There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.
Abstract
Objectives
The objective of this study is the implementation of an automatic procedure to weekly
detect new SARS-CoV-2 variants and non-neutral variants (variants of concern (VOC)
and variants of interest (VOI)).
Methods
We downloaded spike protein primary sequences from the public resource GISAID and
we represented each sequence as k-mer counts. For each week since 1 July 2020, we
evaluate if each sequence represents an anomaly based on a One Class support vector
machine (SVM) classification algorithm trained on neutral protein sequences collected
from February to June 2020.
Results
We assess the ability of the One Class classifier to detect known VOC and VOI, such
as Alpha, Delta or Omicron, ahead of their official classification by health authorities.
In median, the classifier predicts a non-neutral variant as outlier 10 weeks before
the official date of designation as VOC/VOI.
Discussion
The identification of non-neutral variants during a pandemic usually relies on indicators
available during time, such as changing population size of a variant. Automatic variant
surveillance systems based on protein sequences can enhance the fast identification
of variants of potential concern.
Conclusion
Machine learning, and in particular One Class SVM classification, can support the
detection of potentially VOC/VOI variants during an evolving pandemics.
Ten years ago, a correspondence [1,2], signed by more than 70 championed ‘A global initiative on sharing avian flu data’ (GISAID) [3], leading to the GISAID Initiative in 2008. What started out as an expression of intent to foster international sharing of all influenza virus data and to publish results collaboratively has emerged as an indispensable mechanism for sharing influenza genetic sequence and metadata that embraces the interests and concerns of the wider influenza community, public health and animal health scientists, along with governments around the world. Today GISAID is recognised as an effective and trusted mechanism for rapid sharing of both published and ‘unpublished’ influenza data [4]. Its concept for incentivising data sharing established an alternative to data sharing via conventional public-domain archives. In 2006, the reluctance of data sharing, in particular of avian H5N1 influenza viruses, created an emergency bringing into focus certain limitations and inequities, such that the World Health Organization (WHO)’s Global Influenza Surveillance Network (now the Global Influenza Surveillance and Response System (GISRS) [5]) was criticised on several fronts, including limited global access to H5N1 sequence data that were stored in a database hosted by the Los Alamos National Laboratories in the United States (US) [6,7]. This data repository, set up with financial support from the US Centers for Disease Control and Prevention (CDC) as a first attempt to share ‘sensitive’ data from affected countries, but was accessible only to those who were also providing H5N1 sequence data. This limited-access approach restricted wider sharing of data prior to publication, which was vital for broader understanding of the progress of the emergent public and animal health threat. The need for greater transparency in data sharing and for acknowledgement of those contributing samples from H5N1-infected patients and animals and related genetic sequence data was not satisfied by sharing data after formal publication via public-domain databases. Scientists charged with the day to day responsibilities of running WHO Collaborating Centres (CCs) for Influenza, National Influenza Centres and the World Organisation for Animal Health (OIE)/ Food and Agriculture Organization of the United Nations (FAO) [8] reference laboratories, were therefore eager to play a key role and provide scientific oversight in the creation and development of GISAID’s data sharing platform that soon became essential for our work. A unique collaboration ensued, involving, in addition to members of WHO’s GISRS and OIE/FAO reference laboratories, the wider influenza research community along with officials in governmental institutions and non-governmental organisations. Facilitated by a well-connected broadcast executive with background in licensing of intellectual property, an agreement was drawn up on the sharing of genetic data to meet emergency situations, without infringing intellectual property rights - the GISAID Database Access Agreement (DAA). The DAA governs each individual’s access to and their use of data in GISAID’s EpiFlu database [9]. It was this alliance between scientists and non-scientists, with a diversity of knowledge and experience, involved in drawing up an acceptable simple, yet enforceable, agreement which gained the trust and respect of the scientific community and public health and animal health authorities. The essential features of the DAA encourage sharing of data by securing the provider’s ownership of the data, requiring acknowledgement of those providing the samples and producing the data, while placing no restriction on the use of the data by registered users adhering to the DAA. It essentially defines a code of conduct between providers and users of data, cementing mutual respect for their respective complementary contributions, and upholding the collaborative ethos of WHO’s GISRS, initially established 65 years ago this year [5]. Launched in 2008, the EpiFlu database was of key importance in the response to the 2009 influenza A(H1N1) pandemic, allowing countries to readily follow the evolution of the new virus as it spread globally [10]. Acceptance of the GISAID sharing mechanism by providers and users of data, and the confidence of the influenza community, were further illustrated in 2013 by the unprecedented immediate release of the genetic sequences of Influenza A(H7N9) viruses from the first human cases, by Chinese scientists at the WHO Collaborating Centre for Influenza in Beijing [11,12]. Such events reaffirmed GISAID’s applicability to timely sharing of crucial influenza data. The subsequent use of the sequence data to generate, develop and test candidate vaccine viruses by synthetic biology within a few weeks also demonstrated how GISAID successfully bridged this important ‘technological’ gap [13,14]. The paper by Bao et al. from Jiangsu province of China published in this issue once again confirms the importance of the timely sharing of data on the evolution of the A(H7N9) viruses for global risk assessment. The authors analysed the recently isolated H7N9 viruses form the fifth wave in Jiangsu province, and the results showed no significant viral mutations in key functional loci even though the H7N9 viruses are under continuous dynamic reassortment and there is genetic heterogeneity. These findings should help to reduce concerns raised, even though the number of human infection with H7N9 virus increased sharply during the fifth wave in China. GISAID provides the data-sharing platform particularly used by