5D Library › Polypharmacy Supplement Interactions, Chronotherapy Dosing Windows and Cytochrome Enzyme Competition · 中文
Chapter 7. The Seven Relations
A classification of ordered pairs, written in the Han
The oldest Chinese pharmacopoeia does not treat a drug as a thing with properties. It treats a pair of drugs as a thing with properties. The preface of the Shennong bencao jing sets out seven cases for what one substance does to another: acting alone, mutual reinforcement, one assisting another, one being restrained by another, one spoiling another, and mutual opposition (神农本草经辑注 1995). That is a classification of ordered pairs by the sign and the symmetry of their effect on each other, and it is formally the same object as an interaction matrix.
Source text
The seven relations. The preface of the Shennong bencao jing lists the cases as acting alone, mutual need, mutual use, mutual fear, mutual dislike and mutual opposition, together with the instruction that a prescriber must examine them before combining. Two features of the scheme are worth noting before any comparison is attempted. The first is that it is a scheme about pairs, not about single substances, so it is a relational classification and not a list of properties. The second is that two of its cases, one substance restraining another’s harm and one substance spoiling another’s benefit, are distinguished by whether what is affected is toxicity or efficacy, which is a distinction a model of absorption cannot make (神农本草经辑注 1995).
That second feature sets the limit of what the comparison can establish, and the program states the limit rather than working around it.
Calculation
# Chapter 7. p13: The seven relations, written as a matrix. The
# Shennong pharmacopoeia sorts what one substance does to another into
# seven cases: acting alone, mutual reinforcement, one assisting
# another, one being restrained by another, one spoiling another, and
# mutual opposition. That is a classification of ordered pairs by the
# sign and the symmetry of their effect on each other, which is exactly
# what the interaction model of this book produces. Each of the 55
# unordered pairs of the eleven items is scored twice, once in each
# direction, and sorted into the classical cases by the rule below.
# The model sees delivery, not toxicity, so the case in which one
# substance restrains another's harm cannot be told apart from the case
# in which it spoils another's benefit; the two are reported together.
ITEMS = (
("magnesium", 400.00, 24.305, "divalent", 4.0000),
("zinc", 25.00, 65.380, "divalent", 0.1500),
("calcium", 500.00, 40.080, "divalent", 5.0000),
("iron", 18.00, 55.850, "divalent", 0.0800),
("vitamin D3", 0.10, 384.600, "micelle", 0.0004),
("vitamin K2", 0.18, 444.700, "micelle", 0.0005),
("NAD precursor", 300.00, 255.200, "nrk", 1.0000),
("curcumin", 500.00, 368.400, "conjugate", 1.0000),
("piperine", 10.00, 285.300, "conjugate", 0.0500),
("theanine", 200.00, 174.200, "amino", 1.0000),
("caffeine", 100.00, 194.200, "cyp1a2", 0.5000),
)
ENH_MAX, ENH_K = 1.60, 0.020
BOOSTS = {("piperine", "curcumin")} # ordered: first raises second
THR = 0.05
def mmol(dose_mg, molar_mass):
return dose_mg / molar_mass
def delivery(i, partner):
name, dose, mm, site, km = ITEMS[i]
s = mmol(dose, mm)
alone = s / (km + s)
extra = 0.0
if partner is not None and ITEMS[partner][3] == site:
extra = mmol(ITEMS[partner][1], ITEMS[partner][2]) / ITEMS[partner][4]
out = s / (km * (1.0 + extra) + s)
if partner is not None and (ITEMS[partner][0], name) in BOOSTS:
p = mmol(ITEMS[partner][1], ITEMS[partner][2])
out *= 1.0 + ENH_MAX * p / (ENH_K + p)
return out / alone
