Bioengineering horizon scan 2020 - POINT OF VIEW - Office of Research and ...
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
FEATURE ARTICLE
POINT OF VIEW
Bioengineering horizon scan
2020
Abstract Horizon scanning is intended to identify the opportunities and threats associated with
technological, regulatory and social change. In 2017 some of the present authors conducted a horizon
scan for bioengineering (Wintle et al., 2017). Here we report the results of a new horizon scan that is
based on inputs from a larger and more international group of 38 participants. The final list of 20
issues includes topics spanning from the political (the regulation of genomic data, increased
philanthropic funding and malicious uses of neurochemicals) to the environmental (crops for changing
climates and agricultural gene drives). The early identification of such issues is relevant to researchers,
policy-makers and the wider public.
LUKE KEMP*, LAURA ADAM, CHRISTIAN R BOEHM, RAINER BREITLING,
ROCCO CASAGRANDE, MALCOLM DANDO, APPOLINAIRE DJIKENG,
NICHOLAS G EVANS, RICHARD HAMMOND, KELLY HILLS, LAUREN A HOLT,
TODD KUIKEN, ALEMKA MARKOTIĆ, PIERS MILLETT, JOHNATHAN A NAPIER,
CASSIDY NELSON, SEÁN S ÓHÉIGEARTAIGH, ANNE OSBOURN,
MEGAN J PALMER, NICOLA J PATRON, EDWARD PERELLO,
WIBOOL PIYAWATTANAMETHA, VANESSA RESTREPO-SCHILD, CLARISSA RIOS-
ROJAS, CATHERINE RHODES, ANNA ROESSING, DEBORAH SCOTT,
PHILIP SHAPIRA, CHRISTOPHER SIMUNTALA, ROBERT DJ SMITH,
LALITHA S SUNDARAM, ERIKO TAKANO, GWYN UTTMARK, BONNIE C WINTLE,
NADIA B ZAHRA AND WILLIAM J SUTHERLAND*
Introduction (Hanea et al., 2017), with some minor adjust-
Bioengineering is expected to have profound ments (see Methods). We tasked our experts
impacts on society in the near future as applica- with identifying ‘novel, plausible and high-
tions increase across multiple areas, while costs impact’ issues in biological engineering, and
*For correspondence: ltk27@cam. and barriers to access fall. The speed of this they produced a long list of 83 issues. Partici-
ac.uk (LK); w.sutherland@zoo.cam. change and the breadth of the applications pants then scored the issues anonymously (with
ac.uk (WJS)
make the task of forecasting the impacts of bio- a score out of 1,000, reflecting likelihood,
Competing interest: See engineering both urgent and difficult (Gus- impact and novelty), arriving at a short list of 41
page 15 ton, 2014). In 2017 we published the results of a issues to be discussed at a workshop. This was
Funding: See page 15 ’horizon scan’ that looked at emerging issues in coupled with a ‘yes/no’ question to determine
bioengineering (Wintle et al., 2017). Here we whether the issues were novel, based on
Reviewing editor: Helena Pérez
report the results of an updated horizon scan whether the experts had heard of the issue pre-
Valle, eLife, United Kingdom
based on a wider range of inputs (38 partici- viously. After deliberation, participants re-scored
Copyright Kemp et al. This these issues. The issues identified in the latest
pants from six continents and 13 countries, com-
article is distributed under the
pared with 27 participants from the UK and US horizon scan differ substantially from those iden-
terms of the Creative Commons
in the 2017 exercise) and a broader definition of tified in 2017. This change likely stems from an
Attribution License, which
permits unrestricted use and bioengineering. increase in the diversity of the participants,
redistribution provided that the We followed the same structured ‘investigate, improvements in the methods used, a broader
original author and source are discuss, estimate and aggregate’ (IDEA) proto- definition of bioengineering, and changes in the
credited. col for identifying and prioritising issues research landscape since 2017.
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 1 of 20Feature Article Point of View Bioengineering horizon scan 2020
Since it was undertaken, there have been the basis for research roadmaps
developments in a number of the issues identi- (Kennicutt et al., 2015).
fied in the 2017 bioengineering horizon scan. In this article we provide a high-level sum-
Human germline genome editing came to prom- mary of the top 20 issues identified in the bioen-
inence in late 2019 when researcher He Jiankui gineering horizon scan 2020 (while
announced the birth of two girls with CRISPR/ acknowledging that the number of topics cov-
Cas9-edited genomes (Cyranoski, 2019). Mili- ered means that there will be some sacrifice of
tary funding of bioengineering projects also depth for breadth). We take a broader view of
remained substantial: for example, projects bioengineering than we did in 2017, defining it
funded by DARPA included programs to explore as the application of ideas, principles and techni-
the use bioelectronics for tissue repair and ques to the engineering of biological systems.
regeneration (BETR) and to develop mosquito- This means that we now cover more aspects of
repellent skin (ReVector). There have also been bioengineering, as well as issues that contribute
to or result from bioengineering advances (such
breakthroughs in the use of enhanced photosyn-
as funding). To avoid giving a false sense of fore-
thesis for agricultural productivity: a 2018 study
casting precision or overemphasising minor dif-
reported that metabolic engineering strategies
ferences in scoring, the issues are not ranked,
increased photosynthetic efficiency by 17%,
and are instead grouped into issues that are
which resulted in an increase of about 20% in
expected to be most relevant within five years,
biomass in field conditions (South et al., 2019).
within 5–10 years, and on timescales of longer
This technology is now being deployed in sev-
than 10 years (Table 1). Our intent is to spur fur-
eral crops. The use of ‘platform technologies to ther research into these issues and further dis-
address emerging disease pandemics’, another cussion of their implications by researchers,
topic identified in 2017, has taken on particular policy-makers and the wider public.
