5D Library › 复合补剂相互作用、时间疗法给药窗口与代谢酶竞争 · English

第三章 那一把

两类损耗,相乘

前两章量了两件彼此独立的事:一项在自己的通路上输给竞争者多少,以及它在到达时输给肠道条件多少。它们是相乘的,因为一项既被挤掉又赶上坏时机,就两头都吃亏。把一切放在一个时刻吞下的做法,保证了柜子里至多只有一个条件是对的,因为铁要空腹且胃酸充足,脂溶性维生素要脂肪,而锌要没有植酸。

有一对是反方向的。Shoba 等人给志愿者服用加与不加胡椒碱的姜黄素,测到两者同服时姜黄素血浆浓度大幅上升,这是这个柜子里唯一一处主张合并而不是分开的相互作用 (Shoba et al. 1998)。程序里所用的增强倍数刻意远小于那篇论文所报告的数字,这样结论就不依赖对单一研究最有利的那种读法。本章先单独计算共享位点上的竞争与肠道状态:p05 没有计入胡椒碱的增强,第六章在 p11 建立排程时,再把它作为一个独立的增强项并入。

证据

已发表。 Shoba 等人报告,在动物与人体志愿者中,同服胡椒碱使姜黄素的生物利用度大幅提高 (Shoba et al. 1998)。

推导

# Chapter 3. p05: What the whole handful actually delivers. Two losses
# are multiplied for each item. The first is competition at a shared
# site, computed as in p02. The second is the condition of the gut when
# the dose arrives: a fat-soluble vitamin needs fat in the meal, iron
# needs an acid stomach and no polyphenol, zinc needs little phytate.
# Taking everything at one moment means at most one of those conditions
# can be right. The comparison is against a schedule that gives each
# item the condition it wants, which is what the rest of the book builds.
ITEMS = (
    # name, dose mg, molar mass, site, Km mmol, needs
    ("magnesium",     400.00,  24.305, "divalent",  4.0000, "plain"),
    ("zinc",           25.00,  65.380, "divalent",  0.1500, "lowphytate"),
    ("calcium",       500.00,  40.080, "divalent",  5.0000, "plain"),
    ("iron",           18.00,  55.850, "divalent",  0.0800, "acid"),
    ("vitamin D3",      0.10, 384.600, "micelle",   0.0004, "fat"),
    ("vitamin K2",      0.18, 444.700, "micelle",   0.0005, "fat"),
    ("NAD precursor", 300.00, 255.200, "nrk",       1.0000, "plain"),
    ("curcumin",      500.00, 368.400, "conjugate", 1.0000, "fat"),
    ("piperine",       10.00, 285.300, "conjugate", 0.0500, "plain"),
    ("theanine",      200.00, 174.200, "amino",     1.0000, "plain"),
    ("caffeine",      100.00, 194.200, "cyp1a2",    0.5000, "plain"),
)
# the one moment everything is taken: a coffee and a slice of toast
AT_ONCE = {"fat_g": 4.0, "phytate_mg": 900.0, "polyphenol_mg": 320.0,
           "ph": 2.1}
# what each condition looks like when the item gets its own slot
IDEAL = {"fat": {"fat_g": 18.0, "phytate_mg": 0.0, "polyphenol_mg": 0.0,
                 "ph": 2.0},
         "acid": {"fat_g": 2.0, "phytate_mg": 0.0, "polyphenol_mg": 0.0,
                  "ph": 1.6},
         "lowphytate": {"fat_g": 6.0, "phytate_mg": 60.0,
                        "polyphenol_mg": 0.0, "ph": 1.8},
         "plain": {"fat_g": 6.0, "phytate_mg": 120.0,
                   "polyphenol_mg": 40.0, "ph": 1.9}}
FAT_HALF, PHY_MASS, ZN_MASS = 9.0, 660.0, 65.38
PH_HALF, PH_HILL = 3.4, 2.4

def mmol(dose_mg, molar_mass):
    return dose_mg / molar_mass

def site_factor(i, together):
    name, dose, mm, site, km, needs = ITEMS[i]
    s = mmol(dose, mm)
    extra = 0.0
    if together:
        for j, other in enumerate(ITEMS):
            if j != i and other[3] == site:
                extra += mmol(other[1], other[2]) / other[4]
    return (s / (km * (1.0 + extra) + s)) / (s / (km + s))

def condition_factor(needs, env):
    f = 1.0
    if needs == "fat":
        f *= env["fat_g"] / (FAT_HALF + env["fat_g"]) \
            / (18.0 / (FAT_HALF + 18.0))
    if needs == "acid":
        f *= 1.0 / (1.0 + env["polyphenol_mg"] / 120.0)
        f *= (1.0 / (1.0 + (env["ph"] / PH_HALF) ** PH_HILL)) \
            / (1.0 / (1.0 + (1.6 / PH_HALF) ** PH_HILL))
    if needs == "lowphytate":
        ratio = (env["phytate_mg"] / PHY_MASS) / (25.0 / ZN_MASS)
        f *= 1.0 / (1.0 + 0.55 * ratio)
    return min(1.0, f)

if __name__ == "__main__":
    print("   item           site    condition   together   separated")
    tog, sep = [], []
    for i, (name, dose, mm, site, km, needs) in enumerate(ITEMS):
        a = site_factor(i, True) * condition_factor(needs, AT_ONCE)
        b = site_factor(i, False) * condition_factor(needs, IDEAL[needs])
        tog.append(a)
        sep.append(b)
        print("   {:<15}{:<8.4f}{:<12.4f}{:<11.4f}{:.4f}"
              .format(name, site_factor(i, True),
                      condition_factor(needs, AT_ONCE), a, b))
    mean_t = sum(tog) / len(tog)
    mean_s = sum(sep) / len(sep)
    print("   mean delivered fraction, all at once: {:.4f}".format(mean_t))
    print("   mean delivered fraction, separated:   {:.4f}".format(mean_s))
    print("   loss from taking the handful at one moment: {:.4f}"
          .format(1.0 - mean_t / mean_s))
    worst = min(range(len(ITEMS)), key=lambda i: tog[i] / sep[i])
    print("   worst item: {} keeps {:.4f} of what it could"
          .format(ITEMS[worst][0], tog[worst] / sep[worst]))
    losses = sorted((1.0 - tog[i] / sep[i], ITEMS[i][0])
                    for i in range(len(ITEMS)))
    print("   item           loss from the single moment")
    for l, name in reversed(losses):
        print("   {:<15}{:.4f}".format(name, l))
    big = [l for l, _ in losses if l >= 0.35]
    print("   items losing at least 0.35: {} of {}"
          .format(len(big), len(ITEMS)))
    print("   their losses run from {:.4f} to {:.4f}"
          .format(min(big), max(big)))
#    item           site    condition   together   separated
#    magnesium      0.3605  1.0000      0.3605     1.0000
#    zinc           0.2502  0.3377      0.0845     0.8844
#    calcium        0.2463  1.0000      0.2463     1.0000
#    iron           0.3544  0.2414      0.0856     1.0000
#    vitamin D3     0.6709  0.4615      0.3096     1.0000
#    vitamin K2     0.7357  0.4615      0.3396     1.0000
#    NAD precursor  1.0000  1.0000      1.0000     1.0000
#    curcumin       0.7708  0.4615      0.3557     1.0000
#    piperine       0.5562  1.0000      0.5562     1.0000
#    theanine       1.0000  1.0000      1.0000     1.0000
#    caffeine       1.0000  1.0000      1.0000     1.0000
#    mean delivered fraction, all at once: 0.4853
#    mean delivered fraction, separated:   0.9895
#    loss from taking the handful at one moment: 0.5096
#    worst item: iron keeps 0.0856 of what it could
#    item           loss from the single moment
#    iron           0.9144
#    zinc           0.9045
#    calcium        0.7537
#    vitamin D3     0.6904
#    vitamin K2     0.6604
#    curcumin       0.6443
#    magnesium      0.6395
#    piperine       0.4438
#    theanine       0.0000
#    caffeine       0.0000
#    NAD precursor  0.0000
#    items losing at least 0.35: 8 of 11
#    their losses run from 0.4438 to 0.9144

在一杯咖啡与一片吐司旁一次吞下,这个柜子交付的平均分数是 0.4853,而给每项所需条件的排程是 0.9895,损失 0.5096。损失并不是均匀铺开的。十一项里有八项损失至少 0.35,而这些损失从 0.4438 到 0.9144。铁只保住它本可交付的 0.0856,而锌损失 0.9045,因为两者同时被竞争与条件击中。有三项毫无损失,因为没有东西与它们共用通路,它们也不需要任何特定条件。