def classify(a, b):
# a is what the second does to the first, b the other way round
up_a, up_b = a > THR, b > THR
dn_a, dn_b = a < -THR, b < -THR
if not (up_a or up_b or dn_a or dn_b):
return "acting alone"
if up_a and up_b:
return "mutual reinforcement"
if up_a or up_b:
# one of the two is raised, whatever happens to the other
return "one assisting another"
if dn_a and dn_b:
return "mutual opposition"
return "one restraining or spoiling"
ORDER = ("acting alone", "mutual reinforcement", "one assisting another",
"one restraining or spoiling", "mutual opposition")
if __name__ == "__main__":
n = len(ITEMS)
counts = {k: 0 for k in ORDER}
rows = []
for i in range(n):
for j in range(i + 1, n):
a = delivery(i, j) - 1.0
b = delivery(j, i) - 1.0
c = classify(a, b)
counts[c] += 1
rows.append((c, ITEMS[i][0], ITEMS[j][0], a, b))
print(" unordered pairs scored: {}".format(n * (n - 1) // 2))
print(" case pairs share")
for c in ORDER:
print(" {:<30}{:<8d}{:.4f}"
.format(c, counts[c], counts[c] / len(rows)))
print(" the pairs that are not acting alone:")
print(" case first second effects")
for c in ORDER[1:]:
for cc, a_name, b_name, a, b in rows:
if cc == c:
print(" {:<28}{:<13}{:<11}{:+.3f} {:+.3f}"
.format(c[:27], a_name[:12], b_name[:10], a, b))
strongest = min(rows, key=lambda r: min(r[3], r[4]))
print(" strongest opposition: {} and {} at {:+.3f} and {:+.3f}"
.format(strongest[1], strongest[2], strongest[3], strongest[4]))
helps = [r for r in rows if max(r[3], r[4]) > THR]
print(" pairs in which anything is raised at all: {}"
.format(len(helps)))
print(" so in this cupboard the classical cases are not evenly")
print(" filled: {:.4f} of pairs act alone and {:.4f} oppose"
.format(counts["acting alone"] / len(rows),
(counts["mutual opposition"]
+ counts["one restraining or spoiling"]) / len(rows)))
# unordered pairs scored: 55
# case pairs share
# acting alone 47 0.8545
# mutual reinforcement 0 0.0000
# one assisting another 1 0.0182
# one restraining or spoiling 0 0.0000
# mutual opposition 7 0.1273
# the pairs that are not acting alone:
# case first second effects
# one assisting another curcumin piperine +0.556 -0.444
# mutual opposition magnesium zinc -0.333 -0.537
# mutual opposition magnesium calcium -0.328 -0.541
# mutual opposition magnesium iron -0.441 -0.450
# mutual opposition zinc calcium -0.413 -0.422
# mutual opposition zinc iron -0.532 -0.336
# mutual opposition calcium iron -0.535 -0.332
# mutual opposition vitamin D3 vitamin K2 -0.329 -0.264
# strongest opposition: magnesium and calcium at -0.328 and -0.541
# pairs in which anything is raised at all: 1
# so in this cupboard the classical cases are not evenly
# filled: 0.8545 of pairs act alone and 0.1273 opposeOf the 55 unordered pairs, 47 act alone, a share of 0.8545. Seven are in mutual opposition, all of them among the four bulk and trace minerals and the two fat-soluble vitamins, and one is a case of one substance assisting another, which is curcumin and piperine. No pair in this cupboard is a case of mutual reinforcement, and only one pair involves anything being raised at all.
That distribution is itself a finding about modern supplement practice. The classical scheme has six cases for interaction and one for independence, and a modern cupboard fills the independence case 0.8545 of the time and two of the interaction cases the rest. The relations the Shennong preface is most concerned with, the ones where combining is actively good, are empty here. A Han prescriber would have described this cupboard as badly composed, not because the items are wrong but because almost nothing in it is chosen to work with anything else in it.
Four rules about the hour
The second classical source is narrower and more testable. The preface to Li Shizhen’s Bencao gangmu quotes an earlier rule of four cases about when to swallow something: a complaint above the diaphragm, after food; below the abdomen, before food; in the limbs and vessels, on an empty stomach and at dawn; in the marrow, on a full stomach and at night (李时珍 1982).
Source text
The dosing hours. The passage, quoted in the Bencao gangmu’s preface from a Jin dynasty source, assigns a time and a relation to food according to where in the body the complaint is. It is a rule about the route and the destination, not about the substance, which is what makes it comparable to a chronopharmacological argument at all: both claim that the same substance at a different hour does a different amount of good (李时珍 1982).
Calculation
# Chapter 7. p14: Four classical rules about when to swallow something,
# scored against the arithmetic. The preface to the Bencao gangmu quotes
# an earlier rule of four cases: a complaint above the diaphragm, take
# it after food; below the abdomen, before food; in the limbs and
# vessels, on an empty stomach and at dawn; in the marrow, on a full
# stomach and at night. Each case is given here the modern slot whose
# gut condition and clock hour the arithmetic of p11 would choose for
# the class of substance that case is about, and the two are compared.
# The agreement is counted; nothing is adjusted to improve it.
SLOTS = (
("after food, morning", 7.0, 18.0, 2.0),
("before food, midday", 12.0, 2.0, 1.7),
("empty stomach at dawn", 6.0, 0.5, 1.5),
("full stomach at night", 22.0, 8.0, 1.9),
)
CASES = (
# classical case, the modern class it maps to, needs, phase peak
("above the diaphragm", "fat-soluble vitamin", "fat", 6.0),
("below the abdomen", "mineral needing acid", "acid", 12.0),
("limbs and vessels", "trace mineral, empty", "acid", 6.0),
("marrow, at night", "calming mineral", "plain", 22.5),
)
FAT_HALF, PH_HALF, PH_HILL, PHASE_WIDTH = 9.0, 3.4, 2.4, 5.0
E = 2.718281828459045
def condition(needs, slot):
_, hour, fat_g, ph = slot
f = 1.0
if needs == "fat":
f *= (fat_g / (FAT_HALF + fat_g)) / (18.0 / (FAT_HALF + 18.0))
if needs == "acid":
f *= (1.0 / (1.0 + (ph / PH_HALF) ** PH_HILL)) \
/ (1.0 / (1.0 + (1.5 / PH_HALF) ** PH_HILL))
return min(1.0, f)
def phase(peak, slot):
d = abs((slot[1] - peak + 12.0) % 24.0 - 12.0)
return E ** (-(d / PHASE_WIDTH) ** 2)
if __name__ == "__main__":
print(" score of each slot for each case:")
print(" case " + "".join("{:>9d}".format(i + 1)
for i in range(len(SLOTS))))
agree = 0
for idx, (cname, modern, needs, peak) in enumerate(CASES):
scores = [condition(needs, s) * phase(peak, s) for s in SLOTS]
best = max(range(len(SLOTS)), key=lambda i: scores[i])
if best == idx:
agree += 1
print(" {:<22}".format(cname)
+ "".join("{:>9.4f}".format(v) for v in scores))
print(" slot names, in the order of the columns:")
for i, s in enumerate(SLOTS):
print(" {}: {}".format(i + 1, s[0]))
print(" case classical slot arithmetic")
for idx, (cname, modern, needs, peak) in enumerate(CASES):
scores = [condition(needs, s) * phase(peak, s) for s in SLOTS]
best = max(range(len(SLOTS)), key=lambda i: scores[i])
print(" {:<22}{:<22}{}"
.format(cname, SLOTS[idx][0], SLOTS[best][0]))
print(" cases where the two agree: {} of {}"
.format(agree, len(CASES)))
print(" as a share: {:.4f}".format(agree / len(CASES)))
print(" the modern class each case was mapped to:")
for cname, modern, needs, peak in CASES:
print(" {:<22}{}".format(cname, modern))
print(" the mapping is the author's and is where the argument can")
print(" be attacked: a different mapping would give a different")
print(" count, and the rule itself fixes none of it")
worst = min(range(len(CASES)),
key=lambda i: max(condition(CASES[i][2], s)
* phase(CASES[i][3], s) for s in SLOTS))
wv = max(condition(CASES[worst][2], s) * phase(CASES[worst][3], s)
for s in SLOTS)