significance as a result of the COVID-19 pan-
demic. Many of the vaccine candidates for
COVID-19 currently undergoing clinical and pre- The issues most relevant within
clinical evaluation have been developed from five years
platforms for non-coronavirus candidates such as
influenza, SARs and Ebola (WHO, 2020). Access to biotechnology through
Horizon scanning aims to build societal pre- outsourcing
paredness by systematically identifying upcom- Traditionally, the biotechnology sector has had
ing opportunities and threats from high barriers to entry, with organizations need-
technological, regulatory and social change ing to build extensive physical and knowledge-
based assets. New ’cloud labs’ and services labs
(Sutherland and Woodroof, 2009). Horizon
are circumventing this model by using technolo-
scanning with the Delphi technique has a long
gies such as robotics, automation and the inter-
history. It has been used to identify emerging
net to offer widely-accessible standardised
critical issues in areas as diverse as conservation
services with limited need for physical material
biology (Sutherland et al., 2006;
transfer (Jessop-Fabre and Sonnenschein,
Sutherland et al., 2017), invasive species in the
2019). This facilitates both broader access and
UK (Ricciardi et al., 2017), poverty reduction
faster development of new products through the
(Pretty et al., 2010) and biosecurity
sharing of capital and knowledge across projects
(Boddie et al., 2015). Periodic horizon scanning (Lentzos and Invernizzi, 2019). It is also helping
is also undertaken in some areas: in global con- to empower non-traditional researchers by low-
servation, for example, these scans have identi- ering the threshold for participating in cutting-
fied issues such as micro-plastics, gene editing edge research.
for invasive species, and cultivated meat approx- This distributed approach poses a biosecurity
imately six years before they captured public gap as research activities are separated from
attention (Sutherland et al., 2017). Horizon- intent: the cloud lab may not seek additional
scanning activities related to the Antarctic and details on an experiment’s context, including
Southern Ocean (Kennicutt et al., 2014) have why it is being performed. There is also a lack of
also directed funding and policy appropriate biosecurity guidelines and gover-
(Kennicutt et al., 2019), and helped to provide nance models to handle this (Palmer et al.,
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 2 of 20Feature Article Point of View Bioengineering horizon scan 2020
Table 1. Overview of the bioengineering horizon scan 2020.
Summary of the 20 issues identified through the scan; issues are grouped according to likely timeline for realisation.
10 Years
Access to biotechnology through Agricultural gene drives Bio-based production of materials
outsourcing
Crops for changing climates Neuronal probes Live plant dispensers of chemical
expanding new sensory signals
capabilities
Function-based design in protein Distributed pharmaceutical Malicious use of advanced
engineering development and neurochemistry
manufacturing
Philanthropy shapes bioscience Genetically engineered Enhancing carbon sequestration
research agendas phage therapy
State and international regulation Human genomics Porcine bioengineered
of DNA database use converging with replacement organs
computing technologies
Microbiome engineering in The governance of cognitive
agriculture enhancement
Phytoremediation of
contaminated soils
Production of edible
vaccines in plants
The rise of personalised
medicine such as cell
therapies
2015; Dunlap and Pauwels, 2019). As outsourc- 2019). Technical progress is still required for
ing through cloud labs becomes increasingly success in the field. However, deployment may
prevalent in the next five years, these challenges be hindered by a comparative lack of funding
may require the development of new guidelines for plant science, as well as lengthy and expen-
and business and incentive models for responsi- sive regulatory regimes in most jurisdictions.
ble innovation and biosecurity. New models for public-private co-operation will
be needed to advance the translation of basic
Crops for changing climates research through to the field, including business
Climate change is predicted to result in more models that are not based on simple economic
frequent droughts and intensive precipitation returns. The effects of novel traits on biodiversity
events. This will increase soil salinity, elevate and ecosystems will require further scrutiny
average temperatures, and shift the range, before being deployed in a warmer world.
abundance and genotypic diversity of pollina-
tors, pests and pathogens. All of these factors Function-based design in protein
are expected to impact crop yields. In response, engineering
efforts are intensifying to adapt food production Despite a growing understanding of the relation-
using agro-ecological strategies (Altieri et al., ship between protein structure and function,
2015), as well as the provision of well-adapted efficient design of new proteins with a desired
crop varieties by genetic engineering and new action has remained a laborious process. For
breeding technologies (Dhankher and Foyer, example, chimeric antigen receptor (CAR) thy-
2018): Drought-tolerant genetically modified mus lymphocyte (T cell) therapies which combine
(GM) plant varieties have reached the market functional protein moieties to activate T cells
and more are in development (Nuccio et al., against malignant tumours have only recently
2018); the capabilities of plant immune recep- been approved for human use after decades of
tors have been broadened by protein engineer- iteration (Feins et al., 2019). The convergence
ing (De la Concepcion et al., 2019); and the of ongoing developments, including substantial
identification of conserved submergence-acti- improvements in predicting protein structure
vated genes has revealed novel genetic targets from amino acid sequences using machine learn-
for enhancing flood tolerance (Reynoso et al., ing (AlQuraishi, 2019; Yang et al., 2019a),
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 3 of 20Feature Article Point of View Bioengineering horizon scan 2020
could overcome previous technical and compu- State and international regulation of
tational challenges. This indicates a potential deoxyribonucleic acid (DNA) database use
revolution in function-based protein design, Personal genomic sequencing continues to drop
leading to various useful industrial compounds in price and increase in accessibility. The inher-
(such as the development of catalysts for any ent inability to truly anonymise such data, cou-
desired organic reaction) and medical applica- pled with the wealth of information it provides
tions (such as the ability to selectively destroy, on both individuals and families, distinguishes it
suppress or stimulate any malfunctioning tissue, from conventional data types such as finger-
which is the key to treating many refractory dis- prints (identifiable but uninformative) or shop-
eases). However, as this field grows so will the ping habits (Finnegan and Hall, 2017). The
risk of deliberate misuse. Protein engineering drop in price and the use of technologies such
could be used to produce agents that have a as cloud storage have allowed wider use of DNA
higher lethality or specificity than existing agents databases by different actors. While the vulnera-
(including new agents based on novel mecha- bility of cloud infrastructure is a concern, there is
nisms of action). Protein engineering might also greater potential for misuse by states and law
enforcement in the name of security. This has
simplify the production of toxins currently
been seen in efforts to target Muslim Uighurs in
derived from natural sources.
China via blood samples (Wee, 2019), and in a
consumer genetics database allowing the Fed-
Philanthropy shapes bioscience research
eral Bureau of Investigation in the US to com-
agendas
pare genetic data from crime scenes to a
Over the past decade, philanthropic funding
database of over two million profiles without
(including venture philanthropy) of research and
customer consent (Haag, 2019). The potential
innovation has been increasing (Coutts, 2019;
to accrue and analyse vast amounts of genomic
Depecker et al., 2018). This has largely been
information raises concerns over privacy, espe-
driven by the increasing concentration of wealth,
cially mass surveillance (Solove, 2011); the
and erosion of public health and scientific potential expansion of state surveillance powers
research initiatives within key countries. These necessitates dialogue and policy intervention
investments can provide particular research domestically and internationally.
groups or areas with substantial funding over
prolonged periods of time, and they can also
support areas of research that are not usually Issues most relevant in 5–10 years
funded by governments. Philanthropic invest-
ments can also promote innovation, such as Agricultural gene drives
allowing for more exotic approaches not usually Gene drives were initially proposed for the con-
funded by governments. However, these invest- trol of insect vectors for human diseases
(Gantz et al., 2015; Neve, 2018), but recent
ments might also influence the development of
work suggest that they could provide major eco-
biotechnologies in a way that has less of a public
nomic benefits to the agricultural sector (Col-
mandate than government-funded research.
lins, 2018; Neve, 2018). However, while there is
Philanthropic investments also operate without
potential for gene drives to eliminate or sup-
the traditional mechanisms for accountability,
press pest species, their widespread uptake and
transparency or oversight often required by fed-
use could lead to problems in their application
eral or state law (Reich, 2018). Some areas of
and governance (Evans and Palmer, 2018). One
medical research are already considerably
concern is that commercial interests will seek to
underfunded compared to health needs maintain sales of agrochemicals by configuring
(Rafols and Yegros, 2018), and philanthropic gene drives to reduce chemical resistance in tar-
investments may exacerbate this discrepancy in get pest insects and weeds as opposed to caus-
the near-term future. Significant investment into ing sterility in those species. A second concern is
a small range of actors could also undermine that unilateral deployment of gene drives may
diversity, particularly at the international level cause rapid and unintended ecosystem pertur-
(Lentzos, 2019). A possible response would be bations without proper oversight or recall. There
partnerships between public and private invest- have also been questions around their control
ors, though such partnerships might not fully and the lack of public consultation (or participa-
address concerns about accountability, transpar- tion) regarding their release, as well as legal
ency or oversight. implications if populations are eliminated within,
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 4 of 20Feature Article Point of View Bioengineering horizon scan 2020
or new gene configurations are carried to, native research and development (Munos, 2010;
locations (Montenegro de Wit, 2019). Efforts Gassmann et al., 2018; Open Source Pharma,
are already underway to counter, control and 2020). That itself may shape regulatory frame-
even reverse the undesired effects of genome works, and may provide new open or distributed
editing, including DARPA’s Safe Genes program models for drug manufacturing. However, in the
(Wegrzyn, 2019). Policy-makers will need to be absence of appropriate norms or regulations
vigilant to more problematic applications as (Blum, 2010), it may also lead to the
agricultural gene drives become more prevalent. manufacturing, at scale, of drugs that are not
vetted for safety, or administered under appro-
Neuronal probes expanding new sensory priate clinical guidance (Coleman and Zilinskas,
capabilities 2010).
New research into creating probes that mimic
neurons could enable novel medicinal and Genetically engineered phage therapy
enhancement applications such as the creation The World Health Organization (WHO) recently
of new sensory capabilities. Traditionally, neuro- reported a worrying lack of new antibiotics to
nal probes have both structural and mechanical address the dangerous trends of rising resis-
dissimilarities from their neuron targets, leading tance to existing antibiotics (WHO, 2015), and
to neuro-inflammatory responses. However, it is antimicrobial resistance has been identified as a
now possible to fabricate neuron-like electronic potential global catastrophic risk. Phage therapy
probes (with widths similar to those of neurons) has recently seen a renaissance as a potential
and unobtrusively fuse them with live neurons alternative to antibiotic treatment. In particular,
(Yang et al., 2019b). Potentially, the technology the ability to rapidly engineer phage sequences
could be used to add new sensory capabilities and phage cocktails opens up the prospect of
by implanting neuronal probe arrays as a visual personalised treatments for tackling genetically-
cortical prosthesis system. However, such biomi- diverse infections and overcoming problems of
metic sensory probes could introduce unin- antimicrobial resistance (Schmidt, 2019). The
tended vulnerabilities, from a risk of malicious technical advances observed in the medical
attack via the internet to possible mass monitor- application of phage therapy will also have an
ing of implanted civilians by law enforcement impact on other uses of phages as delivery sys-
(Yetisen, 2018). tems in biotechnology. Efforts have also been
significantly buoyed by the development of eas-
Distributed pharmaceutical development ier methods for engineering phages to combat
and manufacturing the inevitable evolution of phage resistance in
Outsourcing and increasingly lower barriers to bacteria (Pires et al., 2016). However, barriers
access in bioengineering are allowing for greater to widespread commercial use persist, including
localisation and geographical distribution of the high costs, instability of the medication, the
manufacturing and development of pharmaceut- need to type the infection (instead of giving a
icals. Bioengineering offers the capacity to cre- broad-spectrum pill) and immunogenicity. This
ate pharmaceutical compounds or their makes it more likely for phage therapy to be
precursors by genetically modifying organisms used as a last resort once other treatments have
to produce them. The prospect of non-tradi- failed.
