Three days after treatment, a significant median reduction of 4.37 log10and 2.54 log10in infectious disease titre and 3.72 log10and 1.33 log10in viral RNA weight was observed for the UZGENT-COV2 cocktail 10mg/kg and 1mg/kg, respectively, compared to the placebo group (Fig.8c and d). In silicomodelling of antibody-epitope relationships, if accurate, can save time and resources allowing for streamlined development of therapeutics. We queried PubMed for publications on epitope DY 268 prediction published between 2013 and 2023 using search terms such as antibody-epitope prediction, in silicoepitope mapping, sequence centered antibody prediction and AI antibody prediction. Several studies describe computational methods for sequence-based antibody modelling, however true experimental validation of such models is definitely scarce. == Added value of this study == Within the context of this study, AI-assisted epitope prediction was used as the sole method for lead selection while BLI, mutation screening and cryo-EM analyses were performed consequently. == Implications of all the available evidence == Results of this study show that overall performance of AI-based methods forde novoprediction lacks accuracy and still requires experimental input. == Intro == Monoclonal antibody cloning from solitary B cells is becoming accessible to a growing number of academic and industrial laboratories. The unprecedented rate of SARS-CoV-2 neutralizing antibody (nAB) development illustrates this convenience. Within a few months from the start of the COVID-19 pandemic multiple medical trials were initiated to evaluate the effectiveness of nAbs leading to their market authorization by respective agencies worldwide. Currently, DY 268 REGN-COV2,1LY-CoV555, either only or in combination with LY-CoV016,2,3,4VIR 7831,5and additional anti-SARS-CoV-2 Spike (S) antibodies are emergency-approved for restorative or preventive use (examined in Kumar, Chandele6). SARS-CoV-2 offers, however, been shown to rapidly evolve, 7occasionally accumulating multiple mutations actually within one sponsor.8,9The resulting continuous emergence of COVID-19 variants of concern (VOC) Rabbit Polyclonal to USP13 with Omicron as notable example has already deemed most of these nABs ineffective.10Since several medical conditions result in impaired immune responses to both SARS-CoV-2 infection and vaccination, effective nAbs must be available for clinical interventions. With this light, quick epitope mapping of potential nAb candidates is essential in finding pipelines. Several experimental techniques at various levels of complexity such as biolayer interferometry, candida display-based DY 268 deep mutational scanning, or cryo-EM allow to obtain that info. In addition, recently artificial-intelligence (AI)-centered methods have DY 268 been used to forecast epitopes for Abdominal muscles with unfamiliar 3D constructions.11,12,13,14 Here, we describe the isolation and characterization of two potently neutralizing antibodies against the SARS-CoV-2 receptor binding website (RBD) from convalescent patient-derived B cells. Combining info from pseudovirus neutralization assays within silicoepitope prediction via MabTope,15we selected for further screening two nABs with neutralization potencies in the picomolar range and expected not to compete with each other.In vivoexperiments inside a hamster model of SARS-CoV-2 infection proven efficacy of the determined nABs cocktail to be lower than anticipated. Although unexpected based on the AI epitope predictions (except for K417, predicted as part of the UZGENT_A3 epitope), our antibody cocktail lost efficacy against selected VOCs (Beta and Omicron). Ultimately, determination of the SARS-CoV-2-RBD-Antibody complex constructions via cryo-EM exposed that the two antibodies interact with the RBD in a very similar manner with mainly overlapping epitopes, explaining their loss of overall performance in neutralizing selected SARS-CoV-2 variants. == Methods == == Ethics statement == == Human being subjects == A total of 151 convalescent COVID-19 individuals with disease severities ranging from slight (CoSer cohort, 72 individuals) to severe (CoVim cohort, 79 individuals) for whom analysis was either confirmed by PCR and/or ELISA (Supplementary Table S2) were enrolled in this study and thus put through a single blood draw. Patient sampling was authorized by the Ghent University or college Hospitals Honest Committee (applications BC-07492 and BC-08071) and all.