CH103 Antibody Lineage - Affinity Maturation
A TMAP of the CH103 broadly-neutralizing antibody lineage (459 heavy-chain GenBank records, KC575845-KC576303) plus the published 363-nt VHDJH unmutated common ancestor (UCA) from Liao et al. 2013. Each record is aligned to the UCA by best ungapped placement, mutation burden is counted as unambiguous nucleotide mismatches, and sequences are represented as nucleotide 5-mers under a Jaccard-metric TMAP. Tree hops from the UCA node correlate with mutation burden (Spearman 0.253), and the graph-local audit shows TMAP edges connect sequences far more similar than chance: mean edge nt-mismatch delta is 4.58 versus 16.38 for random pairs (p < 0.001), and 91.6% of dated edges connect sequences from the same sampling week versus near-total mixing at random.
How it was made
from tmap import TMAP
# Nucleotide 5-mer sets per sequence (UCA reference + 459 observed records)
model = TMAP(
metric="jaccard",
n_neighbors=20,
n_permutations=256,
kc=80,
seed=42,
minhash_seed=42,
).fit(kmer_sets)
# Tree hops from the UCA (root) track affinity maturation
depth = [d for _, _, d in model.tree_.bfs(root=0)]
viz = model.to_tmapviz()
viz.add_color_layout("nt mismatches from UCA", mutation_burden, color="viridis")
viz.add_color_layout("Tree hops from UCA", depth, color="magma")
viz.write_html("ch103_lineage.html")