tional pharmaceutical manufacture has gained
some traction, but with few tangible results. Bar- Human genomics converging with
riers to distributed pharmaceutical manufactur- computing technologies
ing becoming broadly adopted include the scale Human genomics is increasingly incorporating
of production required for individual or commu- technologies such as blockchain, cloud comput-
nity use; meeting appropriate safety standards ing and machine learning. Firms such as Amazon
for manufacturing and administration; and inter- and Google offer cloud computing-based stor-
facing with drug approval pathways. Efforts in age and data analytics services for the petabytes
non-traditional pharma, such as The Open Insulin of genetic data stored online, while companies
Project (Gallegos et al., 2018), are rising in pro- such as Encrypgen and Nebula use blockchain in
file and will likely continue, whether individual systems that reward individuals for sharing their
projects are successful or not. This is supported genetic data. Artificial intelligence and machine
by the Open Pharma movement which seeks to learning are enabling deep analysis of thousands
empower innovation through open-access of molecules with potential to become future
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 5 of 20Feature Article Point of View Bioengineering horizon scan 2020
drugs (Japsen, 2016), as well as human genomic their introduction to and implications for the
data (iCarbonX, 2018). Most recently, deep environment. Certain plant species have natural
learning used molecular structure to predict the mechanisms that enable both uptake and toler-
efficacy of antibiotic candidates (Stokes et al., ance of natural and anthropogenic inorganic pol-
2020). Some uses of these technologies could lutants. Identifying, expressing and potentially
help address current privacy concerns. This engineering these traits is receiving increased
includes the use of blockchain as well as ’secret research interest. Preliminary work on transgenic
sharing’ techniques in which sensitive informa- plants in the lab by overexpression of metal
tion is divided across multiple servers ligands, transporters and specific enzymes has
(Cho et al., 2018). However, as they are applied led to successful phytoextractions of pollutants
to human genomic data in increasingly powerful including explosives and heavy metals. However,
and connected ways, additional ethical issues few experiments have been conducted in the
will arise. Enlivened and global discussion on field on contaminated soils (Fasani et al., 2018),
how best to handle societal implications will where toxicity of various pollutants and the
become necessary (Yakubu et al., 2018). impact of various environmental factors on the
plant-microbiome interaction has limited the
Microbiome engineering in agriculture success of phytoremediation to date. Realising
Progress on microbiome engineering and geno- biotechnological phytoremediation will depend
mic sequencing could allow for beneficial new on a number of factors: a more robust systemic
understanding of plant-microbiome interactions
applications in agriculture, but also risks. Micro-
with pollutants (Basu et al., 2018); the surviv-
biome engineering and the development of syn-
ability of these engineered organisms in the
thetic microbiomes offer wide-ranging uses for
environment; understanding and controlling
mammalian health as well as plant and animal
environmental impacts; and robust societal dis-
productivity, soil health and disease manage-
cussion and carefully designed regulatory
ment. A bottom-up approach to microbiome
regimes.
engineering aims to predictably alter micro-
biome properties and design functions for agri-
Production of edible vaccines in plants
cultural and therapeutic applications.
Plants offer a scalable low-cost platform for
Microbiome engineering strategies could pro-
recombinant vaccine production (Merlin et al.,
vide alternatives to the use of antibiotics for live-
2017). The introduction of the oral polio vaccine
stock management (Broaders et al., 2013).
in the 1960s led to huge interest in developing
These approaches offer the potential for innova-
vaccines that can be delivered without the need
tive, sustainable pathways for plant disease sup-
for injection. Given that plants are widely con-
pression by engineering the microbiomes
sumed, they offer an attractive means of vaccine
indigenous to agricultural soils (Foo et al.,
delivery. Plant-expressed antibodies can protect
2017). Advances in genome sequencing, meta-
against tooth decay. Similarly, expression of nor-
genomics and synthetic biology have already ovirus-like particles in transgenic potatoes could
provided a theoretical framework for construct- raise antibodies against the virus when the mate-
ing synthetic microbiomes with novel functionali- rial is consumed (Tacket et al., 2000). Plant-pro-
ties. New methods, such as in situ mammalian duced vaccines have also been developed for
gut microbiome engineering, could help to over- some animal diseases (Marsian et al., 2019).
come existing limitations and offer new capabili- Oral delivery with minimal processing has the
ties for the future (Ronda et al., 2019). These potential to reduce requirements for extensive
new methods and advances can support better frameworks for production, purification, sterilisa-
design of microbiome modulation strategies in tion, packaging and distribution. A major chal-
mammalian health and agricultural productivity. lenge is the need for improvement of the
Yet, the engineering of agricultural microbiomes chemical and physical stability of vaccines during
on a large scale could also create vulnerabilities transit through the gut in order to ensure effi-
towards malicious intervention. cacy (Berardi et al., 2018). Also, commercialisa-
tion may be difficult under current regulatory
Phytoremediation of contaminated soils regimes (Merlin et al., 2017). Moreover, if pro-
Research in phytoremediation is leading to the duction is scaled up beyond contained green-
creation of engineered plants that could help houses, this will require the deliberate
recuperate contaminated soils, but further field environmental (field) release of plants engi-
trials are needed along with discussions about neered to contain vaccines.
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 6 of 20Feature Article Point of View Bioengineering horizon scan 2020
The rise of personalised medicine such as existing methods, attention is still required in
cell therapies addressing specific impacts on feedstocks,
There is an accelerating trend towards the devel- energy, water and other environmental and soci-
opment and approval of personalised therapeu- etal factors (Matthews et al., 2019). This is
tics. These are medical treatments that are accompanied by technical barriers in product
tailored towards individuals, accounting for their processing. While some bio-based materials are
likely response based on genomic and epige- already on the market, significant private invest-
netic data. In the US in 2018, 42% of all new ment and supportive public policy frameworks
drug approvals by the Food and Drugs Adminis- (including but not limited to carbon pricing, as
tration concerned these treatments well as more speculative nitrogen pricing) will be
(PMC, 2019). However, significant challenges required over the next decade and beyond to
stand in the way of developing and deploying accelerate the widespread worldwide transition
personalised medicine and cell therapies. These to these materials (HM Government, 2018).
includes issues of delivery logistics and cost. The
key factor to clinical adoption of personalised Live plant dispensers of chemical signals
medicine is the value recognition by all health- Plants emit volatile signals that can activate
care stakeholders. Most personalised medicines defence responses in other nearby plants. The
are genetically guided interventions that address concept of using GM plants to deliver these sig-
relatively small subsets of patients with rare nals has made practical progress in recent years.
genetic mutations. The treatment approaches These genetically modified plants are intended
are sometimes costlier due to their increased to be helpers that protect surrounding conven-
sophistication and lower demand. Once these tional crops that are cultivated for consumption.
barriers are overcome there will be some poten- Field trials have evaluated the potential of trans-
tial problems that will need to be mitigated via genic wheat to repel different pests and virus
policy. One is ensuring equitable access. Reim- vectors (Bruce et al., 2015). Despite excellent
bursement from third-party payers such as results in the lab, in planta synthesis of the alarm
health insurance companies is also likely to pheromone failed to reduce aphid numbers.
