5D Library › 亚知觉色胺药代动力学、5-HT2A 受体脱敏与神经可塑性窗口的生物数学模型 · English
第四章 受体下调与耐受棘轮:每次下降,累计上升
已知的部分
证据
受体下调与耐受(动物与人体)。 反复给予 LSD 使大鼠脑内 5-HT2 受体选择性下调,Buckholtz 等人 1990 年报告了这一结果 (Buckholtz et al. 1990)。Gresch 等人发现行为耐受与皮层 5-HT2A 信号的降低相关,并把两者联系起来 (Gresch et al. 2005)。在人体方面,Isbell 等人 1956 年就观察到连续给药期间耐受的发展 (Isbell et al. 1956)。Nichols 的综述整理了这一文献 (Nichols 2016)。
注
这些证据的暴露水平。 上述动物实验与早期人体观察使用的暴露远高于本书所讨论的范围。受体下调在高暴露下存在,这一点有证据;人体在亚知觉暴露下的 5-HT2A 恢复半衰期没有被测量。本章的全部表格都是「如果恢复半衰期是某个值会怎样」,不是「实际的日程应当如何」。
棘轮的数学
把表面受体密度 写成合成、组成性降解、以及由占据驱动的内化三项之和, 归一化使无药时的稳态为 1:
推导
# Chapter 4. Receptor downregulation under repeated exposure: the "tolerance
# ratchet". Surface receptor density R obeys synthesis, constitutive
# degradation, and internalisation driven by occupancy:
# dR/dt = ksyn - kdeg R - kint O(Ce) R
# O(Ce) = Ce / (Ce + 1) (p01, h = 1, dimensionless)
# with R normalised so that the drug-free steady state is 1. Exposure is the
# effect-site curve of p04 repeated on a schedule. Time is in days; the
# receptor recovery half-life tau_R is the parameter that decides everything,
# and it is the one measured only in animals at much higher exposures (Chapter
# 4 text). Author's model; dimensionless units.
import math
DT = 0.002 # days
# receptor recovery half-life, days (illustrative)
TAU_R = 3.0
KDEG = math.log(2.0) / TAU_R
KSYN = KDEG # so that drug-free steady state R = 1
# internalisation rate at full occupancy, per day
KINT = 2.0
KE_CE = math.log(2.0) / 0.25 # effect-site clearance: 6-hour half-life
def run(dose_fraction, interval_days, n_doses, tail_days=30.0, tau_r=TAU_R):
kdeg = math.log(2.0) / tau_r
ksyn = kdeg
R, Ce = 1.0, 0.0
t_end = interval_days * n_doses + tail_days
doses = [i * interval_days for i in range(n_doses)]
di, R_at_dose, Rmin = 0, [], 1.0
n = int(t_end / DT)
for i in range(n):
t = i * DT
if di < len(doses) and t >= doses[di] - DT / 2:
R_at_dose.append(R)
Ce += dose_fraction # instantaneous input, dimensionless
di += 1
O = Ce / (Ce + 1.0)
R += DT * (ksyn - kdeg * R - KINT * O * R)
Ce += DT * (-KE_CE * Ce)
Rmin = min(Rmin, R)
return R_at_dose, Rmin, R
if __name__ == "__main__":
print("Receptor density normalised to 1 without drug. Recovery half-life"
" {:.1f} days.".format(TAU_R))
print(
"Dose fraction 0.5 in the dimensionless units of p01 (occupancy"
" ~0.33).\n"
)
print(" interval (days) R before dose 1, 2, 3, 5, 10 R at dose 10"
" minimum R")
for iv in (1.0, 2.0, 3.0, 5.0, 7.0):
Rd, Rmin, Rend = run(0.5, iv, 10)
picks = [Rd[k] for k in (0, 1, 2, 4, 9)]
print(
" {:15.1f} ".format(iv)
+ " ".join("{:.3f}".format(v) for v in picks)
+ "{:12.3f} {:10.3f}".format(Rd[9], Rmin)
)
print(
"\nThe ratchet is visible in the columns: at a 1-day interval the"
" density"
)
print(
"before each dose keeps falling and settles well below 1, while at a"
" 7-day"
)
print(
"interval it returns almost to 1 before the next dose. The quantity"
" that"
)
print(
"decides which happens is the ratio of the recovery half-life to the"
" dosing"
)
print("interval, not the amount.")
