Abundance Aware Phylogenetic Inference for Dense Biological Sampling

15 concepts6 papers1 proposal

When reconstructing the evolutionary history of rapidly mutating systems—such as B-cell receptors (BCRs) in adaptive immunity, expanding clones in a tumor, or rapidly duplicating gene families—traditional phylogenetic methods run into a fundamental paradox. Intuitively, having more data should make tree reconstruction easier. However, dense biological sampling often yields the opposite result: a massive influx of highly similar or identical sequences that overwhelms standard comparative sequence analysis.

When mutations are sparse relative to the sheer volume of sampled sequences, traditional sequence-only methods suffer from parsimony degeneracy. They generate thousands of distinct tree topologies that explain the mutations equally well. Without an objective way to break these ties, the true biological history remains obscured.

To resolve this ambiguity, a powerful paradigm has emerged: leveraging non-sequence signals—such as genotype abundance and structural context—to guide tree selection and expose the true evolutionary trajectory.

Genotype Abundance as an Evolutionary Tie-Breaker

In a rapidly proliferating population, the frequency of a specific genetic variant is not random noise; it is a direct reflection of its evolutionary success. If a particular genotype is sampled hundreds of times while its mutational neighbors are sampled only once or twice, that sequence represents a highly expanded lineage.

Standard phylogenetic models ignore this count data, treating unique sequences as binary indicators of presence or absence. To exploit this hidden signal, DeWitt et al. introduced an empirical Bayes framework called GCtree. Instead of relying solely on sequence mutations, this approach uses sequence abundance data as an explicit evolutionary signal.

The intuition is straightforward:

  1. Generate Candidate Trees: Traditional parsimony methods are first used to construct the initial forest of equally optimal trees based purely on sequence mutations.
  2. Rank via Abundance: An abundance model evaluates how likely each tree topology is to have generated the observed genotype counts, assuming that highly abundant sequences are more likely to act as persistent ancestral nodes that spawn further mutations.

By ranking the degenerate trees using this probabilistic framework, the authors demonstrated a significant reduction in topological errors (measured by Robinson-Foulds distances), particularly in scenarios plagued by parsimony degeneracy.

Scaling the Search Space with History Graphs

While ranking an initial set of parsimonious trees works well for moderate datasets, dense sampling often causes the number of equally parsimonious trees to explode exponentially. Enumerating and scoring every single tree individually quickly becomes computationally impossible.

To scale abundance-aware inference to massive datasets, Dumm et al. introduced a compact data structure known as a history sDAG (shared directed acyclic graph).

Traditional Enumeration:
[Tree 1]  [Tree 2]  [Tree 3] ... [Tree 10,000] -> Evaluated Individually (Slow)

History sDAG Framework:
   [Shared Ancestral Subgraph]
         /            \
  [Sub-history A]   [Sub-history B]          -> Evaluated Graph-wide (Fast)

Instead of treating every tree as an independent entity, the history sDAG collapses identical sub-histories and shared evolutionary pathways into a single graph. By applying context-sensitive and abundance-aware scoring functions directly over this graph structure, the authors made it possible to efficiently store and explore a vastly larger space of parsimonious trees, unlocking highly accurate lineage tracking for densely sampled BCR sequences in germinal centers.

Expanding Context: Species Trees and Cellular Constraints

The core philosophy of this research direction—using external structural context to resolve sequence-level ambiguity—extends beyond genotype abundance. Depending on the biological scale, different types of contextual constraints can act as evolutionary tie-breakers.

Macro-Evolutionary Context: Species Trees

When reconstructing the history of a gene family, sequence-only trees often fail because they cannot distinguish between true mutations and macro-evolutionary events like horizontal gene transfer (HGT), gene duplications, or gene losses.

To fix this, Bansal et al. developed TreeFix-DTL. This method balances sequence data with the known species tree, explicitly modeling duplication, transfer, and loss (DTL) costs. By penalizing gene trees that require highly improbable macro-evolutionary events, the authors reduced topological errors by an average of 71.1% compared to standard maximum likelihood methods. Similarly, Zheng et al. demonstrated that integrating multiple phylogenetic algorithms to embed gene family histories directly within a species tree context yields significantly higher log-likelihoods and more biologically plausible topologies.

Cellular and Phenotypic Context

At the single-cell level, physical and phenotypic constraints provide the necessary context to resolve degenerate trees:

  • Tumor Phylogenetics: When inferring tumor evolution from ultra-low coverage single-cell DNA sequencing, sequence data alone is often too sparse. Weber et al. (2023) developed Phertilizer, a method that integrates single-nucleotide variants (SNVs) with copy number aberrations (CNAs). The structural changes from CNAs provide a robust framework that guides the placement of cells, leading to superior genotype accuracy and ancestral recall.
  • Immune Cell Isotypes: In B-cell lineage trees, somatic hypermutation occurs alongside class switch recombination—a unidirectional phenotypic shift where B cells change their antibody isotype (e.g., from IgM to IgG). Weber et al. (2024) introduced TRIBAL, an algorithm that models these isotype transitions simultaneously with mutations. Because isotype switching is biologically directional, it acts as a natural constraint, eliminating tree topologies that violate these cellular rules and drastically improving ancestral sequence reconstruction.

Conclusion

Whether dealing with the proliferation counts of B-cell receptors, the copy number profiles of tumor clones, or the embedding constraints of a species tree, the underlying principle remains the same: sequence data alone is not enough when sampling is dense. By treating abundance and biological context as primary evolutionary signals rather than secondary metadata, modern phylogenetic frameworks successfully turn the challenge of dense biological sampling into an informational advantage.

History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_handling_dense_biological_sampling_b22cade6,
  title = {handling dense biological sampling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/handling-dense-biological-sampling/rdr_b22cade6d5f642e3b177b1a850990174},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_handling_dense_biological_sampling_b22cade6_v1,
  title = {handling dense biological sampling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/handling-dense-biological-sampling/rdr_b22cade6d5f642e3b177b1a850990174/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake