No abstract

Understanding affinity maturation of antibodies that can target many variants of HIV-1 is important for vaccine development. While the antigen-binding site of antibodies is known to mutate throughout the co-evolution of antibodies and viruses in infected individuals, the roles of the mutations in the antibody framework region are not well understood. Throughout affinity maturation, the CH103 broadly neutralizing antibody lineage, from an individual designated CH505, altered the orientation of one of its antibody variable domains. The change in orientation was a response to insertions in the variable loop 5 (V5) of the HIV envelope. In this study, we generated CH103 lineage antibody variants in which residues in the variable domain interface were mutated, and measured the binding to both autologous and heterologous HIV-1 envelopes. Our data show that very few mutations in an early intermediate antibody of the lineage can improve binding toward both autologous and heterologous HIV-1 envelopes. We also crystallized an antibody mutant to show that framework mutations alone can result in a shift in relative orientations of the variable domains. Taken together, our results demonstrate the functional importance of residues located outside the antigen-binding site in affinity maturation.

No abstract

No abstract

We develop a chatbot using deep bidirectional transformer (BERT) models to handle client questions in financial investment customer service. The bot can recognize 381 intents, decides when to say I don’t know, and escalate escalation/uncertain questions to human operators. Our main novel contribution is the discussion about the uncertainty measure for BERT, where three different approaches are systematically compared with real problems. We investigated two uncertainty metrics, information entropy and variance of dropout sampling, in BERT, followed by mixed-integer programming to optimize decision thresholds. Another novel contribution is the usage of BERT as a language model in automatic spelling correction. Inputs with accidental spelling errors can significantly decrease intent classification performance. The proposed approach combines probabilities from masked language model and word edit distances to find the best corrections for misspelled words. The chatbot and the entire conversational AI system are developed using open-source tools and deployed within our company’s intranet. The proposed approach can be useful for industries seeking similar in-house solutions in their specific business domains. We share all our code and a sample chatbot built on a public data set on GitHub.