%
Anti-infectives 26%
46% 61%
CNS 39%
14%
33%
Respiratory 7%
67% 20%
Musculoskeletal 5%
Percentage % of R&D SMEs across the 5 therapy areas5
80%
Split between Clinical and Research-only SMEs
Figure 7
In addition to searching publicly available company information, two main data sources were used: UK Biotech Company List2, and BioPharm Insight Database3. A standard set of search criteria was used, including a limit of 10 drugs in the portfolio to ensure we focussed on SME community of companies. Search criteria included: i) key words for the Therapy Area, ii) UK Head Quarters, and iii) ≤10 drugs in the portfolio across all Phase of Discovery and Development. Companies with no news or external communications in the last 3 years were excluded, unless there was clear evidence from other sources that the company was still solvent and operating. Technology platform companies with projects in >1 Therapy Areas, are included in each of the Therapy Area numbers. http://biopharmguy.com/links/country-united-kingdom-all-location.php, BioPharm Insight (www.biopharminsight.com)
9
Asset owner snapshot
State of the Discovery Nation 2019
Cancer has an accessible network of experts that helps small companies Important drivers for this strength are the presence of a cohesive research community, led by Cancer Research UK (CRUK), and the commercial opportunity for new oncology drugs. Large charities like CRUK increase the strategic coordination of research, and via initiatives such as the National Cancer Research Institute, attempt to make research and clinical experts more accessible. CRUK’s public donations increase the funds available for oncology research and early development in academia and in SMEs. The continuing commercial success of oncology companies and drugs through late-stage development and reimbursement attracts venture capital and pharma investment in this therapy area. As a result, the UK oncology SME sector is amongst the best funded and performing in the UK, despite the ongoing challenges faced, as illustrated by the recent £65m Series B funding of Artios Pharma. The lack of commercial pull in anti-infectives means support is needed all the way to adoption Although there are significant number of anti-infective companies evenly spread across the preclinical and clinical phases, the growth drivers for the anti-infectives UK SME sector are very different. The preclinical research part of this sector is growing. This is driven by government-led public awareness of the threat posed by antibiotic resistant microbes, organisations such as the Antimicrobial Resistance (AMR) Centre, and government and charity funding (e.g., National Institute for Health Research, UK Research and Innovation, and the Wellcome Trust). This funding is complementary to the academic strength in basic anti-infective science identified in the State of the Discovery Nation 2018 report10. In contrast to oncology drugs, uncertain pricing and the adoption of new antiinfective treatments as a ‘last resort’ means they are perceived to have a low market potential. This makes securing venture capital funding challenging. Licensing and partnerships are rare as many major pharma companies have stopped late-stage development and commercialisation of anti-infectives. The State of the Discovery Nation 2018 report10 reflected this situation, as the anti-infectives area was rated weak in translational medicine, suggesting that anti-infectives SME landscape could benefit from translational support. However, the lack of commercial pull means that translational support is needed not just across the ‘valley of death’ but also all the way to adoption, or there is a risk of molecules and companies stalling at later phases. In this context, the UK government’s commitment to “develop and test new models for national purchasing arrangements that de-link the price paid for antimicrobials from the volumes sold”11 could make a real impact.
https://md.catapult.org.uk/resources/report-state-of-the-discovery-nation-2018/ https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/773130/uk-amr-5-year-national-action-plan.pdf
10 11
State of the Discovery Nation 2019
Asset owner snapshot
CNS may be an area where UK SMEs can grow in strength Drawing conclusions from the relatively small number of CNS SMEs is complicated by the very diverse nature of this therapy area. It includes Pain, Psychiatry, Neurodegenerative diseases and Dementia, with companies often working on more than one indication. However, there is commercial and public interest for new products in these diseases. Dementia stands out most prominently (nearly 40% of these companies) and, with growing charity and academic sector coordination (e.g., the Dementia Discovery Fund, Alzheimer’s Research UK, and UK Dementia Research Institute), may be an area where UK SMEs can grow in strength. UK’s respiratory musculoskeletal areas have fewer preclinical companies The Respiratory and Musculoskeletal SME sectors pose a different challenge, as few SMEs were identified in these sectors, one-third of which are preclinical. There is a risk that the UK is losing the critical mass of translation experience in these two therapy areas.The small community of clinical stage SMEs in these areas could be difficult to sustain over time due to a lack of local preclinical companies producing assets for clinical research. Unlike antiinfectives, major pharma companies are still investing in both There is a risk that the respiratory and musculoskeletal, creating commercial pull for UK is losing the critical products. However, based on this snapshot, the UK’s Respiratory mass of translation and Musculoskeletal companies may need support to meet the experience in these two global demand.