一次吞下整个柜子时,每一项的损失(p05)。

0.5096 这个数字值得与通常的说法比较。缓释配方与「优化组合」的广告常引用「同服损失三分之一到三分之二」。这个区间大致是对的,属于营销说法罕见地经得起算术检验的情形。经不起检验的是它暗示的补救:损失来自竞争与肠道条件,而一粒缓慢释放内容物的胶囊两者都不改变。

分开到底能收回多少

如果损失是真的,下一个问题就是修它要花多少。把一堆东西分成几组在注意力上并不免费,而正确的问题是每增加一组的边际价值,而不是一份完美排程的价值。

推导

# Chapter 3. p06: How much separating the handful can recover. The items
# of p05 are split into one, two or three groups. Everything inside a
# group is swallowed together, so it competes for shared sites, and the
# group is taken under one condition, chosen from the four the previous
# program defines by whichever suits that group's members best. Every
# assignment is enumerated; the first item is fixed in the first group
# and the group labels are forced into canonical order, which removes
# the relabellings without removing any genuinely different split. The
# answer is therefore the true optimum for one, two and three groups.
ITEMS = (
    ("magnesium",     400.00,  24.305, "divalent",  4.0000, "plain"),
    ("zinc",           25.00,  65.380, "divalent",  0.1500, "lowphytate"),
    ("calcium",       500.00,  40.080, "divalent",  5.0000, "plain"),
    ("iron",           18.00,  55.850, "divalent",  0.0800, "acid"),
    ("vitamin D3",      0.10, 384.600, "micelle",   0.0004, "fat"),
    ("vitamin K2",      0.18, 444.700, "micelle",   0.0005, "fat"),
    ("NAD precursor", 300.00, 255.200, "nrk",       1.0000, "plain"),
    ("curcumin",      500.00, 368.400, "conjugate", 1.0000, "fat"),
    ("piperine",       10.00, 285.300, "conjugate", 0.0500, "plain"),
    ("theanine",      200.00, 174.200, "amino",     1.0000, "plain"),
    ("caffeine",      100.00, 194.200, "cyp1a2",    0.5000, "plain"),
)
IDEAL = {"fat": (18.0, 0.0, 0.0, 2.0),
         "acid": (2.0, 0.0, 0.0, 1.6),
         "lowphytate": (6.0, 60.0, 0.0, 1.8),
         "plain": (6.0, 120.0, 40.0, 1.9)}
TYPES = tuple(sorted(IDEAL))
FAT_HALF, PHY_MASS, ZN_MASS = 9.0, 660.0, 65.38
PH_HALF, PH_HILL = 3.4, 2.4

def mmol(dose_mg, molar_mass):
    return dose_mg / molar_mass

def site_in_group(i, group):
    name, dose, mm, site, km, needs = ITEMS[i]
    s = mmol(dose, mm)
    extra = sum(mmol(ITEMS[j][1], ITEMS[j][2]) / ITEMS[j][4]
                for j in group if j != i and ITEMS[j][3] == site)
    return (s / (km * (1.0 + extra) + s)) / (s / (km + s))

def condition(needs, env):
    fat_g, phytate, polyphenol, ph = env
    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 + polyphenol / 120.0)
        f *= (1.0 / (1.0 + (ph / PH_HALF) ** PH_HILL)) \
            / (1.0 / (1.0 + (1.6 / PH_HALF) ** PH_HILL))
    if needs == "lowphytate":
        ratio = (phytate / PHY_MASS) / (25.0 / ZN_MASS)
        f *= 1.0 / (1.0 + 0.55 * ratio)
    return min(1.0, f)

def score(assign, k):
    total = 0.0
    for g in range(k):
        group = [i for i, a in enumerate(assign) if a == g]
        if not group:
            continue
        site = {i: site_in_group(i, group) for i in group}
        best = None
        for ty in TYPES:
            env = IDEAL[ty]
            v = sum(site[i] * condition(ITEMS[i][5], env) for i in group)
            if best is None or v > best:
                best = v
        total += best
    return total / len(ITEMS)

