File size: 4,242 Bytes
638ea38 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 | import itertools
import numpy as np
import pandas as pd
from scipy import stats
from statsmodels.stats.multitest import multipletests
from common import RES, OX, AGENTS, hom, axes, centre, effects, agent_effects
Z = hom()
ag, ox, os_, ref = axes(Z.columns)
Zc = centre(Z)
E_ox, E_os = effects(Zc, ox, os_)
cd = (E_ox - E_os).abs()
A = agent_effects(Zc, ag)
A.round(4).rename_axis("gene").to_csv(RES / "agent_effects.csv")
stress_cols = np.where(ox | os_)[0]
R = Zc.iloc[:, stress_cols].corr(method="spearman")
lab = ag[stress_cols]
pairs = []
for i, j in itertools.combinations(range(len(stress_cols)), 2):
same = lab[i] == lab[j]
axis_i = "oxidative" if lab[i] in OX else "osmotic"
axis_j = "oxidative" if lab[j] in OX else "osmotic"
kind = "same agent" if same else ("same axis, different agent" if axis_i == axis_j else "different axis")
pairs.append((lab[i], lab[j], kind, axis_i if axis_i == axis_j else "mixed", R.iloc[i, j]))
P = pd.DataFrame(pairs, columns=["agent1", "agent2", "kind", "axis", "rho"])
P.to_csv(RES / "condition_pair_correlations.csv", index=False)
names = [f"{lab[i]}|{Zc.columns[stress_cols[i]].split(':')[0]}" for i in range(len(stress_cols))]
pd.DataFrame(R.values, index=names, columns=names).round(4).rename_axis("condition").to_csv(
RES / "condition_correlation_matrix.csv")
anova = []
for axis, members in (("oxidative", ["h2o2", "paraquat", "4nqo"]), ("osmotic", ["nacl", "sorbitol"])):
X = [Zc.loc[:, ag == a].values for a in members]
n = np.stack([np.isfinite(x).sum(axis=1) for x in X], axis=1)
tot = np.stack([np.nansum(x, axis=1) for x in X], axis=1)
ok = (n > 0).all(axis=1)
with np.errstate(invalid="ignore", divide="ignore"):
means = tot / n
grand = tot.sum(axis=1) / n.sum(axis=1)
ssb = np.nansum(n * (means - grand[:, None]) ** 2, axis=1)
ssw = sum(np.nansum((x - m[:, None]) ** 2, axis=1) for x, m in zip(X, means.T))
dfb, dfw = len(members) - 1, n.sum(axis=1) - len(members)
F = np.where(ok, (ssb / dfb) / (ssw / dfw), np.nan)
eta = np.where(ok, ssb / (ssb + ssw), np.nan)
p = stats.f.sf(F, dfb, dfw)
padj = np.full(len(p), np.nan)
padj[ok] = multipletests(p[ok], method="fdr_bh")[1]
anova.append((axis, len(members), sum(x.shape[1] for x in X), int(ok.sum()), float(np.nanmedian(eta)),
float(np.mean(padj[ok] < 0.05)), float(np.mean(p[ok] < 0.05))))
if axis == "oxidative":
het = pd.DataFrame({"F": F, "p": p, "p_bh": padj, "eta2_agent": eta}, index=Zc.index)
het.round(5).rename_axis("gene").to_csv(RES / "oxidative_agent_heterogeneity.csv")
anova = pd.DataFrame(anova, columns=["axis", "agents", "conditions", "genes", "median_eta2_agent", "frac_bh_lt_0.05", "frac_p_lt_0.05"])
anova.to_csv(RES / "agent_heterogeneity_summary.csv", index=False)
std = A / A.std()
S_ox = std[["h2o2", "paraquat", "4nqo"]].mean(axis=1)
S_os = std[["nacl", "sorbitol"]].mean(axis=1)
cd_bal = (S_ox - S_os).abs()
pd.DataFrame({"ox_balanced": S_ox, "osm_balanced": S_os, "cd_balanced": cd_bal}).round(4).rename_axis("gene").to_csv(
RES / "agent_balanced_effects.csv")
loao = []
for drop in AGENTS:
keep_ox = [a for a in ("h2o2", "paraquat", "4nqo") if a != drop]
keep_os = [a for a in ("nacl", "sorbitol") if a != drop]
d2 = (std[keep_ox].mean(axis=1) - std[keep_os].mean(axis=1)).abs()
keep = ag != drop
a2, b2 = effects(Zc, ox & keep, os_ & keep)
loao.append((drop, stats.spearmanr(cd, (a2 - b2).abs(), nan_policy="omit").statistic,
stats.spearmanr(cd_bal, d2, nan_policy="omit").statistic))
L = pd.DataFrame(loao, columns=["omitted_agent", "rho_pooled", "rho_agent_balanced"])
L.to_csv(RES / "leave_one_agent_out.csv", index=False)
pc = {}
for a, b in itertools.product(("h2o2", "paraquat", "4nqo"), ("nacl", "sorbitol")):
pc[f"{a}|{b}"] = (A[a] - A[b]).abs()
PC = pd.DataFrame(pc)
PC.round(4).rename_axis("gene").to_csv(RES / "agent_pair_cd.csv")
top = PC.rank(ascending=False) <= np.ceil(0.1 * len(PC))
k = top.sum(axis=1)
pd.DataFrame({"n_pairs_top_decile": k}).rename_axis("gene").to_csv(RES / "agent_robust_cd.csv")
|