therapy areas
In summary, the UK R&D SME landscape across five therapy areas indicates substantial differences in numbers of early- and latestage SMEs across the therapy areas. Oncology is the only area with a global scale cluster, with potential in CNS, scale issues appearing in preclinical musculoskeletal and respiratory, and market issues in anti-infectives. A predictable market attracts investors and provides the much-needed pull for SMEs to translate vital new medicines. As seen in oncology, it appears that a strong coordinating organisation or a government-driven initiative, which brings together scientific and clinical expertise and is able to provide funding, can provide essential network access and strategy development benefits to smaller companies. CNS, musculoskeletal and respiratory combine commercial pull and an active community of experts and companies that could benefit from support.
THE SME POINT OF VIEW KEY MESSAGES • AI and cell and gene therapies are deemed the hottest areas • Mixed overall sentiment, with significant Brexit concerns • Data science is needed across all areas • CCMs are needed, especially in oncology and CNS • About two-thirds awareness and two-thirds interest in MDC’s themes • SMEs do not recommend the current infrastructure support in general • MDC’s ‘Net Promoter Score’ is increasing — the SME community is increasingly likely to recommend working with MDC • MDC should continue to do more collaboration building and enable access to ‘science industrialisation’ services • 75% recommend government support around funding and tax incentives
State of the Discovery Nation 2019
The SME point of view
An online survey and in-depth interviews were conducted to understand the opportunities and challenges felt by companies in the sector, and to determine what support they were asking for. Half of respondents to our online survey12 felt that AI and cell and gene therapies were the areas of discovery that had increased most in value in 2018. There was less agreement on the innovation that has decreased the most in value, with a quarter identifying small molecules or high-throughput screening. The sector’s feelings about the future were mixed, with 80% viewing Brexit negatively, and recruitment issues being the most commonly expected consequence. Data science is needed across all areas, and CCMs were particularly sought after in oncology and CNS disorders. There is increasing recognition for MDC: 69% of companies were interested in MDC’s themes, and companies are increasingly willing to recommend working with MDC to their peers. SMEs asked for a range of support from MDC including access to high-risk science and collaboration building. SMEs did not recommend most of the current UK non-financial support mechanisms, with three quarters recommending government support via grant funding and tax incentives.
Sector recognition of MDC Most respondents were aware of the range of support that MDC can offer and felt that MDC areas of expertise were of interest to them. The community is increasingly positive about MDC: its ‘Net Promoter Score’ was the most improved of 13 business support structures13, and it is now third highest ranked, behind Innovate UK’s grants and the Medicines and Healthcare Products Regulatory Agency (MHRA). The major requests of MDC (Figure 8) were: • Collaboration: connecting and linking collaborators, and brokering introductions between larger and smaller players • Access to science: providing niche technologies, and access to datasets and screening libraries • New models of engagement: engaging earlier with companies and being able to invest in joint projects
In 2018 the MDC and BIA repeated its annual online survey of the community to understand the opportunities and challenges encountered by companies in the sector, and what support was being requested. Academic Health Science Centres, University technology transfer offices, University contracting department, Genomics England, Academic Health Science Networks, The Knowledge Transfer Network, NICE, The Cell & Gene Therapy Catapult, NOCRI and the CRNs for clinical trials, National Biologics Manufacturing Centre, MDC, Innovate UK’s competitive grants, MHRA
12
13
The SME point of view
Other
State of the Discovery Nation 2019
New Engagement Access to Models Funding
Collaboration
Publicity/ Awareness Access to Science
MDC was asked to support collaboration and access to science Figure 8
A small number commented that MDC could potentially compete with the supply companies that it is here to support. MDC’s goal is to provide a unique combination of capabilities that the community cannot otherwise readily access. MDC has multiple collaborations with supply companies to co-develop new products for them to adopt or access, thereby supporting the continued success of UK CROs.