become an issue for targeted treatments Other studies have demonstrated the feasibility
(Bilkey et al., 2019; Genetics Home Reference, of making insect sex pheromones to trap male
2019). Public health policy must adapt to this insects (Ding et al., 2014). Further finessing of
new frontier of healthcare while addressing its the pheromone blend may be enabled by syn-
potentially detrimental effects on equality of thetic biology. This could open up the possibility
healthcare access and treatment. of using plants as chemical-producing green fac-
tories, or field-based disruptors and dispersers
of insect pests. Unlike current GM solutions for
The issues most relevant in 10 protection from insect herbivory, the use of
years or more pheromones is a non-lethal and less-persistent
Bio-based production of materials intervention, and chemically-manufactured pher-
Biological engineering and production methods omones have been in use for many years. Ques-
facilitate the transformation of renewable plant tions remain as to whether the broader adoption
feedstocks and microorganisms into substitutes of pheromones will simply displace pests to
for a wide range of existing and new materials, unprotected crops.
including plastics and other materials that are
produced from fossil fuels Malicious use of advanced neurochemistry
(European Commission, 2017). These develop- Agents that could attack the central nervous sys-
ments are being driven by increasing govern- tem were investigated during the Cold War but
ment, private and civil society efforts to lack of knowledge only permitted the develop-
decarbonise economies. New opportunities may ment of sedating agents. Concerns over such
be created for small, bio-based production facili- agents and manipulations continues
ties and clean bio-refineries to be located close (Ward, 2019), but could be empowered through
to the markets for these materials, potentially advances in neuroscience and other fields. A
replacing much of the petrochemical sector, and driving force in these advances is significant gov-
there are potential roles for rural areas in grow- ernment interest and investments, including an
ing bio-based feedstocks. While bio-based pro- investment of almost $1bn by the US govern-
duction promises to be more sustainable than ment in the Brain Initiative (NIH, 2019).
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 7 of 20Feature Article Point of View Bioengineering horizon scan 2020
Resulting drugs and nootropics offer health ben- endogenous retroviruses, which otherwise pose
efits, but could also be maliciously used a risk of cross-species transmission (Niu et al.,
(Nixdorff et al., 2018). Governments could use 2017). Such advances hold promise as one tech-
neuro-chemicals to make a populace more sub- nological way to address the global shortage of
servient. Advanced applications in undeclared transplant organs. Over 6,500 patients died
biological warfare could include fostering emo- while on waiting lists in the USA alone in 2017
tional resentment in a targeted population. (UNOS, 2019). Several challenges remain,
These drugs could be appealing to governments including engineering sufficient immune compat-
around the world as a tool for counter-insur- ibility in the organs for successful human trans-
gency or non-lethal law enforcement. The use of plantation, and determining the expected
these new chemicals for law enforcement and in lifespan of the porcine organs in humans. There
non-traditional conflicts may greatly erode the are differing views over the acceptability of por-
norms against chemical agent use on the battle- cine xenotransplantation within major religions,
field, threatening the Chemical Weapons Con- such as Islam and Judaism (Nuffield Council on
vention in the long term. Bioethics, 1996). Before commercial develop-
ment, consideration must be given to questions
Enhancing carbon sequestration surrounding the ethics of using animals for trans-
Metabolic engineering manipulates cells to pro- plantation, cost and access, and using a techni-
duce target molecules by optimising endoge- cal solution for an essentially social problem that
nous metabolic pathways or by reconstructing could be addressed through other approaches,
these pathways in alternative species. ’Next
such as opt-out organ donation schemes.
level’ metabolic engineering aims to design met-
abolic networks de novo, thus bypassing the The governance of cognitive enhancement
bottlenecks and inefficiencies of evolution
Cognitive enhancement is already a widely
(Erb, 2019). Thus far, experimental success is
embraced idea throughout society – caffeine is
lacking. However, recent research in photosyn-
the most widely consumed drug on earth. Novel
thesis may be promising, and examples include
methods of cognitive enhancement such as noo-
engineering a new molecule to perform a
tropics, wakefulness enhancers, or the potential
designed synthetic photorespiration bypass
to directly modulate brain function through
(Trudeau et al., 2018) and developing an opti-
implants or biotechnology are emerging. Uptake
mised carbon dioxide fixation pathway using
of these is being driven by both a productivity-
enzymes from bacteria, archaea, plants and
focused culture, commercial opportunities and
humans (Schwander et al., 2016). Other meth-
increased understanding of neurochemistry.
ods have included laboratory evolution of a bac-
Although some cognitive enhancers require pre-
terium able to use CO2 for growth
(Gleizer et al., 2019). These approaches hold scriptions, others only have to meet basic safety
potential for more efficient carbon sequestration guidelines and are available to purchase online.
and biomass production, as well as for advanc- While numerous trials have supported the safety
ing the development of photovoltaics (the pro- of most nootropics and wakefulness enhancers,
duction of electricity from light) and light- there are few long-term longitudinal studies
sustained biomanufacturing. Yet, such develop- (Fond et al., 2015). A large section of those
ments remain speculative. There are still signifi- who have embraced cognitive enhancement –
cant technical challenges to overcome, and a the ’do-it-yourself’ experimenters – may also be
long path to widespread commercial deploy- ignored by the research community. Lax regula-
ment. Moreover, the field will need to engage tion around safety standards for these products
with its socio-political, ethical and environmental and tools has led to calls to tighten regulatory
dimensions. loopholes, and for academic researchers to part-
ner with and include communities in research on
Porcine bioengineered replacement organs cognitive enhancers (Wexler, 2017). Regulatory
Pigs represent a promising candidate species for frameworks are necessary to both minimise risks
production of human-compatible replacement and gather long-term safety data from end-
organs for xenotransplantation. A recent users, as well as to provide health and safety
advance in porcine genome editing using guidance for international trade of cognitive
CRISPR/Cas9 addresses one of the key scientific enhancing drugs and devices (Maslen et al.,
challenges: successful inactivation of porcine 2015).
