Benchmarking tree and ancestral sequence inference for B cell receptor sequences

Abstract

B cell receptor sequences evolve during affinity maturation according to a Darwinian process of mutation and selection. Phylogenetic tools are used extensively to reconstruct ancestral sequences and phylogenetic trees from affinity-matured sequences. In addition to using general-purpose phylogenetic methods, researchers have developed new tools to accommodate the special features of B cell sequence evolution. However, the performance of classical phylogenetic techniques in the presence of B cell-specific features is not well understood, nor how much the newer generation of B cell specific tools represent an improvement over classical methods. In this paper we benchmark the performance of classical phylogenetic and new B cell-specific tools when applied to B cell receptor sequences simulated from a forward-time model of B cell receptor affinity maturation towards a mature receptor. We show that the currently used tools vary substantially in terms of tree structure and ancestral sequence inference accuracy. Furthermore, we show that there are still large performance gains to be achieved by modeling the special mutation process of B cell receptors. These conclusions are further strengthened with real data using the rules of isotype switching to count possible violations within each inferred phylogeny.

Using B cell receptor lineage structures to predict affinity

“…For the metrics in this paper, that space is constructed from sequences and trees and is vast, complicated, and high-dimensional. In order to measure performance, we began with a previously-described simulation framework [ 15 ], extending it to allow a more comprehensive variation of parameters, and rewriting to optimize for speed. We then performed scans across all reasonably plausible values of parameters that could affect performance.…”

“…Since in this paper we focus on affinity maturation rather than VDJ rearrangement, we refer to that paper for all details on its implementation and validation. This naive sequence is then passed to the package [ 15 ] for GC reaction simulation. For this paper we have extended the original software by adding a number of new parameters to allow for more comprehensive variation, as well as optimizing for speed.…”

“…The correspondence between nucleotide sequence and affinity is by default determined by amino acid hamming distance to target sequence. For a detailed description refer to the supplement of [ 15 ], but the gist is that at each generation a limiting amount of antigen is apportioned among the cells based on their affinity (inverse dissociation constant) using equations of chemical binding equilibrium, and each cell’s λ is calculated based on its acquired antigen and the carrying capacity. The net result is a monotonic increase in mean number of offspring as a cell’s sequence draws closer to a target sequence.…”

Echidna: integrated simulations of single-cell immune receptor repertoires and transcriptomes

“…While this produces extremely diverse repertoires, it ignores many of the complexities not immediately observable underlying actual repertoire diversity (Marcou et al, 2018;Greiff et al, 2017;Slabodkin et al, 2021). We, therefore, have introduced the ability for more sophisticated repertoire simulations involving VAEs (Davidsen et al, 2019;Friedensohn et al, 2020), which under default parameters were trained on publicly available data of naive B and T cell repertoires. Importantly, we have additionally included the option for users to supply their own repertoires as training data to the VAE, thereby enabling the simulation of a repertoire resembling specialized experimental conditions.…”

“…Interpreting such datasets, however, remains challenging as the accompanying computational pipelines and software are still in their infancy (Yermanos, Agrafiotis, et al, 2021b;Borcherding et al, 2020;Sturm et al, 2020). Although multiple tools have been developed to simulate immune receptors and single-cell transcriptomes (Marcou et al, 2018;Weber et al, 2020;Yermanos et al, 2017;Davidsen et al, 2019;Safonova et al, 2015), these represent separate platforms and thus there remains a lack of software capable of simulating scSeq data of immune repertoires and transcriptomes. We therefore developed Echidna, an R package that simulates immune repertoires and their corresponding transcriptomes.…”