Broader government support for the sector 75% of recommendations were for government stimulus funding, covering increased grant funding, tax incentives, and more efficient funding mechanisms. The next largest theme (‘making connections across the sector’) was only raised by 6%, and no other area got more than one mention. The sector noted concern over the move from ringfenced Health and Life Sciences to ‘Smart’ Innovate UK competitions, which has decreased the opportunity for SMEs to access medicines discovery funding.
State of the Discovery Nation 2019
The SME point of view
A big pharma view: fund SMEs and their substrate too When the themes of this report were discussed with large pharma drug discovery developers, one responded with insights into the large company issues around drug development, which is given here to complement the priorities of smaller companies.
“…The field still does not understand the fundamental biology of many diseases well enough to validate drug targets: this will come from increased understanding of disease at the cellular level, both mechanistically and longitudinally and across populations. It will require access to well curated patient cohorts with samples, omics technologies and data management. Cancer has demonstrated the progress that can be made when the cause of a disease (genetic mutations) is identified and becomes affordable to study at massive scale. How can the research community replicate this in cardiac disease or CNS disease by mapping the signatures of disease at a cellular level? The infrastructure and technologies that drive this understanding may not be in small companies and are often better suited to precompetitive, charitable, or government funding, but they are the bedrock that allow innovation and SMEs to start up. These datasets are the input for AI and set the standard for cell models and should not be neglected as a long-term investment…”
AI IN DRUG DISCOVERY KEY MESSAGES • AI is now a core part of drug discovery across all areas • AI is most valuable when applied to a portfolio alongside domain expertise • The industry wants benchmarks and comparisons between AI systems • Small companies prefer to sell assets; large companies prefer partnerships • AI budgets are growing, and pharma companies are moving AI in-house • Automated data generation or validation is highly valuable • The right data are key: acquisition and preparation of data take 75% of resources • Pharma data are valuable but difficult for SMEs to access • MDC can help by making high-value datasets available and linking the sector
State of the Discovery Nation 2019
AI in drug discovery
UK AI in the drug discovery sector is in a good position. Increasing end-user (pharma/biotech) acceptance of the value of AI has The field is still enabled investment in people, data, and algorithm development, being held back by and has led to financial and organisational successes. 90% the availability of of UK SMEs are using or need data science in drug discovery. well-curated, largeHowever, there is a trend towards in-sourcing and partnerships scale data, and models that smaller AI and biotech companies will need to navigate. The field is still being held back by the availability of well-curated, to make pharma large-scale data, and models to make pharma data available data available on acceptable terms; it is a challenge, but possible. Ways to benchmark the performance of different AI approaches would be welcomed. Small companies are being encouraged to look into pharma’s ‘pain points’ of data curation and document/literature mining, natural language processing, as well as building less specialised interfaces to their products to help a broader group of users. MDC can help both by making datasets available and collaborating to validate SME approaches, and organisationally by increasing customer and producer awareness.
AI user needs AI is now a core part of SME and large-company drug discovery. While computational chemistry has been using data science and machine learning techniques for years, increased confidence in the field, increased input data quantities, and increased availability of tools and computing resource mean that AI approaches are being used in all areas of modern drug discovery. 90% of UK SMEs needed data science, with half of these specifically needing AI and machine learning. Lead generation and optimisation were the most frequently cited area of data science need followed by Target ID to clinical research. Within the techniques underlying data science, data mining (including AI and machine learning) was most cited, more so than data acquisition, filtering, curation and information extraction from unstructured sources (Figure 9).
AI approaches are being used in all areas of modern drug discovery
90% of UK SMEs needed data science
AI in drug discovery
State of the Discovery Nation 2019
Figure 9
This relative de-prioritisation of data acquisition and integration does not match with the many requests for improved data access, which may indicate that data access is a governance and operational issue for SMEs, not a lack of data science. AI is now seen as a core part of industrial drug discovery, but measuring impact and validation are seen as highly complex issues There is a need to improve decision making across target identification and validation, drug discovery and design including absorption, distribution, metabolism and excretion (ADME), efficacy, toxicity, and process improvement. AI is also used for clinical trial optimisation, but this is out of scope for this report. 50% of respondents observed that we do not yet have enough data to prove the later clinical benefit of using AI in the early stages of drug development, due to the time delay and potential confounding factors. Analyses, such as AstraZeneca’s 5R paper14, will start to inform such debate as benchmarking data accumulates. In the interim, validation of AI methodologies (through test datasets, pilot projects, partnerships, expert use, and relevant open competitions) is highly important.
