Maximum-Entropy Models of Sequenced Immune Repertoires Predict Antigen-Antibody Affinity

The immune system has developed a number of distinct complex mechanisms to shape and control the antibody repertoire. One of these mechanisms, the affinity maturation process, works in an evolutionary-like fashion: after binding to a foreign molecule, the antibody-producing B-cells exhibit a high-frequency mutation rate in the genome region that codes for the antibody active site. Eventually, cells that produce antibodies with higher affinity for their cognate antigen are selected and clonally expanded. Here, we propose a new statistical approach based on maximum entropy modeling in which a scoring function related to the binding affinity of antibodies against a specific antigen is inferred from a sample of sequences of the immune repertoire of an individual. We use our inference strategy to infer a statistical model on a data set obtained by sequencing a fairly large portion of the immune repertoire of an HIV-1 infected patient. The Pearson correlation coefficient between our scoring function and the IC50 neutralization titer measured on 30 different antibodies of known sequence is as high as 0.77 (p-value 10−6), outperforming other sequence- and structure-based models.Author Summary: Affinity maturation is a very complex biological process which enables activated B-cells to produce antibodies with increased affinity for a given antigen. Once B-cells begin to proliferate, each of the progeny cells introduces mutations in the antigen binding region in order to explore different affinities for the antigen. Selection rounds occurring in the so-called germinal centers in lymph nodes and spleen prune out poorly binding receptors and clonally expand good binders. Thanks to high-throughput sequencing techniques it is now possible to have access to a fairly representative sample (of the order of 105 to 106 sequences) of the immune repertoire of a given individual. Our approach is to first exploit this large amount of sequence data to infer a statistical model for the sequenced portion of the immune repertoire, and then to use the inferred probability of this model as a score when predicting the neutralization power of a given antibody sequence for the antigen of interest. The results we obtained on a specific data set of sequences of an HIV-1 patient show that our score correlates very well with experimentally assessed neutralization power of specific antibodies of known sequence. The performance of the method crucially relies on the ability of our model to account for long-range intragenic epistatic interactions between residues along the whole antibody chain.