We demonstrate the use of a Generative Adversarial Network (GAN), trained from a set of over 400,000 light and heavy chain human antibody sequences, to learn the rules of human antibody formation. The resulting model surpasses common in silico techniques by capturing residue diversity throughout the variable region, and is capable of generating extremely large, diverse libraries of novel antibodies that mimic somatically hypermutated human repertoire response. This method permits us to rationally design de novo humanoid antibody libraries with explicit control over various properties of our discovery library. Through transfer learning, we are able to bias the GAN to generate molecules with key properties of interest such as improved stability and developability, lower predicted MHC Class II binding, and specific complementarity-determining region (CDR) characteristics. These approaches also provide a mechanism to better study the complex relationships between antibody sequence and molecular behavior, both in vitro and in vivo . We validate our method by successfully expressing a proof-of-concept library of nearly 100,000 GAN-generated antibodies via phage display. We present the sequences and homology-model structures of example generated antibodies expressed in stable CHO pools and evaluated across multiple biophysical properties. The creation of discovery libraries using our in silico approach allows for the control of pharmaceutical properties such that these therapeutic antibodies can provide a more rapid and cost-effective response to biological threats.

No abstract

(2016) CHO cell production and sequence improvement in the 13C6FR1 anti-Ebola antibody, mAbs, 8:2, 347-357, DOI: 10.1080/19420862.2015 To link to this article: https://doi.org/10. 1080/19420862.2015 TM was highly constrained at the time because it was in preclinical development and a novel production system (tobacco plants) was being used for manufacturing. To increase the production of ZMapp TM for an uncertain future demand, a consortium was formed in the fall of 2014 to quickly manufacture these anti-Ebola antibodies in Chinese hamster ovary (CHO) cells using bioreactors for production at a scale appropriate for thousands of doses. As a result of the efforts of this consortium, valuable lessons were learned about the processing of the antibodies in a CHO-based system. One of the ZMapp TM cocktail antibodies, known as c13C6FR1, had been sequenceoptimized in the framework region for production in tobacco and engineered as a chimeric antibody. When transfected into CHO cells with the unaltered sequence, 13C6FR1 was difficult to process. This report describes efforts to produce 13C6FR1 and the parental murine hybridoma sequence, 13C6mu, in CHO cells, and provides evidence for the insertion of a highly conserved framework amino acid that improved the physical properties necessary for high-level expression and purification. Furthermore, it describes the technical and logistical lessons learned that may be beneficial in the event of a future Ebola virus or other pandemic viral outbreaks where mAbs are considered potential therapeutics.

In order to avoid the metabolic burden of protein expression during cell growth, and to avoid potential toxicity of recombinant proteins, microbial expression systems typically utilize regulated expression vectors. In contrast, constitutive expression vectors have usually been utilized for isolation of protein expressing mammalian cell lines. In mammalian systems, inducible expression vectors are typically utilized for only those proteins that are toxic when overexpressed. We developed a tetracycline regulated expression system in CHO cells, and show that cell pools selected in the uninduced state recover faster than those selected in the induced state even though the proteins showed no apparent toxicity or expression instability. Furthermore, cell pools selected in the uninduced state had higher expression levels when protein expression was turned on only in production cultures compared to pools that were selected and maintained in the induced state through production. We show a titer improvement of greater than twofold for an Fc‐fusion protein and greater than 50% improvement for a recombinant antibody. The improvement is primarily due to an increase in specific productivity. Recombinant protein mRNA levels correlate strongly with protein expression levels and are highest in those cultures selected in the uninduced state and only induced during production. These data are consistent with a model where CHO cell lines with constitutive expression select for subclones with lower expression levels.

No abstract