14
Morgan P et al. Impact of a five-dimensional framework on R&D productivity at AstraZeneca. Nature Reviews Drug Discovery 2018; 17:167–181
AI in drug discovery
State of the Discovery Nation 2019
In the context of uncertain validation, differing views on business models are held. 40% of smaller companies believe that they could create and identify their own insilico assets and commercialise these to pharma. 50% of larger companies prefer a business model of partnership and close integration or consulting on in-house programmes. Pharma companies preferred to use external SMEs as a technical resource when a technology was unavailable in-house. SMEs were more interested in selling a high-value product e.g., chemical equity to pharma companies. A continuation of the current mix of fee-for-service and risk-sharing work is expected by all. However, pharma customers and leaders increasingly see AI as a core competency. They are also seeing higher budgets and ease of internal approval for AI programmes, meaning a higher proportion of skills and work brought in-house is likely. Validation of AI methodologies is highly important The community is sceptical of black-box algorithms but shows increased interest in models that can inform either the model’s own accuracy or the underlying biology that is being represented. Users have more confidence in AI models that give outputs that are interpretable by the scientist, as opposed to outputting a single score. There are real opportunities for companies to create simpler There are real user interfaces and functionality that is accessible to bench opportunities for scientists without needing an informatician to interpret or guide.
Automation and the value of well-prepared data
companies to create user interfaces and functionality that is accessible to bench scientists
Automated synthesis and analysis of molecules designed in-silico could enable both rapid validation of the performance of the assets and rapid feedback to improve the algorithms. However, the integration of AI design and automated synthesis in production mode is seen as years away, with current systems at proof-of-concept stage. Any algorithm must be able to show excellent reproducibility across a range of datasets and at large scale as the pharma discovery process is already highly automated and optimised.
AI in drug discovery
State of the Discovery Nation 2019
Data availability is a key enabler of AI in drug discovery. While there are some AI approaches which make the most of sparse Data curation and or incomplete data, all data science approaches benefit from preparation accounts for increased amounts of relevant, well-curated and linked data. 75% of the budget for Data curation and preparation accounts for 75% of the budget data science for data science, with only a small proportion taken in the development and use of the algorithms. There were particular requests for data associated with ADME and toxicity, synthesis prediction, compound activity, and linked genetic-phenotypic datasets. Perceptions of data value have increased over the last five years, and so achieving value for money in data generation, acquisition or sharing is essential.
SMEs accessing pharma companies’ data 83% of pharma companies felt that the perceived value of their in-house datasets had increased, and for most this was a barrier to sharing their data. IMI15 initiatives have been set up enabling pharma companies to share data, but most UK SMEs are not part of these initiatives and do not have their own in-house databases to contribute. There was a willingness in pharma companies to try new models of sharing data that do not involve giving away all of the value or confidential assets within the data. This approach could allow algorithms to be trained on data without the data being exposed to the algorithm developer, for example. This approach was considered to be difficult to implement, but possible, and represented a chance for SMEs to access some of the value in pharma datasets on reasonable terms for both parties. In particular, the sector needs methods of sharing pharma company bioactivity data that do not expose precise biological targets (either explicitly or implicitly) or compound structures, while still retaining enough information to inform machine learning.
15
Innovative Medicines Initiative, https://www.imi.europa.eu/
State of the Discovery Nation 2019
Issues with valuation of AI It is difficult to assess the value of an AI approach. Ideally, there is a validation set of data, unseen by the algorithm developer, which the customer can run the algorithm on to assess performance against the standard. While there are open competitions built around this model, most respondents felt that this was not covering the need in drug discovery. This may be due to the highly specialised areas of prediction that AI is now reaching, or the lack of validation data. This is a market failing, preventing customers from independently verifying the performance of their products. It is difficult to assess the value of an AI approach Some respondents were sceptical about the value of algorithms alone, noting that the publication and sharing of many AI techniques, in particular deep learning toolkits, mean that the tool is rarely differentiated. However, others were convinced that valuable innovation in algorithms continues. Areas where data science has already seen extensive use, such as molecular modelling, are less innovative for data science and more about data access and appropriate use of the approaches. The less wellcovered fields of data extraction and improvement, and natural language processing, were seen as more innovative and highly valuable in curating and enabling the use of public and private scientific and clinical literature. There is no consensus on certain approaches falling in or out of fashion, although there was increased use of ensemble or multi-algorithmic approaches. The field remains fluid, with different algorithms being used successfully, but with difficulties in comparing results. In addition, AI that helps the user understand the basis or accuracy of its predictions is more valuable, both to the informatic and bench science community.
