The original interview can be accessed from here (in Turkish).
Can Data analytics and Artificial Intelligence help us overcome the COVID-19 Crisis? We asked this question faculty members Lale Akarun and Arzucan Özgür of the Department of Computer Engineering. Lale Akarun is the president of the industry-academia Industry 4.0 platform.
Lale Akarun – The Digital technologies that are at the basis of Digital Transformation have been the subject of research for at least 30 years. They are mature and ready to use for some time but acceptance has been slow due to the drawbacks of causing unemployment and the low return on investment, except for maybe a few high-risk sectors. COVID-19 has acted as an accelerator: We have switched to remote teaching because apparently, teaching in a 100-student classroom seemed to be a risky venture.
Digital Transformation refers to the digitalization of all business areas, the organization, marketing, product development, value chain, production. We have just conducted an assessment of 100 SME’s in Turkey with regard to their digital transformation maturity levels. Our observations show that digital transformation has to start not at production, but in other areas. Companies have to collect data, use data analytics and use it to empower the workers so as to increase productivity. This is the case for classroom teaching: Empowered by videoconferencing and teaching tools such as Moodle, we do a better job of teaching during this pandemic.
Arzucan Özgür:
Interactions between pathogen (e.g., the SARS-CoV-2 virus that causes the Covid-19 disease) and host (e.g., human) proteins play important roles in infection mechanisms. In order for a chemical molecule to be a drug, it must first be able to bind to a target protein of the pathogen or host organism. It also needs to satisfy other features such as being non-toxic, having minimal side effects, and being synthesizable. Information about pathogen-host protein-protein interactions as well as protein-chemical molecule interactions is of vital importance for drug discovery. However, due to the huge number of possible interactions among biochemical molecules, it is not feasible to test and validate each possible interaction through wet lab experiments.
Artificial intelligence algorithms can aid drug discovery by narrowing the hypothesis space and suggesting the most probable hypotheses for validation in the laboratory using wet lab experiments. For example, for Covid-19, artificial intelligence algorithms can be used to estimate the drugs (among all the available drugs) that are most likely to affect target proteins by making use of the available data such as known protein-protein and protein-drug interactions. In order to develop new drugs, in recent years, intensive studies have been conducted on deep learning-based generative models that suggest new chemical compounds, which are likely to affect target proteins and are likely to satisfy the desired properties such as nontoxicity and synthesizability.
Artificial intelligence can also aid in accelerating research by facilitating access to the available scientific knowledge. More than 40,000 scientific articles related to Covid-19 have been published in the last five months, and new articles continue to be published rapidly. It is not possible to manually read all of these articles and consume the knowledge in them. Natural language processing (NLP) techniques can be employed to automatically process these articles and make the important information easily accessible to the researchers. In addition, the data automatically extracted from all scientific articles can be analyzed holistically. Such an analysis may enable the discovery of hidden links that can improve our understanding of the disease and shed light on new drug/vaccine development studies.
