"Wie in der Probe­blutung zei­gen die [Spaltstellen-]­Anti­körper eine sehr gute Spe­zi­fi­tät (…). So­weit sehen die Blots sehr gut aus. Es gibt auch keine Kreuz­re­aktio­nen mit E-Coli Pro­teinen bei der in vitro Trans­lation."

A customer's quote

The Art of Selecting an Epitope

The two figures vividly illustrate that selecting an epitope solely based on software algorithms is an essentially impossible task.

Three different prediction programs, taken as an example, suggest a variety of different regions that might be suitable as an antigenic peptide. In contrast, the epitope mapping experiment (anonymized data from a recent customer's project) reveals that a rabbit immunized with the full-length protein choses another set of epitopes, that does not match any of the software suggested candidates.

Shown are the results from a representative partial sequence.

Comparison of epitope prediction methods versus experimental epitope mapping
Epitope prediction versus epitope mapping (~300 amino acid partial sequence)
Antibody epitope mapping using a peptide microarray
Antibody epitope mapping using a peptide microarray (~300 amino acid partial sequence shown in the left panel)
Another vivid example of epitope mapping using a peptide microarray

Specific cross-reactivity with unrelated proteins

Typically, antibody epitopes are about 5-8 amino acids in length. The smallest epitope, which we have mapped so far, comprised only three amino acids and part of the adjacent peptide bond. This poses a possible problem for both peptide and monoclonal antibodies: For statistical reasons, identical (or closely related) sequence stretches of a relevant size may occur in various unrelated proteins. Indeed, identities of - let's say five - consecutive amino acids are essentially inevitable.

This may be sufficient to elicit a specific (!) cross-reactivity with an protein otherwise unrelated in sequence and function. "Specific" - because this antibody was raised (and purified) against this sequence and hence the reactivity is legitimate and desired with the "correct" antigen but disastrous with the "wrong" protein. Therefore, it is mandatory to put a special focus on this fact when designing a peptide antigen.

ambiguous sequence
Ambiguous peptide sequences should be taboo: Possible x-reactivity of anti-albumin antibodies with LMBR1-DCP2 and vice versa

Conclusions:

Epitope selection requires more than just applying a prediction algorithm to the protein sequence.

Therefore, our proprietary work-flow comprises a wide variety of different aspects and methods, including (where available) 3D models and other background information, such as species comparison, secondary structure, domain structure, ...

These considerations lead to a pre-selection of different epitope sequences, which we will discuss in detail with our customers to derive a common educated decision.