The UK’s position and SME advice for MDC and broader government The UK was seen positively in the international context, with a strong mix of data science and drug discovery expertise enabling small and growing AI in drug-discovery companies to compete internationally, despite lower capital availability than in the US.
AI in drug discovery
AI in drug discovery
State of the Discovery Nation 2019
The success of BenevolentAI, Exscientia and Genomics plc, amongst others, demonstrate that AI in drug-discovery companies can grow and thrive in the UK. AI companies feel more hindered by the lack of available talent and affordable data than by a lack of investment capital. Despite the difficulty for investors in assessing a combination of tech and biotech business fundamentals, this is a ‘hot’ sector, and the field is considered to be the area of medicines discovery that has most increased in value over the last year.
AI in drug-discovery companies can grow and thrive in the UK
MDC was asked to work on collecting and making available datasets around ADME, toxicity, and linked genetic and phenotypic data The survey demonstrated that MDC is well known and supported by the AI Drug Discovery community. Collecting and making available datasets around ADME, toxicity, and linked genetic and phenotypic data was identified as a focus area for MDC. The fast changes in the demand and supply for AI in drug discovery led to requests for MDC to organise events and networking to support the community, improve mutual understanding between producers and customers, and matchmake deals nationally and internationally. MDC’s role in advising companies on formation and early stage was valued, and collaborations with SMEs have brought additional funds and resources to the sector.
18%
Training data scientists
12%
Funding Translations
42%
Making data available
12%
Linking Companies
10%
Data Standards
6% Other
Making data available is the most valued government support for UK data science Figure 10 — recommendations for support from broader government
State of the Discovery Nation 2019
Over 40% of respondents want broader government support in increasing data access to support the adoption of AI for drug discovery (Figure 10). Industry also called for ‘industrial’ training of data scientists and focussed funding for the translation of new academic approaches, suggesting that many of the nation’s most impactful interventions only require limited public sector investment on priority areas. This activity, to enable AI in drug discovery, is well suited to be delivered via MDC as a trusted central provider that can leverage both private and public investment.
AI in drug discovery
COMPLEX CELL MODELS KEY MESSAGES • CCMs are a valuable preclinical research enhancement, if used correctly • Improving reproducibility and developing validation data are current priorities • Adoption is small-scale and experimental, with regulation the enabler for growth • UK SMEs have potential, reputation and skills to use CCMs well • Need to validate models and support pre-sales, to bring together the community and advise • The field is broad, so MDC should prioritise high-risk, high-value areas
State of the Discovery Nation 2019
CCMs cover a broad field aiming to improve on the predictability of current twodimensional (2D) and non-human-derived cell models, while managing increased assay difficulty and cost. CCMs are technological approaches to improve the predictions from cell models, for example by organising the cells into three-dimensional (3D) structures, mimicking differentiated human tissues, approximating fluid composition and flow, and linking different tissue types into a model system. The use of CCMs is being driven by a combination of decreased trust in the translational value of animal models and increased availability of data to support the validity of complex human cell models. Landmark publications from the Institute of Cancer Research and the Hubrecht Institute16 have linked the predictions from CCMs with clinical response in patients. There is an increasing awareness of the issues with 2D models, such as cells on flat surfaces being less physiologically relevant or single populations of cells lacking cell-cell interactions. In this context, value comes from appropriate use and understanding the challenges and limitations of CCMs.
Areas of complex cell model needs 50 45 40
Percentage %
35 30 25 20 15 10 5 0
Areas of CCM need
Figure 11
16
http://science.sciencemag.org/content/359/6378/920 and https://www.ncbi.nlm.nih.gov/pubmed/29224780
Complex cell models
Complex cell models
State of the Discovery Nation 2019
CCMs are needed by the vast majority (84%) of companies for target identification and validation followed by drug efficacy, biomarker validation, and toxicity (Figure 11). There was less need for CCMs in functional and pathway imaging, and for multi-organ or organ-on-a-chip (OOAC) approaches. Interviews identified that OOAC systems require further validation and industrialisation before they can be adopted into widespread use. Experts producing, commercialising and using CCMs emphasised that the use of these approaches must balance the increased complexity and cost of setting-up, maintaining and interrogating the CCM with the increased accuracy of predicting drug efficacy and safety. Drug discovery scientists need a model that is validated, reproducible, and easy to use, despite the increased complexity. Excellent data skills are needed to manage and interpret the increased and novel outputs from new models.
