Sponsored content brought to you by
The aerospace and electronics industries have long applied computational models to not only design airplanes and chips, but also to simulate and optimize their performance. There have, until recently, been few comparable tools for biotechnology.
This is due, at least in part, to the complexity of biology. Consider even a “simple” example, such as engineering a cell to make a monoclonal antibody. Modeling this process requires a thorough understanding of the mechanisms at play across biological and physical scales. A particular antibody may be best produced by transcribing heavy and light chain mRNA at particular rates due to differences in degradation or translation rates, and hence benefit from different promoter strengths for each transcription unit; and its expression may be enhanced if its codon distribution is optimized for its host. The best clone may be the one with the best tradeoff in growth and productivity in the target fed-batch or perfusion process. And the best process may be one that modulates media composition to support the cells’ increased catabolic needs for their enhanced growth and antibody secretion, while maintaining glucose in sufficiently low quantity to limit glycation of the antibody.
At Asimov, we develop computer-aided design tools that enable us to predict and optimize the performance of genetic designs, and combine those with data- and physics-based models of bioreactors and bioprocesses. Combining genetic design and bioprocess models allows us to improve the genetic design or the bioprocess not only independently, but also holistically.
Our holistic application of genetic, cellular, and bioprocess models has enabled us to boost antibody titers more than two-fold over traditional approaches, and allows us to achieve antibody titers up to 11 g/L in fed batch culture and lentiviral functional titers up to 109 TU/mL in batch culture.
We tackle this holistic design problem with a portfolio of mechanistic, data-driven machine learning, and hybrid models predicting process performance from genetic design across protein biologics, cell and gene therapy, and mRNA therapeutic platforms. As appropriate, we bring to bear models of:
- codon optimization and signal peptide cleavage, to maximize expressibility and secretability of the gene of interest, or efficiency of viral vector packaging, given a fixed amino acid sequence;
- genetic construct expression dynamics, to design a plasmid for maximal expression in the producer cell line;
- CHO metabolism and phenotype in a fed batch bioprocess, to optimize feed schedule and media composition to maximize titer or improve product quality attributes;
- viral vector expression and packaging dynamics modulated by bioreactor conditions, to optimize transfection or induction timing, process parameters, and harvest timing.
-
Figure 2. A multiscale CHO cell metabolism and bioprocess model, trained on historical bioprocess data, allows Asimov scientists to choose process conditions to improve monoclonal antibody titers
by more than 25%.
The true potential of computer-aided genetic and bioprocess design, though, lies in democratizing its application. In this spirit, we are deploying them in our web-based Kernel software platform for the wider industry and academia to design and manufacture new biotherapeutics.
To learn more about how Asimov is advancing mammalian cell biology, visit our website www.asimov.com.