def enumerate_assignments(k):
    n = len(ITEMS)
    assign = [0] * n

    def go(i, used):
        if i == n:
            yield tuple(assign)
            return
        for g in range(min(used + 1, k)):
            assign[i] = g
            yield from go(i + 1, max(used, g + 1))
    yield from go(1, 1)

if __name__ == "__main__":
    print("   groups   best mean delivered fraction   splits tried")
    best_by_k = {}
    for k in (1, 2, 3):
        best, arg, n = -1.0, None, 0
        for a in enumerate_assignments(k):
            n += 1
            v = score(a, k)
            if v > best:
                best, arg = v, a
        best_by_k[k] = (best, arg)
        print("   {:<9d}{:<31.4f}{}".format(k, best, n))
    one = best_by_k[1][0]
    for k in (2, 3):
        v, arg = best_by_k[k]
        print("   {} groups recover {:+.4f} over one, a gain of {:.4f}"
              .format(k, v - one, v / one - 1.0))
    print("   the best three-group split:")
    v, arg = best_by_k[3]
    for g in range(3):
        group = [ITEMS[i][0] for i, a in enumerate(arg) if a == g]
        if group:
            print("   group {}:".format(g + 1))
            line = "   "
            for nm in group:
                if len(line) + len(nm) + 2 > 72:
                    print(line)
                    line = "   "
                line += "  " + nm
            print(line)
    print("   mean delivered fraction with three groups: {:.4f}".format(v))
    print("   remaining shortfall against a perfect schedule: {:.4f}"
          .format(1.0 - v))
    print("   the gain from the second group is {:.4f} and from the"
          .format(best_by_k[2][0] - one))
    print("   third only {:.4f}, so most of the money is in the first"
          .format(best_by_k[3][0] - best_by_k[2][0]))
    print("   split, which is the practical conclusion")
#    groups   best mean delivered fraction   splits tried
#    1        0.6284                         1
#    2        0.8386                         1024
#    3        0.9241                         29525
#    2 groups recover +0.2102 over one, a gain of 0.3345
#    3 groups recover +0.2957 over one, a gain of 0.4705
#    the best three-group split:
#    group 1:
#      magnesium  vitamin D3  NAD precursor  curcumin  theanine  caffeine
#    group 2:
#      zinc  calcium  vitamin K2  piperine
#    group 3:
#      iron
#    mean delivered fraction with three groups: 0.9241
#    remaining shortfall against a perfect schedule: 0.0759
#    the gain from the second group is 0.2102 and from the
#    third only 0.0855, so most of the money is in the first
#    split, which is the practical conclusion

把一切放在一起但为整堆选出单一最佳条件,已经得到 0.6284,高于 p05 的 0.4853,因为 p05 把条件固定在咖啡加吐司,而不是让它被挑选。两组达到 0.8386,三组达到 0.9241。第二组带来的增益是 0.2102,第三组只带来 0.0855,而相对完美排程的剩余差距是 0.0759。

这就是本章的实际结论,而它是关于注意力的结论,不是关于生物化学的结论。第一次分开值第二次的两倍半,所以一个只会可靠地做一件事的读者,应该把这堆东西分开一次,把铁与常量矿物质放在一天的两端,然后就停在那里。第三组买到的不足十分之一,还引入了第八章所显示的、人们会漏掉的那个时段。

假说 · 尚未验证

作者的诠释(尚未验证): 每增加一个服用时段的边际价值下降得比依从性更快,所以对几乎每一个真实的柜子,两组排程使交付的营养素最大,而三组排程并不,尽管它在纸面上得分更高。什么会证明它错: 在一项同时测量依从性与某项营养状态标志物的试验里,如果三个月后三组排程所测得的状态高于两组排程,那么纸面上的排序毕竟才是对的。

小结

竞争与肠道条件相乘,所以一次吞下整个柜子比一份适合每项的排程少 0.5096,其中十一项里有八项损失至少 0.35,而铁只保住 0.0856。分成两组收回 0.2102,第三组只再收回 0.0855,剩余差距 0.0759。第四章问:除了让各项互不干扰之外,时钟还买到了什么。

参考文献

Shoba, Guido, David Joy, Thangam Joseph, Muhammed Majeed, Rajagopalan Rajendran, and P. S. S. R. Srinivas. 1998. “Influence of Piperine on the Pharmacokinetics of Curcumin in Animals and Human Volunteers.” Planta Medica 64 (4): 353 to 356.

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).