CCMs are needed by the vast majority (84%) of companies for target identification and validation
Excellent data skills are needed to manage and interpret the increased and novel outputs from new models
Users need a model that is validated, reproducible, and easy to use
CNS
Oncology
Other Disease Non Disease Specific*
Oncology and CNS are seen to be the areas where CCMs could have most impact Figure 12
*Non disease specific (e.g. Safety/Toxicology, Immunology)
State of the Discovery Nation 2019
Complex cell models
Early areas of commercial interest include toxicity, oncology, immunology, and neurology. CCMs were considered to be most impactful in oncology and CNS disorders (Figure 12). This mirrors the current technological advancement and commercial opportunity for cancer models, and the strengths of CNS-humanised models versus their animal comparators for complex biological diseases. Current usage is small-scale and experimental, with regulatory support expected to be the major growth driver. Broad innovation continues but customers value usability over novelty, so industrialisation is a key hurdle to overcome. The UK has a good reputation in the field, but companies have received less support than the US. MDC was asked to support companies in demonstrating and validating models in selected areas with high commercial interest that are too risky for company funding, and to link-up the community of users and suppliers.
CCMs user priorities Identified core needs include: predictivity, reproducibility, validation, and ease of interpretation. The driving need is improved predictivity across the stages of drug discovery, or between the CCM and clinical performance. Quantitative in-vitro to invivo extrapolation should be improved by the use of complex human models. This should enable segmentation and ranking of molecules, and improved understanding of therapeutic margins and in vivo potency, as well as adding data to increase the understanding of the disease. Results should be consistent across multiple human donors, consistent over time and reflect the mature human phenotype. The core needs of users were: predictivity, reproducibility, validation, and ease of interpretation Although there is no definitive publication or data that ‘proves’ a model, a reasonable number of widely accepted reference compounds should have been run on multiple occasions with consistent results across molecular and functional endpoints. The current gold standard of validation is checking if the drug that was active or toxic in the clinical setting matches with a signal in the CCM. Ideally, the model’s expression profile should match clinical samples, or the CCM should give the correct prediction in situations when the 2D model does not correlate with the clinical result, but this is not commonly done. The validation standard A wider initiative is used by CROs is understandably ‘good enough’ rather than ‘beyond needed to increase the all doubt’, so a wider initiative is needed to increase the validity of validity of CCMs CCMs and to compare across competing models. The costs and associated risks are too high to allow CROs to develop new models from scratch.
Complex cell models
State of the Discovery Nation 2019
3D models can give more information, leading to increased data size and complexity, as well as new data points that do not have equivalence in prior assays. This requires links to advanced analytical approaches for interrogation of models that can measure conventional markers in 3D. Skilled analytical, data and biological scientists are, therefore, needed to interpret novel findings, even when the models achieve the predictivity and reproducibility mentioned before. Skilled data scientists and biological scientists are needed to interpret novel findings CCMs are currently better proven in certain areas than others. Early usage was expected in drug safety testing as it is common across all drugs, with cardiovascular and hepatic toxicity as the initial examples. Primary cells or cells derived from inducible pluripotent stem cells (iPSC) can be used as long as the model is reproducible and well-validated for a given end-point. Cancer CCMs have more competition from animal models such as xenografts, Immunology and but there has been a range of success across the tumour types, neurology CCM and particular interest in modelling the tumour environment. models are in demand Improved immunological and neurological CCM models are in demand, due to the issues in translating data from existing models to the clinic.
