Umjetna inteligencija u slučaju COVID 19
Possibilities of applying AI in a pandemic
Ruđer Bošković Institute, Centre for Informatics and Computing
Centre of Excellence Data Science DATACROSS RA3
K. Skala, D. Tomić, T. Lipić, D. Davidović, J. Mesarić
Domain problem: AI for social good (AI4SG) - AI for Healthcare - infectious Diseases
New achievements and results enable the rapid and efficient application of AI methods and technology using eInfrastructure to produce results quickly and efficiently.
Addressing issues of dealing with infectious disease outbreaks with AI based solutions can be observed from two perspectives:
(1) Individual human-level:
· concerns on clinical health aspects:
· disease diagnosis:
· reliable detection vs. large-scale screening
· from: biochemistry data, medical imaging/radiomics data (X-Ray, CT, MR), ...
· examples:
· Wang, S., Kang, B., Ma, J., Zeng, X., Xiao, M., Guo, J., ... & Xu, B. (2020). A deep learning algorithm using CT images to screen for Corona Virus Disease (COVID-19). medRxiv.
·
· clinical treatment/prevention:
· drugs discovery approaches (vaccine, antibody and small molecule development) / drug repurposing, predicting vaccine immunogenicity,
· virtual screening tools (pyRX, AutoDock Vina) and databases (ttps://zinc.docking.org/), sequenced virus genome data, ...
· examples:
· Coronavirus puts drug repurposing on the fast track, https://www.nature.com/articles/d41587-020-00003-1
· Coronavirus Deep Learning Competition, https://www.youtube.com/watch?v=1LJgkovowgA
· Tomic, Adriana, et al. "SIMON, an automated machine learning system, reveals immune signatures of influenza vaccine responses." The Journal of Immunology 203.3 (2019): 749-759.
· Lavecchia, Antonio. "Deep learning in drug discovery: opportunities, challenges and future prospects." Drug discovery today (2019).
· clinical prediction:
· We suggest using the Vini in silico model of cancer (developet on Ruđer Bošković Institute) [1] for VDS (virtual drug screening) against COVID-19 virus. Since there is no KEGG pathway for COVID-19 yet, VDS should be performed on ACE2 receptor of coronavirus [2], and the VINI model used in ultra-docking mode [3]. On the pilot phase we obtain that the HIV drugs applicable in COVID19 treatment.
·
(2) Social system-level:
· concerns on predicting public reaction towards disease outbreaks:
· public health aspects:
· epidemics monitoring,
· tracking spread of coronavirus
· Available dataset - 2019 Coronavirus dataset (January - February 2020):
· https://www.kaggle.com/brendaso/2019-coronavirus-dataset-01212020-01262020/activity
· detect signs of potential disease outbreaks from the collected (human behavioural) information
· tracking s
· epidemics forecasting
· AI + complex network epidemiology models (http://www.gleamviz.org/, Vespignani)
·
· epidemics control
· methodology: AI (deep RL), complex networks (settings of disease maximization/minimization problem, network dismantling)
· examples:
· Probert, William JM, et al. "Context matters: using reinforcement learning to develop human-readable, state-dependent outbreak response policies." Philosophical Transactions of the Royal Society B 374.1776 (2019): 20180277.
· Bryan Wilder, Sze-Chuan Suen, and Milind Tambe. Preventing infectious disease in dynamic populations under uncertainty. In Thirty-Second AAAI Conference on Articial Intelligence, 2018.
· Ren, Xiao-Long, et al. "Generalized network dismantling." Proceedings of the National Academy of Sciences 116.14 (2019): 6554-6559.
· COVID-19 and artificial intelligence: protecting health-care workers and curbing the spread, https://www.thelancet.com/journals/landig/article/PIIS2589-7500(20)30054-6/fulltext
· [artificial] complex systems (indirectly) affected with the outbreak:
· socio-economical, financial systems
· political systems
· impacts of outbreaks on political stability
· Morens, David M., Gregory K. Folkers, and Anthony S. Fauci. "The challenge of emerging and re-emerging infectious diseases." Nature 430.6996 (2004): 242-249.
We can also reason about the issues, from the point of view of two conceptual framework frameworks used in AISG survey: AEC (agent – environment – community) and DPP (descriptive – predictive – prescriptive) framework.
FRAMEWORK
|
DESCRIPTIVE
|
PREDICTIVE
|
PRESCRIPTIVE
|
AGENT
|
Disease diagnosis
|
Disease development predictions
|
Treatment recommendation
|
ENVIRONMENT
|
Drug discovery/development
|
Epidemic prediction
|
Preventing spread of illness
|
COMMUNITY
|
(subtyping)
|
(predictive phenotyping)
|
|
AI SERVICE AND RESOURCES FROM SCIENCE
· Epidemic Intelligence Information System (EPIS)
Web based communication platform that allows nominated public health experts to exchange technical information to assess whether current and emerging public health threats have a potential impact in the EU.
AI SERVICES FROM INDUSTRY
HealthMap, https://en.wikipedia.org/wiki/HealthMap
BlueDot - Automated infectious disease surveillance, https://bluedot.global/ [natrag na događanja]
|