Reinforcement Learning Based Diagnosis and Prediction for COVID-19 by Optimizing a Mixed Cost Function From CT Images.

A novel coronavirus disease (COVID-19) is a pandemic disease has caused 4 million deaths and more than 200 million infections worldwide (as of August 4, 2021). Rapid and accurate diagnosis of COVID-19 infection is critical to controlling the spread of the epidemic. In order to quickly and efficiently detect COVID-19 and reduce the threat of COVID-19 to human survival, we have firstly proposed a detection framework based on reinforcement learning for COVID-19 diagnosis, which constructs a mixed loss function that can integrate the advantages of multiple loss functions. This paper uses the accuracy of the validation set as the reward value, and obtains the initial model for the next epoch by searching the model corresponding to the maximum reward value in each epoch. We also have proposed a prediction framework that integrates multiple detection frameworks using parameter sharing to predict the progression of patients' disease without additional training. This paper also constructed a higher-quality version of the CT image dataset containing 247 cases screened by professional physicians, and obtained more excellent results on this dataset. Meanwhile, we used the other two COVID-19 datasets as external verifications, and still achieved a high accuracy rate without additional training. Finally, the experimental results show that our classification accuracy can reach 98.31%, and the precision, sensitivity, specificity, and AUC (Area Under Curve) are 98.82%, 97.99%, 98.67%, and 0.989, respectively. The accuracy of external verification can reach 93.34% and 91.05%. What's more, the accuracy of our prediction framework is 91.54%. A large number of experiments demonstrate that our proposed method is effective and robust for COVID-19 detection and prediction.

Structural basis of a high-affinity antibody binding to glycoprotein region with consecutive glycosylation sites (preprint)

Consecutive glycosylation sites occur in both self and viral proteins. Glycan-shielding of underneath peptide region is a double-edged sword, that avoids immune attack to self-proteins, but helps viruses including HIV-1 and SARS-CoV2 to escape antibody binding. Here we report a high-affinity antibody, 16A, binding to linear peptide containing consecutive glycosylation sites. Co-crystallization of 16A Fab and glycopeptides with GalNAc modifications at different sites showed that STAPPAHG is the sequence recognized by 16A antibody. GalNAc modification at Threonine site on STAPPAHG sequence significantly increased the affinity of Fab binding by 30.6 fold (KD=6.7nM). The increased affinity is conferred by hydrophilic and pi-stacking interactions between the GalNAc residue on Threonine site and a Trp residue from the CDR1 region of the heavy chain. Furthermore, molecular modeling suggested that GalNAc on T site causes more favorable conformation for antibody binding. These results showed that glycan modification most proximal to linear peptide core epitope significantly increases antigenicity of a glycopeptide epitope. The antibody recognition mode by peptide-binding CDR groove with a glycan-binding edge, may shed light on designing of linear glycopeptide-based vaccines for cancer and viral diseases. Teaser A high-affinity antibody was found to bind densely glycosylated glycoprotein region by a peptide binding groove of the antibody’s variant region, with a glycan-binding edge specific to glycosylation site most proximal to core peptide epitope.