Pricing & regulators drive adoption The development of regulatory guidelines is a main driver of adoption and standardisation of complex assays. Within pharma, there is a trend towards the increasing adoption of CCMs, but this is more at the experimentation stage rather than large-scale use. Broader adoption will be driven by the data that are being requested by regulators, and by reduction in the cost and time needed to validate and adopt a CCM. Pharma have their own standard processes for in vitro testing, which will take evidence, time and money to change. It is accepted that CCMs will not replace animals in the next five years, but they should reduce the number of animals being used. Broader adoption will be driven by the data that are being requested by regulators Given that preclinical budgets are not increasing, CCM funding will most likely be at the cost of animal budgets. The price of the model is important and current pricing suggests they are being used in the later stages of preclinical drug discovery. However, as the models are not yet seen to be fully industrialised, the current sales by UK SMEs are smaller test projects costing tens of thousands for tens of vials17. Precise pricing is complex as the comparison of value is not yet objectively accepted, so pricing follows the current assumptions of value: more than 2D and less than in-vivo.
https://cn-bio.com/innovate-uk-awards-grant-to-cn-bio-innovations-and-astrazeneca-to-use-organs-on-chips-to-evaluate-the-use-of-existing-drugs-to-treat-fatty-liver-disease/
17
Complex cell models
State of the Discovery Nation 2019
Innovation in CCMs There is significant innovation in CCM design and supply, both in the development of new models and in the improvement in performance of current models. Much of this innovation is being carried out within universities, often in collaboration with industry partners. Cost remains important, as well as usability via improved robustness and repeatability. Vendors were felt to have neglected usability in the past, and customer feedback was leading to simplification of the culture process and reducing the time to maturity. Some respondents noted that transfer of organoids between institutions was not yet commonplace, partly through concerns that the protocols would not replicate. CCMs are currently low-to-medium throughput by nature of their complexity, shape, and fluid component, and that scaling-up may not be completely solvable, which will define the point of the process where they are used to evaluate drug candidates. Gene editing (e.g. CRISPR) is now affordable for CCMs and represents an opportunity for producing disease-like phenotypes. Innovations in 3D-extracellular matrices are being researched by CROs to give increased control and relevance. Cell differentiation and maturation remains a challenge and a focus in areas such as neurology. There was little commercial demand for complex, multi-organ systems, and also significant challenges in delivery of systems combining more than 2-3 organ models. However, OOAC technologies are continuing to mature, and the use of flow is important for a variety of organ types. Far-reaching IP ownership Innovations in matrices of protocols (e.g., stem cell differentiation and CRISPR) is one of and fluid flow can drive the major blockers to innovation being exploited commercially.
up OOAC impact
UK opportunities and calls for MDC action The UK has CCM expertise of international quality. A combination of robust scientific approaches and an experienced community in models and drug discovery, in general means that the reputation of UK CCMs is high. There are companies in the UK who could take advantage of the unmet need for CCMs, for example iPSC suppliers or CROs, and there are UK academic institutes developing novel complex models. However, other countries have invested strategically in OOAC and cell models, such as the US National Institute of Health (NIH) Tissue Chip for Drug Screening program18, and the UK will need similar support to remain at the forefront of this technology. Assure and validate models via demonstration in MDC labs
18
https://ncats.nih.gov/tissuechip
Complex cell models
State of the Discovery Nation 2019
The calls for MDC action were for independent validation and assurance of models via demonstration by assessing responses to known molecules. This should be as open as possible - as much of the current work is not published - and it should support the presales process by demonstrating third-party use. The choice of model is important. CROs believe that MDC should focus on areas that are higher-risk projects, or in partnership with companies to jointly reduce the risk. SMEs can identify niche models that currently are not well-served by the UK CRO community. The sector asked MDC to pick a couple of long-term areas in which to develop expertise and capability. Suggestions included oncology and immuno-oncology (needing first to identify specific opportunities), neuroinflammation, neurodegeneration and fibrosis. While SMEs asked if MDC could aggregate or make validation sets of compounds for reviewing models, pharma companies said they were willing to share compounds directly for CCM makers to test on their CCMs. Collaborative guidance of the standard controls by tissue type is being published from a group of 18 pharma companies, so coordination with such initiatives will be important. Most respondents noted that the pace of change within the field made it important for MDC to help create a network of suppliers and users across the country to share information and techniques quickly with smaller companies. This network would include all the related disciplines across academia, pharma and suppliers. In particular, user group meetings that brought lab staff together would be very valuable. Finally, MDC could support network members as an honest broker between model users and suppliers, as well as supplying expert advice to support and critique the experimental approaches and data from CCM companies.
