Survey Outcome on Immunogenicity Risk Assessment Tools for Biotherapeutics: an Insight into Consensus on Methods, Application, and Utility in Drug Development.

A survey conducted by the Therapeutic Product Immunogenicity (TPI) community within the American Association of Pharmaceutical Scientists (AAPS) posed questions to the participants on their immunogenicity risk assessment strategies prior to clinical development. The survey was conducted in 2 phases spanning 5 years, and queried information about in silico algorithms and in vitro assay formats for immunogenicity risk assessments and how the data were used to inform early developability effort in discovery, chemistry, manufacturing and control (CMC), and non-clinical stages of development. The key findings representing the trends from a majority of the participants included the use of high throughput in silico algorithms, human immune cell-based assays, and proteomics based outputs, as well as specialized assays when therapeutic mechanism of action could impact risk assessment. Additional insights into the CMC-related risks could also be gathered with the same tools to inform future process development and de-risk critical quality attributes with uncertain and unknown risks. The use of the outputs beyond supporting early development activities was also noted with participants utilizing the risk assessments to drive their clinical strategy and streamline bioanalysis.

Assay format diversity in pre-clinical immunogenicity risk assessment: Toward a possible harmonization of antigenicity assays.

A major impediment to successful use of therapeutic protein drugs is their ability to induce anti-drug antibodies (ADA) that can alter treatment efficacy and safety in a significant number of patients. To this aim, in silico, in vitro, and in vivo tools have been developed to assess sequence and other liabilities contributing to ADA development at different stages of the immune response. However, variability exists between similar assays developed by different investigators due to the complexity of assays, a degree of uncertainty about the underlying science, and their intended use. The impact of protocol variations on the outcome of the assays, i.e., on the immunogenicity risk assigned to a given drug candidate, cannot always be precisely assessed. Here, the Non-Clinical Immunogenicity Risk Assessment working group of the European Immunogenicity Platform (EIP) reviews currently used assays and protocols and discusses feasibility and next steps toward harmonization and standardization.

Improved prediction of HLA antigen presentation hotspots: Applications for immunogenicity risk assessment of therapeutic proteins.

Immunogenicity risk assessment is a critical element in protein drug development. Currently, the risk assessment is most often performed using MHC-associated peptide proteomics (MAPPs) and/or T-cell activation assays. However, this is a highly costly procedure that encompasses limited sensitivity imposed by sample sizes, the MHC repertoire of the tested donor cohort and the experimental procedures applied. Recent work has suggested that these techniques could be complemented by accurate, high-throughput and cost-effective prediction of in silico models. However, this work covered a very limited set of therapeutic proteins and eluted ligand (EL) data. Here, we resolved these limitations by showcasing, in a broader setting, the versatility of in silico models for assessment of protein drug immunogenicity. A method for prediction of MHC class II antigen presentation was developed on the hereto largest available mass spectrometry (MS) HLA-DR EL data set. Using independent test sets, the performance of the method for prediction of HLA-DR antigen presentation hotspots was benchmarked. In particular, the method was showcased on a set of protein sequences including four therapeutic proteins and demonstrated to accurately predict the experimental MS hotspot regions at a significantly lower false-positive rate compared with other methods. This gain in performance was particularly pronounced when compared to the NetMHCIIpan-3.2 method trained on binding affinity data. These results suggest that in silico methods trained on MS HLA EL data can effectively and accurately be used to complement MAPPs assays for the risk assessment of protein drugs.

Post hoc assessment of the immunogenicity of bioengineered factor VIIa demonstrates the use of preclinical tools.

Immunogenicity is an important consideration in the licensure of a therapeutic protein because the development of neutralizing anti-drug antibodies (ADAs) can affect both safety and efficacy. Neoantigens introduced by bioengineering of a protein drug are a particular cause for concern. The development of a bioengineered recombinant factor VIIa (rFVIIa) analog was discontinued after phase 3 trials because of the development of ADAs. The unmodified parent molecule (rFVIIa), on the other hand, has been successfully used as a drug for more than two decades with no reports of immunogenicity in congenital hemophilia patients with inhibitors. We used computational and experimental methods to demonstrate that the observed ADAs could have been elicited by neoepitopes in the engineered protein. The human leukocyte antigen type of the patients who developed ADAs is consistent with this hypothesis of a neoepitope-driven immune response, a finding that might have implications for the preclinical screening of therapeutic protein analogs.

Definition of supertypes for HLA molecules using clustering of specificity matrices.

Major histocompatibility complex (MHC) proteins are encoded by extremely polymorphic genes and play a crucial role in immunity. However, not all genetically different MHC molecules are functionally different. Sette and Sidney (1999) have defined nine HLA class I supertypes and showed that with only nine main functional binding specificities it is possible to cover the binding properties of almost all known HLA class I molecules. Here we present a comprehensive study of the functional relationship between all HLA molecules with known specificities in a uniform and automated way. We have developed a novel method for clustering sequence motifs. We construct hidden Markov models for HLA class I molecules using a Gibbs sampling procedure and use the similarities among these to define clusters of specificities. These clusters are extensions of the previously suggested ones. We suggest splitting some of the alleles in the A1 supertype into a new A26 supertype, and some of the alleles in the B27 supertype into a new B39 supertype. Furthermore the B8 alleles may define their own supertype. We also use the published specificities for a number of HLA-DR types to define clusters with similar specificities. We report that the previously observed specificities of these class II molecules can be clustered into nine classes, which only partly correspond to the serological classification. We show that classification of HLA molecules may be done in a uniform and automated way. The definition of clusters allows for selection of representative HLA molecules that can cover the HLA specificity space better. This makes it possible to target most of the known HLA alleles with known specificities using only a few peptides, and may be used in construction of vaccines. Supplementary material is available at http://www.cbs.dtu.dk/researchgroups/immunology/supertypes.html.

Reliable prediction of T-cell epitopes using neural networks with novel sequence representations.

In this paper we describe an improved neural network method to predict T-cell class I epitopes. A novel input representation has been developed consisting of a combination of sparse encoding, Blosum encoding, and input derived from hidden Markov models. We demonstrate that the combination of several neural networks derived using different sequence-encoding schemes has a performance superior to neural networks derived using a single sequence-encoding scheme. The new method is shown to have a performance that is substantially higher than that of other methods. By use of mutual information calculations we show that peptides that bind to the HLA A*0204 complex display signal of higher order sequence correlations. Neural networks are ideally suited to integrate such higher order correlations when predicting the binding affinity. It is this feature combined with the use of several neural networks derived from different and novel sequence-encoding schemes and the ability of the neural network to be trained on data consisting of continuous binding affinities that gives the new method an improved performance. The difference in predictive performance between the neural network methods and that of the matrix-driven methods is found to be most significant for peptides that bind strongly to the HLA molecule, confirming that the signal of higher order sequence correlation is most strongly present in high-binding peptides. Finally, we use the method to predict T-cell epitopes for the genome of hepatitis C virus and discuss possible applications of the prediction method to guide the process of rational vaccine design.