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 8 of 20Feature Article Point of View Bioengineering horizon scan 2020
Discussion deployment of GMOs and are now prominent in
discussions around gene drives (Evans et al.,
Emergent themes 2019). Such concerns also factored into many of
Seven underlying themes emerged from the the issues we have identified, most notably edi-
workshop discussion: 1) political economy and ble vaccines and live plant dispensers. Further
funding; 2) ethical and regulatory frameworks; 3) development of bioengineered products will
climate change; 4) transitioning from lab to field; require appropriate regulation. Additionally, the
5) inequalities; 6) technological convergence; necessary social, environmental and human
and 7) misuse of technology. None of these health risk assessments need to take place to
were judged precise enough to qualify as hori- transition bioengineering from the lab into the
zon-scanning items, although some sub-compo- wider world.
nents were. These themes represent underlying A fifth theme is the potential for bioengineer-
commonalities and drivers across issues. ing to exacerbate existing inequalities in wealth
First, participants expressed concern about and health. This factored into several issues
the political economy of bioengineering (that is, including the rise of personalised medicine,
how political and economic institutions influence replacement organs, and the regulation of cog-
bioengineering, including the role of regulation nitive enhancement. In contrast, distributed
and politics) and, related to this, about funding. pharmaceutical development and manufacturing
These concerns centred around a view that was an emerging area fuelled in part by the
research funded by the military, industry or phi- desire to deliver more equitable, cheap and
lanthropy was less accountable than civilian gov- accessible medicine. Ensuring that the benefits
ernment-funded research and could create real of bioengineering are spread fairly and widely
or perceived conflicts of interest (see, for exam- will be a defining feature of future debates.
ple, Licurse et al., 2010). Enhancements also come with risks, especially at
Second, a recurring theme across several the earliest stages. Many of these are expected
issues was the need for ethics and better regula- to be borne by unwilling or uninformed recipi-
tory frameworks to manage the problems ents (as in the case of the CRISPR twins) before
expected to emerge from technologies on the being marketed to the wealthy. These problems
horizon. This was true for most issues of inequality also highlight the need for horizon-
highlighted in the scan, ranging from carbon scanning efforts to make efforts to include rep-
sequestration to bioengineered replacement resentatives from more oppressed and marginal-
organs. This underscores the need for greater ised groups.
engagement between ethicists, social scientists, The sixth theme is that the convergence of
policy-makers and the cutting-edge of different technologies will be crucial in the future
bioengineering. development of bioengineering. Many of the
Third, climate change is likely to be a critical issues in this horizon scan are driven by progress
driver of bioengineering in the future. Our list in adjacent fields. Both neuronal probes and
includes an application to both adaptation malicious uses of neurochemistry will be enabled
(crops for changing climates) and negative emis- by progress in neuroscience, and the overlap of
sions (sequestration). Others, such as live plant human genomics with computing technologies
dispensers, could be boosted in relevance as a brings both opportunities and threats. As auto-
way to enhance agricultural productivity in the mation and measurement, neuroscience, chemis-
face of detrimental climate impacts. Progress in try and artificial intelligence continue, they will
climate policies will shape the development and shape both what is possible and what is pursued
demand of bioengineering technologies. Cli- in bioengineering. This poses a challenge for
mate change impacts will also create new prob- regulators, who may need to think about policy
lems that could be addressed through that cuts across bioengineering into other areas,
bioengineering and policy. This includes changes such as cybersecurity. It also highlights a need
in the range of vector-borne diseases, such as for continued horizon scanning and foresight
the expansion of tropical infectious diseases. exercises to engage a broad range of technolog-
A fourth theme is that of transitioning from ical expertise so that key points of intersection
lab to field. The deliberate release of a new bio- and convergence are not overlooked.
engineering product into the environment Last, our scan highlights ongoing concerns
entails risks in both practice and perception. around the misuse of technology by state or
Concerns over the unintended consequences of non-state actors. Examples included various bio-
environmental release have hindered the weapons and the misuse of DNA databases.
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 9 of 20Feature Article Point of View Bioengineering horizon scan 2020
Table 2. Overview of the bioengineering horizon scan 2017.
Summary of the 20 issues identified in 2017; issues are grouped according to likely timeline for realisation.
10 Years
Artificial photosynthesis and Regenerative medicine: 3D New makers disrupt
carbon capture for producing printing body parts and pharmaceutical makers
biofuels tissue engineering
Enhanced photosynthesis for Microbiome-based Platform technologies to address
agricultural productivity therapies emerging disease pandemics
New approaches to synthetic Producing vaccines and Challenges to Taxonomy-Based
gene drives human therapeutics in description and management of
plants biological risk
Human genome editing Manufacturing illegal drugs Shifting ownership models in
using engineered biotechnology
organisms
Accelerating defense agency Reassigning codons as Securing the critical infrastructure
research in biological engineering genetic firewalls needed to deliver the
bioeconomy
Rise of automated tools for
biological design, test and
optimisation
Biology as information
science: impacts on global
governance
Intersection of information
security and bio-
automation
Effects of the Nagoya
Protocol on biological
engineering
Corporate espionage and
biocrime
The 2017 scan noted themes of equality, bio- and political economy in the 2020 exercise. This
informatics and regulation, all of which feature reflects the significant deviation in issues
prominently in the 2020 scan (see Table 2 for a between the two studies.
summary of the previous scan). The 2017 exer- Some issues from 2017 also appear in the
cise discussed the intersection between biotech- 2020 exercise in a slightly altered form: concerns
nology and information and digital technologies. about the military use of bioengineering are now
Technological convergence also features in the more specific (for example, ‘Malicious use of
present scan, but with a broader scope encom- advanced neurochemistry’), and there are new
passing neuroscience (adding new sensory capa- concerns about the misuse of DNA databases.
bilities) and neurochemistry (malicious uses of Both scans also focussed on different meth-
advanced neurochemistry) as well as other fields. ods for the production of replacement organs.
Both scans featured a strong emphasis on the The 2017 exercise identified 3D printing cells on
potential for bioengineering to amplify or allevi- organ-shaped scaffolds, while the 2020 exercise
ate inequalities. In the 2017 scan this included examined the potential for porcine genome
the potential for human genomics to create new editing to allow for xenotransplantation. Finally,
’sociogenetic’ classes, while differences in both scans assessed the issue of pharmaceutical
healthcare and access to cognitive enhancement manufacturing becoming increasingly distrib-
were the flagship issues in this 2020 scan. The uted. The 2017 exercise focused on start-up
thematic convergence between the two scans entrepreneurs and biohacking communities,
demonstrates that many of the underlying trends whereas the 2020 exercise took a broader look
in bioengineering include important structural at the possibility of decentralisation.
issues involving ethics and regulation. These will The differences between the scans are likely
likely influence the field for years to come. There due to three reasons. First, we used a wider defi-
were also several differences in themes, includ- nition of bioengineering which encompassed
ing the greater importance of climate change issues such as biomechanical implants. Two of
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 10 of 20Feature Article Point of View Bioengineering horizon scan 2020
the issues identified in this scan would not have under future triennial iterations, and by clearly
been covered by the 2017 definition: neuronal describing the methods used, have made the
probes expanding new sensory capabilities and process open for uptake by others.
the governance of cognitive enhancement. Sec- Future pathways for forecasting bioengineer-
ond, half of the participants (19/38) were not ing issues are manifold. Further updates of this
involved in the 2017 scan; the new participants scan could be paired with systematic reviews of
were also more geographically diverse (see their accuracy and efficacy, as well as deeper
Methods), and included a higher proportion of dives into the issues that have been identified.
social scientists. Third, there have been signifi- Extensions of the horizon-scanning process
cant changes in research and the world at large. could include: focusing on specific areas of bio-
For example, all the research underpinning the engineering, such as catastrophic risks; incorpo-
issue of neuronal probes has occurred in the last rating decision-support tools such as fault-trees;
three years. Similarly, recent research in climate examining the development of bioengineering
change has highlighted the continued increase issues in tandem with overlapping technological
in emissions and warming (Friedlingstein et al., areas such as artificial intelligence; and produc-
2019), and that tipping points are more proba- ing a policy-focused scan which involves greater
ble than previously expected (Steffen et al., engagement with regulators.
2018; Lenton et al., 2019).
Limitations and ways forward Methods
While useful, horizon scanning has its limits. Cri- Our study made use of the Investigate Discuss
tiques have suggested that the Delphi technique Estimate Aggregate (IDEA) protocol. In this pro-
(of which the IDEA protocol is a relatively recent cess, participants were asked to investigate and
evolution) can give unjustified confidence in submit candidate issues, privately and anony-
results that are essentially the subjective judge- mously score the gathered issues, and discuss
ments of experts (Sackman, 1975). However, in their thinking with others. They then provided a
the absence of data, expert elicitation is war- second score which was mathematically aggre-
ranted, and structured approaches such as Del- gated (Hanea et al., 2018a). The element of dis-
phi and the IDEA protocol have been found to cussion is powerful, as the sharing of information
improve group judgement and outperform other between participants has been shown to
forecasting methods, such as prediction markets improve the accuracy of Delphi-style forecasts
(Hanea et al., 2017). While it is difficult to evalu- (Hanea et al., 2018b). The IDEA protocol has
ate the efficacy of the Delphi technique due to also performed well relative to prediction mar-
inconsistencies in its application (de Loë et al., kets in early studies (Hanea et al., 2017).
2016), those that do exist are promising. A Despite being a relatively recent evolution of the
review of a long-term Delphi in predicting devel- Delphi technique, the IDEA protocol has already
opments in the health sector found that results been successfully applied to a range of areas
were accurate in 14/18 identified issues including natural resource management
(Parente and Anderson-Parente, 2011). The (Hemming et al., 2018) and assessing pollinator
method continues to show significant utility in abundance in response to environmental pres-
both accurately sighting emerging develop- sures (Barons et al., 2018). Aside from seeking
ments and exploring the implications of poten- a shared understanding of terms and reducing
tial issues on the horizon. linguistic ambiguity, consensus is not sought
We acknowledge that the issues identified in during discussion and scores are kept anony-
this horizon scan are ultimately representative of mous during both rounds. This is done to avoid
the participants involved. While the 2020 scan is undesirable group dynamics and peer pressure
an improvement on previous efforts in terms of distorting individual judgements. Our use of the
diversity, the majority of respondents were still IDEA protocol can be split into three phases: i)
from a developed economy background. The recruitment and issue gathering; ii) initial scor-
scan did capture a large cross-section of aca- ing; and iii) workshop preparation, deliberation
demic sub-fields in bioengineering, but under- and re-scoring.
represented industry, communities and policy-
makers. Moreover, we achieved a rough gender Phase one: recruitment and issue gathering
balance with 21 male participants (55%) and 17 Our study drew on a group of 38 participants
female participants (45%). We intend to make from six continents. Participants came from
the process increasingly global and diverse countries including the UK, US, Canada,
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 11 of 20Feature Article Point of View Bioengineering horizon scan 2020
Australia, Germany, Croatia, Thailand, France, participant’s issues scores. Z-scores are created
Chile, Peru, Switzerland, Malaysia, Zambia and by subtracting the mean and dividing by the
Pakistan. Recruitment was done via a panel of standard deviation for each issue against the
six initial experts (EP, PM, SÓhÉ, CR-R, CR, LS participant’s set. This ensures that variations in
and BW). The panel aimed to ensure a balance the range of participants’ scoring is accounted
across areas such as plant sciences, medicine, for. We then ranked the average z-scores across
bioindustry and biosecurity. They also sought to the issues and selected the highest ranked 41
have a mix of approximately half new partici- (approximately cutting the long list in half).
pants and half participants from the 2017 exer- We discussed two potential reforms on the
cise scan. Selected bioengineering scholars and previous scoring approach: breaking scoring
practitioners were asked to submit two to five down across the three criteria, and including
issues each. Our initial request was for issues uncertainty estimates. We decided against both
that were ‘novel, plausible and high-impact’. We potential reforms. Experts are poor at estimating
asked participants to provide issues that were at their own uncertainty and this could incentivise
a specific level of granularity. As with the previ- overconfidence. We decided that greater disag-
ous scan we asked participants not to focus on a gregation in voting was likely to impose a
general topic, such as ’gain of function’ research, greater burden on participants while providing
nor on multiple topics simultaneously. Instead little additional benefit. Moreover, keeping the
they were guided to focus on one area within a protocol similar to the 2017 scan was desirable
general topic and its implications, such as an for comparison.
emerging regulatory change for GMOs. After One amendment was made to the previous
duplicates were merged, a long list of 83 issues horizon-scanning methodology: the introduction
was generated from the initial submissions. This of ’devil’s advocates’ into the process.