Independent validation and assurance of models via demonstration by assessing reponses to known molecules
Pharma companies said they were willing to share compounds directly for CCM makers to test on their CCMs
Create a network of suppliers and users across the country to share information and techniques quickly with smaller companies
Complex cell models
State of the Discovery Nation 2019
Broader government recommendations The request for broader government support for CCMs was focussed on stimulus funding (Figure 13). Over 75% of respondents want specific funding for early-stage companies, academia and translation. The combination of high market demand and lack of funding suggests that CCMs are an area of current market failure, with a key need for validation and funding to translate excellent basic research into industrial adoption.
Academic Funding
Support Translation Access to Resources
Company Funding
Training
MDC What government support is most valued for CCMs? Figure 13
Acknowledgments Thank you for reading the State of the Discovery Nation 2019. Thanks also to the experts across the UK SME community for their contributions to it. The state of the drug discovery nation is improving and can benefit from continued support from the MDC, BIA, and broader government. We will continue to conduct research to ensure that we continue listening to the sector and shape MDC’s strategy in response to evidence. Please contact MDC if there is additional evidence that we should be considering. Our ambition is to go beyond incremental change, to transform the industry, and we look forward to working with you to achieve it.
APPENDIX — MARKET REVIEW METHODOLOGIES
State of the Discovery Nation 2019
Appendix - Market review methodologies
Methodology to create the company database The development of the database of MDC relevant SMEs was performed in four phases: developing the initial list of companies; classifying by sector and segment; supplementary data gathering; and filtering and validation. The initial list of companies was compiled using data from the OLS 2017 companies list, which was published in May 2018. This list of companies was expanded using data from commercial sources, industry bodies, and Innovate UK, coupled with internal data from the MDC staff and CRM systems. Once a long list of companies had been produced, an exercise was performed to remove duplicates and to classify new companies based on their primary sector and segment using the same descriptors as OLS. A focussed data-gathering exercise was then performed using Companies House accounts information (e.g. to determine the number of employees, and the last reported status of the company). The company list was then filtered to identify the MDC in-scope companies using the following criteria.
Main Criteria
Examples of Exclusions
Company is an SME based on the number of staff (i.e. fewer than 250 staff)
Large pharma companies
Company is involved in MDC relevant activities
Service and Supply companies involved in logistics, legal etc. Core companies involved in Advancedtherapy medicines or the manufacture of medicines
Active UK registered company (based on data from Companies House)
Dormant, in liquidation, not a going concern, non-UK registered companies.
Company has sufficient (online) data freely available to make an assessment
Newly incepted companies / in stealth mode
A validation review was then performed within the Catapult of final list of companies, and once complete, the list was locked down for a detailed analysis. In performing this study, it is recognised that this group of companies changes quickly, as commercial and research programmes succeed and fail. Therefore, the results should be treated as a snapshot of the market at a point in time.
Appendix - Market review methodologies
State of the Discovery Nation 2019
Methodology to estimate the number of employees The number of employees can be used as a proxy for the size of companies where sufficient data exists to make estimates for this. Companies do not always disclose the number of employees that they have. OLS put employment in ranges due to the proprietary nature of their source data. To develop our estimate of the number of employees across the sector, MDC gathered data from the Annual Report and Accounts for a sample of 12% of companies in order to estimate an average number of employees in each band. These data are presented below and suggest that companies tend to have staff below the midpoint of the band.
Employment Band
No. of companies with known no. of employees
Average no. of employees for that band
No of companies with an estimate no. of employees
0-4
65
2
794
5-9
36
6
181
10-19
24
13
136
20-49
24
29
117
50-99
20
69
46
100-249
15
158
33
Survey details In 2018, MDC and the BIA repeated their annual joint online survey of the community to understand the opportunities and challenges felt by companies in the sector, and what support they were asking for. The online survey contained detailed sections on two areas of particular interest: AI in drug discovery and CCMs, as well as a repeat of the 2017 questions on perspectives of government support mechanisms. The survey was sent out in collaboration with the BIA and filled in by 60 companies (representing a similar number of companies to the previous report, as last year’s survey also included academic and government responses). The survey population was representative of senior commercial leadership: a quarter were Chief Executive Officers, half were senior R&D, and the remainder mainly Chief Officers or board. One-third of respondents worked in companies active in small molecule drug discovery, and a quarter worked in CROs. In terms of disease focus, one-third were cross-disease, and one-third were spread across oncology, antiinfectives and immunology.