GISRS, through which sequence data considered by the WHO CCs in selecting viruses recommended for inclusion in seasonal and pre-pandemic vaccines are shared openly and on which research scientists, public and animal health officials and the pharmaceutical industry depend. Such openness of the most up-to-date data assists in an understanding of and enhances the credibility of the WHO recommendations for the composition of these seasonal and potential-pandemic vaccines. Furthermore, in promoting the prompt sharing of data from potential pandemic zoonotic virus infections, as well as from seasonal influenza viruses, GISAID ensures a key tenet of the WHO Pandemic Influenza Preparedness (PIP) Framework [15], highlighting the critical role it plays in mounting an effective mitigating response. GISAID’s ability to facilitate efficient global collaborations, such as the Global Consortium for H5N8 and Related Influenza Viruses [16,17], is central to monitoring phylogeographic interrelationships among, for example, H5 subtype viruses in wild and domestic birds in relation to their incidence, cross-border spread and veterinary impact, and assessing risk to animal and human health [18]. Traditional public-domain archives such as GenBank, where sharing and use of data takes place anonymously, fulfil a need for an archive of largely published data; however, that conventional method of data exchange notably has not been successful in encouraging rapid sharing of important data in epidemic or (potential) pandemic situations, such as those caused by Middle East respiratory syndrome coronavirus (MERS-CoV) and Ebola viruses. While the GISAID EpiFlu database is hosted and its sustainability ensured through the commitment of the Federal Republic of Germany [19], the establishment of GISAID and development of the EpiFlu database was reliant to a large extent on philanthropy of one individual and voluntary contributions and generosity of many others, together with some initial financial provision by the US CDC and the German Max Planck Society. That GISAID has become accepted as a pragmatic means of meeting the needs of the influenza community in part reflects the particular characteristics of influenza and the continual need for year-round monitoring of the viruses circulating worldwide, essential for the biannual vaccine recommendations and assessment of the risk posed by frequent zoonotic infections by animal influenza viruses [20]. In the meantime, calls for an equivalent mechanism to promote the timely sharing of data in other urgent epidemic settings go largely unfulfilled [21,22]. A recent publication considered whether the ‘paradigm shift’ in data sharing by GISAID could be applied more generally to assist in preparedness for and response to other emergent infectious threats, such as those posed by Ebola virus [21] and Zika virus [23]. Such a trusted system could complement and take full advantage of the latest advances in rapid sequencing of specimens in the laboratory and in the field, for outbreak investigation [24]. Given the crucial importance of genetic data in improving our understanding of the progress of an emergent, potentially devastating epidemic, the effectiveness of GISAID in influenza pandemic preparedness is self-evident and provides important lessons for future pandemic threats. While the genetic makeup and the necessary associated data of the different viruses are distinct requiring separate databases/compartments for unambiguous analysis, the modi operandi for sharing genetic data are generic and the GISAID mechanism could be applied to other emerging pathogens. Indeed, the wider implementation of such a data sharing mechanism should be key in concerted efforts to contain spread of disease in animals and threats to human health, in realising the concept of One Health.
Background Artificial intelligence (AI) research in healthcare is accelerating rapidly, with potential applications being demonstrated across various domains of medicine. However, there are currently limited examples of such techniques being successfully deployed into clinical practice. This article explores the main challenges and limitations of AI in healthcare, and considers the steps required to translate these potentially transformative technologies from research to clinical practice. Main body Key challenges for the translation of AI systems in healthcare include those intrinsic to the science of machine learning, logistical difficulties in implementation, and consideration of the barriers to adoption as well as of the necessary sociocultural or pathway changes. Robust peer-reviewed clinical evaluation as part of randomised controlled trials should be viewed as the gold standard for evidence generation, but conducting these in practice may not always be appropriate or feasible. Performance metrics should aim to capture real clinical applicability and be understandable to intended users. Regulation that balances the pace of innovation with the potential for harm, alongside thoughtful post-market surveillance, is required to ensure that patients are not exposed to dangerous interventions nor deprived of access to beneficial innovations. Mechanisms to enable direct comparisons of AI systems must be developed, including the use of independent, local and representative test sets. Developers of AI algorithms must be vigilant to potential dangers, including dataset shift, accidental fitting of confounders, unintended discriminatory bias, the challenges of generalisation to new populations, and the unintended negative consequences of new algorithms on health outcomes. Conclusion The safe and timely translation of AI research into clinically validated and appropriately regulated systems that can benefit everyone is challenging. Robust clinical evaluation, using metrics that are intuitive to clinicians and ideally go beyond measures of technical accuracy to include quality of care and patient outcomes, is essential. Further work is required (1) to identify themes of algorithmic bias and unfairness while developing mitigations to address these, (2) to reduce brittleness and improve generalisability, and (3) to develop methods for improved interpretability of machine learning predictions. If these goals can be achieved, the benefits for patients are likely to be transformational.
This is an open access article distributed in accordance with the Creative Commons
Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute,
remix, adapt, build upon this work non-commercially, and license their derivative
works on different terms, provided the original work is properly cited, appropriate
credit is given, any changes made indicated, and the use is non-commercial. See:
http://creativecommons.org/licenses/by-nc/4.0/.
scite shows how a scientific paper has been cited by providing the context of the citation, a classification describing whether it supports, mentions, or contrasts the cited claim, and a label indicating in which section the citation was made.
Smart Citations