Almost five years have passed since the emergence of the new coronavirus (SARS-CoV-2) in 2019. Looking back on this series of events, what do you feel and think? I am sure there are many thoughts and opinions. Here I would like to comment on my feelings from a medical-scientific point of view in relation to the paper accepted for NeurIPS 2024.
From a Clinician's Point of View
I work as a researcher at COGNANO and also as an internist in a general hospital. So I have been treating patients with COVID-19 since the beginning. In the early days of the outbreak, the typical CT imaging appearance was unique, with multiple bilateral peripheral fine ground-glass opacities. The course of the blood tests was different from that of normal pneumonia, with progressively worsening pneumonia despite decreasing CRP. In general, the fatal course of any infectious disease is seen mainly in people with severe underlying diseases or in the elderly, but COVID-19 can cause death in relatively young people in their 50s and 60s who have no underlying diseases. I remember thinking: 'This is certainly an emerging infectious disease that we cannot be too careful about. Various therapeutic agents have been used, but they rarely have dramatic antiviral efficacy, and the basic coping strategy is still to control the patient's immune response for the time being. In fact, the cause of death is the excessive immune response to the virus, known as ARDS.
COVID-19 Drug Discovery
The speed of drug development for SARS-CoV-2 was unprecedented in the history of drug discovery. It usually takes about 15 years to develop a new drug, but for COVID-19, several small molecule compounds and antibody drugs were developed and used in actual clinical practice within a few years. Small molecule drugs are generally designed to suppress viral growth after infection, but are not expected to prevent infection. Antibody drugs can have a prophylactic effect, but SARS-CoV-2 has an excellent ability to evade the immune system because it is an RNA virus that mutates easily and rapidly develops resistance. The development of new antibody drugs against COVID-19 is not cost-effective for the pharmaceutical industry and has therefore become a low priority for the time being. Fortunately, the mutated viruses have become less virulent and deaths in relatively young people in their 50s and 60s with no underlying disease have almost disappeared. Clinically and medically, the problem appears to be largely solved, and the disease was reclassified as a category 5 infection in Japan in May 2023. The economic aftermath remains, but people's exaggerated fears are fading and the era is now called the post-coronary era.
Only a Few Facts Have Been Elucidated
However, scientific questions remain. Why have mutated viruses become less virulent?　Is it because most people have acquired resistance through infection or vaccination?　What are the laws of viral mutation and antibody change?　Why do people who do not die from COVID-19 not develop an exaggerated immune response?　Will the virus become more virulent in the future?　Even with the vast amount of information that has been gathered in a short period of time, and the huge amount of research and money spent on SARS-CoV-2 around the world, it seems safe to say that no one can answer these questions perfectly at this point.
The same was true in the past with the H1N1 (Spanish flu) subtype of influenza that caused a global pandemic in 1918 and is thought to have killed 10 times as many people as the new coronavirus, and why it is now less virulent and has ended remains unknown. This time, however, a huge amount of time-series data on genetic mutations of the new coronaviruses has been recorded on a global scale and in real time, which may provide clues to elucidate the relationship between immune response and viral mutations. This time COGNANO has published time series big data on antibody changes in alpacas to SARS-CoV-2 mutations. Alpacas have specialised antibodies, and comprehensive next-generation sequencing analysis allows us to capture a huge snapshot of the immune response of antibodies in vivo. Our dataset shows that antibodies that can and cannot respond to different viral variants are found with some regularity. It also shows that the immune response is not static, with antibodies being produced once and continuously, but rather quite fluid. This is a unique and unprecedented data set.
Prospects for The Combined Field of Machine Learning and Molecular Biology
Machine learning is exactly the right tool for discovering and unravelling complicated laws and predicting new ones from vast and complex data. In a sense, the accumulation of sufficient medical molecular biology data and breakthroughs in machine learning have brought the whole science to maturity, and it seems fateful that the Nobel Prize in Chemistry was awarded to Demis Hassabis and John Jumper, who developed the AlphaFold. It is inevitable that the field of machine learning drug discovery will evolve in the future. The application of machine learning is also strongly expected for antibody therapeutics, but the fact is that there is currently an overwhelming lack of basic training data. We believe that the data provided by COGNANO will complement these problems. Please look forward to our future developments.