included 10 merged issues. Goodwin and Wright, 2010 have noted that
most forecasting methods are inadequate for
Phase two: scoring identifying high-impact, low-probability events
Participants were asked to vote on the ’suitabil- (some times called ’black swan events’). How-
ity’ of these issues. This involved assigning a ever, the Delphi technique can be better suited
score of 0–1,000 to each of the issues. Partici- to the task if it includes devil’s advocates who
pants were asked to ensure that each score was can advocate for less likely but significant issues.
unique (no identical scores within a given score- We empowered two individuals during the first
sheet). The suitability scores reflected a combi- phase of the process to propose more specula-
nation of plausibility, novelty and impact. Nov- tive and transformative issues. Two different par-
elty was also captured by respondents noting ticipants were then asked during the third phase
whether they had heard of the issue previously (workshop deliberation) to provide more critical
(through a ’yes/no’ response). We then calcu- inputs and actively push against the prevailing,
lated the percentage of participants who had dominant view during discussions. In each case
heard of each issue. These novelty scores were their designation was not revealed to the group.
published alongside all issues in the short list. The devil’s advocates appear to have been a
This was conducted by sending the participants useful addition and were disproportionately suc-
both the long list of issues, along with a tem- cessful in suggesting issues. Six of the nine
plate score-sheet and instructions. At this stage issues they proposed in the first round made it
participants were reminded that "our aim is to through to the short list, and four of the six
identify plausible, novel bioengineering-related issues they proposed in the second round made
issues with important future implications for soci- it through to the final list of 20; with 38 partici-
ety that are not too broad or already well pants, we would expect approximately only one
known’. They were given approximately three issue for every second participant to make it
weeks to complete their scoring. All anonymised through to the final list. 68% of participants had
score-sheets are provided in Supplementary file heard of the issues proposed by the devil’s
1; this file also includes the z-scores of the top advocates, making these issues moderately
20 issues identified in the 2017 scan. Participants more novel than the rest. Overall, an average of
were also able to provide comments on the dif- 70% of participants had heard of each issue. The
ferent issues on the voting sheet. These critiques level of novelty of the issues suggested by dev-
led to a further eight issues being merged into il’s advocates is partly skewed by two more well-
four. Comments were kept to stimulate future known issues which both scored 82.35%. When
discussion. We calculated the z-scores for each both of these issues were excluded, the devil’s
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 12 of 20Feature Article Point of View Bioengineering horizon scan 2020
Table 3. A comparative analysis of the groups involved with phases one and two, and phase three (the workshop).
Characteristics Phases one and two Phase three (workshop)
Sample Size 38 25
Gender 21 male participants (55%) and 17 female participants (45%) 13 females (52%) and 12 males (48%).
Balance
Geographical 13 countries (UK, US, Canada, Australia, Germany, Croatia, Thailand, 10 countries (UK, US, Canada, Australia, Germany, Croatia,
Coverage France, Chile, Peru, Switzerland, Malaysia, Zambia and Pakistan) Thailand, France, Chile, Switzerland and Pakistan)
Disciplinary 15 participants from humanities and social sciences (39%) and 23 from 9 participants from humanities and social sciences (36%) and
Distribution natural sciences (61%) 16 from natural sciences (64%)
advocates suggestions were significantly more z-scores of the top and bottom 20 issues, but a
novel at an average of 61%. much smaller spread of scores within the top 20.
Participants were then given time to discuss the
Phase three: workshop preparation, final list and whether any amendments were
deliberation and re-scoring needed. The group was content with the spread
The 41 issues with the highest scores were kept of the final list and that it accurately reflected
as a part of a shortlist. These were sent back to the deliberations and hence decided that no
participants on the 13th of September 2019. alterations were needed.
Participants were assigned ’cynic’ roles for each A comparison of the rankings of the top 20
issue. This involved doing deeper background issues after the first and second round of scoring
research into the topic. Each issue had at least can be found in Supplementary file 2. There
two cynics, ensuring that at least three partici- was a noticeable difference between the two
pants (the cynics and proposer) had an in-depth rankings. For example, 11 out of the final top 20
knowledge of the area. The workshop was held (55%) issues had been ranked outside of the top
in Cambridge on the 9th of October 2019 with 20 during the first round of scoring. Indeed, four
25 participants; 13 could not attend due to other of the top five issues (80%) were outside the top
obligations. This resulted in a group with 20 after the first round of scoring. This suggests
approximately the same characteristics as the that deliberation was effective in shifting partici-
group that was involved in he first two phases. pants’ perspectives and scores. The novelty
The characteristics of both groups are compared scores are also summarised in
in Table 3. Overall, the gender balance was Supplementary file 2. The final list of issues had
maintained (although the slight skew was a slightly higher degree of novelty, but this was
reversed towards female participants), the disci- minor. The short list of issues resulted in an aver-
plinary split between social and physical scien- age score of 70.52% and a median of 73.53%.
tists was approximately the same, and the By contrast the final list had an average of
geographical coverage became less balanced 68.97% and a median of 67.75%.
due to the loss of participants from Peru, Zambia
and Malaysia. Acknowledgements
These discussions were overseen by an expe- First and foremost we thank Clare Arnstein for
rienced facilitator (WJS, with LK and AR acting her indispensable work in helping to organise
as scribes) and followed a deliberate structure. the horizon-scanning process and workshop. We
Each issue was discussed for approximately ten thank Andrew Balmer, George Church, Nick
minutes before being voted on anonymously. Matthews, Robert Meckin, Penny Polson, Fate-
During discussions, proposers of the issue were meh Salehi and Andrew Watkins for their
asked not to speak until at least three other involvement in submitting issues. This project
respondents had contributed. This was done to was made possible through the support of a
avoid biasing the conversation and allowing the grant from Templeton World Charity Founda-
cynics time to provide an orientating, more neu- tion. The opinions expressed in this publication
tral intervention. The standardised z-scores for are those of the authors and do not necessarily
each issue were calculated and ranked at the reflect the views of Templeton World Charity
end of the workshop, resulting in a top 20 list. Foundation.
The decision to keep the list to 20 was made by
consensus by the workshop group and was influ- Luke Kemp is in the Centre for the Study of Existential
enced by a significant difference between the Risk (CSER) and the Biosecurity Research Initiative at St
Kemp et al. eLife 2020;9:e54489. DOI: https://doi.org/10.7554/eLife.54489 13 of 20You can also read
