| import itertools | |
| import numpy as np | |
| import pandas as pd | |
| from Bio.PDB import PDBParser, PPBuilder | |
| from common import RES, EXT, abc | |
| from structure_utils import alphafold_model, correspondence, kabsch, tm_score | |
| AF = EXT / "alphafold" | |
| A = abc().set_index("gene") | |
| def ca_chain(gene): | |
| s = PDBParser(QUIET=True).get_structure(gene, alphafold_model(A.uniprot[gene], AF)) | |
| res = [r for r in s[0].get_residues() if "CA" in r] | |
| seq = "".join(str(pp.get_sequence()) for pp in PPBuilder().build_peptides(s[0])) | |
| return seq, np.array([r["CA"].coord for r in res], dtype=float) | |
| chains = {g: ca_chain(g) for g in A.index} | |
| rows = [] | |
| for g1, g2 in itertools.combinations(A.index, 2): | |
| (s1, x1), (s2, x2) = chains[g1], chains[g2] | |
| i1, i2 = correspondence(s1, s2) | |
| L = min(len(s1), len(s2)) | |
| d0 = 1.24 * (L - 15) ** (1 / 3) - 1.8 | |
| R, pc, qc = kabsch(x1[i1], x2[i2]) | |
| d = np.linalg.norm((x1[i1] - pc) @ R + qc - x2[i2], axis=1) | |
| tm_single = np.sum(1 / (1 + (d / d0) ** 2)) / L | |
| tm_opt, rmsd = tm_score(x1[i1], x2[i2], L) | |
| rows.append((g1, g2, A.subfamily[g1] == A.subfamily[g2], round(tm_single, 4), round(tm_opt, 4), round(rmsd, 2), len(i1), L)) | |
| out = pd.DataFrame(rows, columns=["gene1", "gene2", "same_subfamily", "tmscore_single_superposition", "tmscore", "rmsd_core", | |
| "aligned", "L"]) | |
| out.to_csv(RES / "abc_struct_sim.csv", index=False) | |
| par = pd.read_csv(RES / "abc_paralog_pairs.csv") if (RES / "abc_paralog_pairs.csv").exists() else None | |
| pd.set_option("display.width", 200) | |
| if par is not None: | |
| keys = {frozenset(p) for p in par[["gene1", "gene2"]].itertuples(index=False)} | |
| out["paralog"] = [frozenset((a, b)) in keys for a, b in zip(out.gene1, out.gene2)] | |
| out.to_csv(RES / "abc_struct_sim.csv", index=False) | |
| from scipy import stats | |
| I = pd.read_csv(RES / "abc_identity.csv", index_col=0) | |
| genes = list(A.index) | |
| n = len(genes) | |
| T = pd.DataFrame(np.nan, index=genes, columns=genes) | |
| for g1, g2, tm in zip(out.gene1, out.gene2, out.tmscore): | |
| T.loc[g1, g2] = T.loc[g2, g1] = tm | |
| iu = np.triu_indices(n, 1) | |
| Iv, Tv = I.loc[genes, genes].values, T.values | |
| obs = stats.spearmanr(Iv[iu], Tv[iu]).statistic | |
| rng = np.random.default_rng(0) | |
| B = 10000 | |
| null = np.array([stats.spearmanr(Iv[iu], Tv[np.ix_(p, p)][iu]).statistic for p in (rng.permutation(n) for _ in range(B))]) | |
| p_mantel = (np.sum(np.abs(null) >= abs(obs)) + 1) / (B + 1) | |
| pd.DataFrame([("identity vs TM-score, Mantel", obs, p_mantel, len(iu[0]), B)], | |
| columns=["test", "rho", "p_gene_label_permutation", "pairs", "permutations"]).to_csv( | |
| RES / "abc_identity_tm_mantel.csv", index=False) | |