print(" the case with the weakest best score is '{}' at {:.4f}"
.format(CASES[worst][0], wv))
# score of each slot for each case:
# case 1 2 3 4
# above the diaphragm 0.9608 0.0646 0.0789 0.0546
# below the abdomen 0.3278 0.9587 0.2369 0.0167
# limbs and vessels 0.8560 0.2271 1.0000 0.0707
# marrow, at night 0.0556 0.0122 0.1054 0.9900
# slot names, in the order of the columns:
# 1: after food, morning
# 2: before food, midday
# 3: empty stomach at dawn
# 4: full stomach at night
# case classical slot arithmetic
# above the diaphragm after food, morning after food, morning
# below the abdomen before food, midday before food, midday
# limbs and vessels empty stomach at dawn empty stomach at dawn
# marrow, at night full stomach at night full stomach at night
# cases where the two agree: 4 of 4
# as a share: 1.0000
# the modern class each case was mapped to:
# above the diaphragm fat-soluble vitamin
# below the abdomen mineral needing acid
# limbs and vessels trace mineral, empty
# marrow, at night calming mineral
# the mapping is the author's and is where the argument can
# be attacked: a different mapping would give a different
# count, and the rule itself fixes none of it
# the case with the weakest best score is 'below the abdomen' at 0.9587The arithmetic picks the same slot as the classical rule in all four cases, a share of 1.0000, and the weakest of the four best scores is 0.9587. That result should be treated with suspicion rather than satisfaction, and the program says so in its own output: the mapping from classical category to modern class is the author’s, and a different mapping would give a different count. “A complaint above the diaphragm” was read here as standing for a fat-soluble vitamin, and nothing in the classical text licenses that reading.
What the comparison does establish is narrower and still worth having. The classical rule varies two things, the relation to food and the hour, and those are exactly the two variables that Chapters 2 and 4 found to matter. A rule that had varied the phase of the moon instead would not have been comparable at all. So the claim this chapter makes is that the Bencao gangmu’s four cases are about the right variables, and the claim it declines to make is that they got the assignments right for reasons anyone in the sixteenth century could have known.
Hypothesis · not established
The author’s interpretation (not yet tested): classical dosing-hour rules in the Chinese materia medica tradition encode the relation to food more reliably than they encode the hour, because the relation to food has an immediately observable consequence, such as nausea or its absence, while the hour does not. What would show it wrong: if a systematic coding of dosing-time instructions across the classical corpus showed the hour specified as often and as consistently as the relation to food, with no difference in how often later editions altered one or the other, the asymmetry would not exist.
Summary
The seven relations of the Shennong preface are a classification of ordered pairs and therefore an interaction matrix. Applied to a modern cupboard, 47 of the 55 pairs act alone, 7 are in mutual opposition and 1 is a case of assistance, with mutual reinforcement empty, which says the cupboard was not composed as a formula. The four dosing-hour rules quoted in the Bencao gangmu agree with the arithmetic in all four cases, but the agreement depends on the author’s own mapping, so the honest claim is that the rules vary the right two variables. Chapter 8 puts adherence and safety in front of all of it.
References
Fan, Y. (2026). Polypharmacy Supplement Interactions, Chronotherapy Dosing Windows and Cytochrome Enzyme Competition: Michaelis-Menten Inhibition Matrices, Chelation Absorption Losses and Circadian Receptor Sensitivity [复合补剂相互作用、时间疗法给药窗口与代谢酶竞争] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090250 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).