In 2021, the FDA approved the 100th monoclonal antibody product, marking a significant milestone since the approval of the first monoclonal antibody (mAb) nearly thirty-five years ago. And the approvals continue to accumulate. As of the first half of this year, six more antibody therapeutics had been granted first approvals in either the US or EU, 20 investigational antibody therapeutics were in regulatory review and the FDA approved the first mAb for use in any animal species.
The pharmaceutical industry registered one of its best years in terms of FDA approval in 2020 with the authorization of 53 drugs. Of these, 40 were new chemical entities and 13 were biologics, including ten mAbs and two antibody-drug conjugates, a rising format in the antibody space. The entire pharma sector, including new drug approvals and manufacturing, is pivoting towards biologics, a segment that is set to overtake small molecules and account for 55% of all innovative drug sales by 2027. By then, it is estimated that mAbs will dominate biologics sales.
Modern small molecule drug discovery typically takes between 10 to 15 years and 90% of drug candidates still fail in clinical trials. For antibody therapeutics, the average clinical development and regulatory review timeline are around eight years with a 22% success rate from Phase 1 to approval.
However, as the structure and design of druggable targets and candidate molecules becomes more complex, and given increasingly stringent regulatory standards, the challenges to generating highly specific and novel biological molecules continue to expand. A 2021 survey of industry professionals revealed that antibody drug discovery processes were the biggest challenge in antibody engineering and therapeutic R&D even as novel targets for antibodies emerge as the predominant development theme.
As a result, sophisticated technologies are playing a critical role in R&D pipelines in augmenting sensitivity, specificity, and flexibility and accelerating development and approval timelines.
There are two main approaches to antibody discovery — animal immunization and in vitro surface display techniques.
Antibodies derived from animal immunization have the advantage of having undergone in vivo affinity maturation, which enables the development of high-affinity clones from initial low-affinity germline antibodies. However, there are several drawbacks to this approach including the requirement for antigens to be immunogenic and concerns about immunodominance and animal ethics. Moreover, there have also been several reports that hybridoma technology used to produce murine-derived mAbs, for instance, has limited therapeutic efficacy and could result in undesirable allergic reactions.
In vitro surface display techniques, including the development of antibody phage display libraries, emerged as an alternative to the traditional hybridoma technology. In this approach, a library of antibodies, either constructed or purchased, are screened in vitro for binding to a target antigen. A library of antibodies (constructed using molecular techniques or purchased) can be screened in vitro for binding to a target antigen. Libraries of varying size and diversity can be generated using phage display or other combinatorial methods, with larger, more diverse libraries are more likely to produce mAbs with the highest affinity and specificity for the target antigen. Though this approach has many advantages over the traditional technique, by sacrificing the continuous evolution process of somatic hypermutation, it ends up discretizing the affinity maturation process.
Each of these approaches has its own advantages and limitations. Moreover, most antibody lead candidates identified using either method of discovery require additional protein engineering before they can be considered suitable for therapeutic use. As a result, improving the antibody discovery process will require harnessing the combined benefits of both technologies. More importantly, the increasing availability of large datasets, antibody sequences, structures, biophysical properties, etc., and the development of advanced computational methods, which improve antibody design and drug interaction predictions, is opening up a new era in biopharmaceutical informatics.
A 2017 article on Biopharmaceutical Informatics: supporting biologic drug development via molecular modelling and informatics formally introduced it as an interdisciplinary field combining knowledge-based informatics approaches (e.g. database creation, curation, and mining) and physics-based molecular modelling and simulation tools (from computational biophysics) that would advance the efficient and cost-effective translation of biologic drug candidates into drug products.
Since then, significant scientific and computational advances have been specific to antibody-based biologic drug candidates.
Take, for instance, the selection and screening process of phage antibody libraries, which has traditionally been a laborious and expensive bottleneck to the expanded use of mAbs. Here, innovations in constructing display libraries, phage, and yeast, have helped reduce many of the challenges associated with the downstream development of therapeutic leads. The shift to NGS-based screening from standard screening methods has resulted in dramatic improvements in throughput as well as a significant drop in the working time of phage antibody library discovery platforms.
Even though high-throughput mAbs discovery phage libraries represent a significant advance, the primary application is still only to increase affinity or specificity to the target antigen. Additional engineering is often required in the lead-candidate optimization stage, a time and cost-intensive process in the preclinical discovery and development cycle where multiple parameters, including expression levels, viscosity, pharmacokinetics, solubility, and immunogenicity, have to be addressed in parallel.
Today deep-learning models are capable of predicting molecular phenotype from sequence data and deep learning-based approaches are able to significantly reduce time, cost, and downstream risk by efficiently identifying the most developable antibody molecule.
At the same time, there is also growing interest in a new class of innovations around in silico engineering and design of antibody candidates. Though still very much in the research stage, In contrast to traditional methods of candidate generation such as hybridoma or phage display, in silico pipelines hold the potential for cheaper and faster drug development with deep-learning-based approaches demonstrating great success in addressing some of the key challenges of computational antibody design.
Even as computational antibody design becomes a reality, there are still several challenges that have to be addressed across in vivo, in vitro, and in silico approaches to antibody discovery and development. There is a critical need for high-throughput experimental techniques that can work with low quantities of protein at the antibody discovery stage, especially in determining in vivo properties that may lack key experimental data. At the same time, the systematic generation of high-quality data from these techniques will help enhance rational design and computationally de-risk candidate molecules.
Despite significant advances in deep learning applications in antibody development, there are still several data-related challenges that need to be addressed. These approaches have consistently demonstrated superior performance in big data domains. However, given the limited availability of antibody structure data, more advanced transfer learning approaches may be required to address the data gap.
In the meantime, the number of novel antibody therapeutics undergoing a first regulatory review reached a record level in 2021. However, any chance of setting an annual record for approvals was hampered by deficiencies in some applications. Perhaps, advancements in biopharmaceutical informatics will be able to mitigate some of the factors currently limiting the success rate of antibody drug discovery.