Designer Antibodies

Over the past few months, we have discussed topics such as bioinformatics, protein structure, crystallography, machine learning, and ELISA testing. All of these have something important in common: they are fundamental to our specialty at Macromoltek: in silico antibody design. During phase 1 of “Project Blog,” all of our posts have built toward the question, “How are antibodies computationally designed?”

Target Discovery

In order to begin designing a new antibody, you first need to find a target, or antigen. Teams of medical researchers spend their careers searching for antigens involved in an endless variety of diseases. Once a molecule is known to be implicated in the development of a disease, it becomes a target candidate. An important constraint in target discovery is that the molecule must also be “accessible” to antibodies. If the target is hidden within the cell, an antibody will not be able to bind it, and therefore cannot disable it or mark it for destruction. Because they are necessarily exposed to the outside of the cell, surface proteins are often good candidates for antibody therapies. One such example of a viable target is Programmed Cell Death 1 (PD-1) Protein, recently discussed on our Instagram page. This protein prevents our native antibodies from eliminating our healthy cells, but cancer cells can utilize it to mask themselves from the immune system. By designing an antibody to target this protein, we can “unmask” the cancer cell so it can no longer avoid detection!

The In Vivo Process

Once a target has been discovered, the design process can begin. Traditionally, antibodies are created in vivo by injecting the target protein into a lab animal. The animal’s immune system will produce antibodies against the disease, which are then harvested and purified. This method is great for producing antibodies for laboratory use, but because the antibodies originated in another species, they aren’t readily usable as treatments. Instead of letting them do their job, the human body will perceive their foreign-looking amino acid sequences as a threat and try to interfere! This problem can be overcome by adjusting the sequence to appear more human — or humanizing the antibody — but this may also change its structure. This, in turn, might cause it to be a less effective treatment for the disease. There are a number of ways in which changing the structure might harm the quality of the antibody; the antibody may bind to more than just the target (non-specific binding), to itself (aggregation), or to lose binding altogether! If any of these occur, the antibody has to be designed and fabricated all over again. This method takes years — or even decades — to produce a viable therapeutic. The aim of in silico design is to utilize bioinformatics and computational modeling to skip all of these steps — saving years in the process.

The In Silico Process

Any in silico design has to start with some kind of structural information. This is why crystallography is so important. Though there are ways to predict protein structure without using crystallography, such models are significantly less reliable. The target structure is used as a starting point to produce antibodies — much like in a living system — but here, the “organism” is a complicated algorithm which simulates the creation of a new antibody. This algorithm takes all of the different features of an organic antibody into account to create an effective treatment from scratch! Of course, the number of possible sequence combinations is nearly limitless, so special methods, such as simulated annealing, are used to optimize the design. Once a sufficient number of sequences have been generated, they’re tested in silico, reviewed by experts, and narrowed down to a small pool of the best candidates.These are then grown in a lab and subjected to further testing in vitro. The in vitro data informs changes to the algorithm, completing a feedback loop that progressively improves in silico antibody results.

The laboratory component is necessary in both the in silico and traditional in vivo methods, to ensure good performance. Assays — such as ELISA, which you may remember from a previous post — provide key information regarding protein expression, folding, and proper antigen binding.

The In Silico Advantage

The most obvious advantage over traditional methods is the time to completion. The process of harvesting, isolating, determining sequence, and finally humanizing animal antibodies takes years. Every problem means the sequence requires adjustment, the antibodies must be harvested again, or a new target needs to be selected — and the process has to start anew. These major in vivo roadblocks are nothing to in silico design. New models can be created in a matter of hours, and honed into a functional antibody product with a few weeks of laboratory testing. Moreover, antibodies can be designed against antigens far too toxic to inject into a living creature. Consider, for example, cobra venom — an animal would be unable to generate antibodies before the symptoms became lethal. The in vivo strategy would fail completely.

Of course, in silico design is not without its flaws. The antibodies produced are only as good as the algorithm that produces them. There are aspects of antibody binding that are not yet understood. In the past decades, the field of antibody research has seen a massive proliferation of new data describing antibody behavior and conformation. Using this data, we seek to actively improve the speed, efficiency, and accuracy of our antibody design process. Macromoltek continues to revolutionize the pharmaceutical industry by providing efficient ways to fight life-threatening diseases and solve real-world problems.

Looking for more information about Macromoltek, Inc? Visit our website at
Interested in molecular simulations, biological art, or learning more about molecules? Subscribe to our Twitter and Instagram!