FULLY AI-DRIVEN HUMANOID VHH PHAGE LIBRARY

Abstract Background & Significance VHHs are small and stable fragments that have great potential as therapeutics due to their size, stability, versatility, for oral administration. The traditional source of is camelids, but humanization usually needed therapeutic development. A human VHH library highly desirable the generation VHHs, natural VH domains unstable standalone units. We developed a humanoid AI-designed sequences both resemble camelid in terms stability such high content no longer needed. Methods In this study, we present fully AI-driven approach de novo design phage library. Firstly, public data nearly one million private were collected. Secondly, autoregressive AI model was trained on another mixed humans camels. Thirdly, CDR1, CDR2, CDR3 regions all generated by mentioned two models. Finally, an ultra large quantity (4E10) utilized build Humanoid Results order verify effectiveness our method, randomly synthesized expressed 26 antibodies from based At same time, 3 molecules reported previous literature included positive controls. First all, success rate expression 96.1%, which much higher than 72% Progen 66% ESMdesign. average titer 59.6mg/L, 1.5 times value control group. hydrophobicity 80% comparable Moreover, immunogenicity less group according proprietary algorithms. diversity naturalness also excellent. Conclusions conclusion, solution could stably massively generate human-like satisfying multiple requirements (including titer, low hydrophobicity, expression, diversity, naturalness) simultaneously. As powerful fragment, has accelerate nanobody bispecific antibody drug