关于「三十克上限:一个把吸收与利用混为一谈的数字」,《亮氨酸阈值与单餐上限神话》给出的定量结论是什么?
结论:第一章以 Bateman 曲线校准原文数字,让「上限」说法自己给出预测:开放模型的 12 小时合成比值为 1.322,封顶模型只有 1.057,后者落在由原文近似值推出的区间 1.095 至 1.624 之外。
数据口径:本书把它重写成可以计算的模型,并与已发表的实验对照:特罗梅伦等人 2023 年的实验显示,训练后一次 100 克乳蛋白比 25 克带来更大、更久的肌原纤维合成反应。 书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不对任何疾病或个人健康结果作出承诺。
对照数据表
下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 1.057, 1.095, 1.322, 1.624, 12
fit of K (g leucine in pool) to the 0-4 h and 4-12 h ratios
hill_h K_g ratio_0_4 ratio_4_12 ratio_0_12 sq_error
1 0.300 1.149 1.296 1.243 0.01341
2 0.200 1.145 1.431 1.322 0.00396
3 0.175 1.097 1.435 1.304 0.01193
chosen: h = 2, K = 0.200 g
12 h mean rate %/h 25 g: 0.0496 100 g: 0.0657
12 h ratio 100/25: open model 1.322, 30 g cap model 1.057
time-weighted reported ratio 1.333, 95% CI 1.095 to 1.624
cap prediction inside that CI: False查看完整验算程序
# Chapter 1. p12: Put the "30 g per meal" cap and the open-ended model to
# the Trommelen 2023 numbers. Synthesis rate (%/h) = B + VM*R(L), where L
# is leucine in the circulating pool (10 % of the protein pool of p11) and
# R(L) = L^h/(K^h + L^h): Michaelis-Menten when h = 1, a sigmoid
# "threshold" when h > 1. Targets (Trommelen et al. 2023): myofibrillar
# synthesis about 20 % higher with 100 g than 25 g over 0-4 h and about
# 40 % higher over 4-12 h. B = 0.030 and VM = 0.040 %/h are illustrative.
# The cap model feeds min(D, 30) g into the same machinery.
# The confidence interval uses an illustrative between-person CV of 0.25
# with 12 per group, the group size of the trial.
# 两个目标比值来自原文;其余参数为示意。
import math
from statistics import NormalDist
F, KE, KA0, DH = 0.70, 0.6, 0.3618, 48.36 # from p11
LEU, B, VM = 0.10, 0.030, 0.040
TARGET = (1.20, 1.40)
def ka(dose):
return KA0 / (1.0 + dose / DH)
def pool(t, dose):
a = ka(dose)
return F * dose * a / (a - KE) * (math.exp(-KE * t) - math.exp(-a * t))
def rate(t, dose, k, h):
lv = pool(t, dose) * LEU
return B + VM * lv ** h / (k ** h + lv ** h)
def mean_rate(dose, a, b, k, h, n=400):
w = (b - a) / n
return math.fsum(rate(a + (i + 0.5) * w, dose, k, h)
for i in range(n)) / n
def ratios(k, h, hi=100, lo=25):
r1 = mean_rate(hi, 0, 4, k, h) / mean_rate(lo, 0, 4, k, h)
r2 = mean_rate(hi, 4, 12, k, h) / mean_rate(lo, 4, 12, k, h)
r12 = mean_rate(hi, 0, 12, k, h) / mean_rate(lo, 0, 12, k, h)
return r1, r2, r12
def fit(h):
best = None
for i in range(5, 301, 5):
k = i / 1000.0
r1, r2, r12 = ratios(k, h)
err = (r1 - TARGET[0]) ** 2 + (r2 - TARGET[1]) ** 2
if best is None or err < best[0]:
best = (err, k, r1, r2, r12)
return best
if __name__ == "__main__":
print("fit of K (g leucine in pool) to the 0-4 h and 4-12 h ratios")
print("hill_h K_g ratio_0_4 ratio_4_12 ratio_0_12 sq_error")
fits = {}
for h in (1, 2, 3):
e, k, r1, r2, r12 = fit(h)
fits[h] = (e, k)
print("{:6d} {:5.3f} {:9.3f} {:10.3f} {:10.3f} {:8.5f}".format(
h, k, r1, r2, r12, e))
hb = min(fits, key=lambda x: fits[x][0])
kb = fits[hb][1]
print("chosen: h = {}, K = {:.3f} g".format(hb, kb))
cap = [mean_rate(min(d, 30), 0, 12, kb, hb) for d in (25, 100)]
opn = [mean_rate(d, 0, 12, kb, hb) for d in (25, 100)]
print("12 h mean rate %/h 25 g: {:.4f} 100 g: {:.4f}".format(*opn))
print("12 h ratio 100/25: open model {:.3f}, 30 g cap model {:.3f}"
.format(opn[1] / opn[0], cap[1] / cap[0]))
obs = (4 * TARGET[0] + 8 * TARGET[1]) / 12.0
se = math.sqrt(2 * math.log(1 + 0.25 ** 2) / 12)
z = NormalDist().inv_cdf(0.975)
lo, hi = obs * math.exp(-z * se), obs * math.exp(z * se)
print("time-weighted reported ratio {:.3f}, 95% CI {:.3f} to {:.3f}"
.format(obs, lo, hi))
print("cap prediction inside that CI: {}".format(lo <= cap[1] / cap[0]))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《肌蛋白合成亮氨酸阈值、单餐蛋白吸收上限神话与肌少症抗阻剂量母典》第 1 章「三十克上限:一个把吸收与利用混为一谈的数字」(版本 1.0.0). 5D Library. DOI 待发布
DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-188 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026188,
author = {Fan, Yafu},
title = {Leucine Thresholds for Muscle Protein Synthesis, the Per-Meal Protein Cap Myth and the Resistance-Training Dose against Sarcopenia},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}