print(
"\nSteady-state pre-dose density against interval, for three recovery"
)
print("half-lives (dose fraction held at 0.5, 20 doses):")
print(" interval tau_R = 1 d tau_R = 3 d tau_R = 7 d")
for iv in (1.0, 2.0, 3.0, 4.0, 5.0, 7.0, 10.0):
row = []
for tau in (1.0, 3.0, 7.0):
Rd, _, _ = run(0.5, iv, 20, tau_r=tau)
row.append(Rd[-1])
print(" {:8.1f} {:13.3f} {:13.3f} {:13.3f}".format(iv, *row))
print(
"\nRead the tau_R = 7 d column: even a 10-day interval leaves the"
" density"
)
print(
"below 1. With a recovery half-life longer than the interval, every"
" dose"
)
print(
"starts from a lower density than the last, and the sequence"
" converges to a"
)
print(
"new steady state - the ratchet. The critical interval, defined as"
" the one"
)
print("at which the steady-state pre-dose density is 0.95:")
for tau in (1.0, 3.0, 7.0):
lo, hi = 0.2, 60.0
for _ in range(45):
mid = 0.5 * (lo + hi)
Rd, _, _ = run(0.5, mid, 20, tau_r=tau)
if Rd[-1] < 0.95:
lo = mid
else:
hi = mid
print(" tau_R = {:4.1f} days -> critical interval {:5.1f} days"
.format(tau, hi))
print(
"\nThe honest point about all of this: tau_R for 5-HT2A in humans at"
)
print(
"sub-perceptual exposures is not measured. The animal literature that"
)
print(
"establishes downregulation used repeated exposures far above"
" these, and"
)
print(
"the recovery half-life at low occupancy need not be the same. So the"
" table"
)
print(
"above is a map of what WOULD follow from a given tau_R, and the"
" schedules"
)
print("recommended in popular accounts cannot be derived from it.")
# Receptor density normalised to 1 without drug. Recovery half-life 3.0 days.
# Dose fraction 0.5 in the dimensionless units of p01 (occupancy ~0.33).
#
# interval (days) R before dose 1, 2, 3, 5, 10 R at dose 10 minimum R
# 1.0 1.000 0.796 0.666 0.541 0.476 0.476 0.440
# 2.0 1.000 0.824 0.741 0.684 0.668 0.668 0.582
# 3.0 1.000 0.860 0.807 0.780 0.776 0.776 0.655
# 5.0 1.000 0.911 0.891 0.885 0.884 0.884 0.725
# 7.0 1.000 0.944 0.936 0.935 0.935 0.935 0.756
#
# The ratchet is visible in the columns: at a 1-day interval the density
# before each dose keeps falling and settles well below 1, while at a 7-day
# interval it returns almost to 1 before the next dose. The quantity that
# decides which happens is the ratio of the recovery half-life to the dosing
# interval, not the amount.
#
# Steady-state pre-dose density against interval, for three recovery
# half-lives (dose fraction held at 0.5, 20 doses):
# interval tau_R = 1 d tau_R = 3 d tau_R = 7 d
# 1.0 0.745 0.471 0.272
# 2.0 0.892 0.667 0.441
# 3.0 0.951 0.776 0.559
# 4.0 0.977 0.842 0.642
# 5.0 0.989 0.884 0.704
# 7.0 0.997 0.935 0.789
# 10.0 1.000 0.970 0.864
#
# Read the tau_R = 7 d column: even a 10-day interval leaves the density
# below 1. With a recovery half-life longer than the interval, every dose
# starts from a lower density than the last, and the sequence converges to a
# new steady state - the ratchet. The critical interval, defined as the one
# at which the steady-state pre-dose density is 0.95:
# tau_R = 1.0 days -> critical interval 3.0 days
# tau_R = 3.0 days -> critical interval 8.0 days
# tau_R = 7.0 days -> critical interval 18.2 days
#
# The honest point about all of this: tau_R for 5-HT2A in humans at
# sub-perceptual exposures is not measured. The animal literature that
# establishes downregulation used repeated exposures far above these, and
# the recovery half-life at low occupancy need not be the same. So the table
# above is a map of what WOULD follow from a given tau_R, and the schedules
# recommended in popular accounts cannot be derived from it.棘轮在表里直接可见。恢复半衰期 天时,1 天间隔下每次给药前的密度持续下降并稳定在远低于 1 的水平,而 7 天间隔下几乎回到 1。第二张表把这件事一般化:决定结果的是恢复半衰期与给药间隔之比,不是量的大小。 天时,即使 10 天的间隔也留在 1 以下。
程序最后用二分法给出「临界间隔」——使稳态给药前密度达到 0.95 的间隔: 天时为 3.0 天,3 天时为 8.0 天,7 天时为 18.2 天。
注
这一节的结论只有一条,而且是条件式的: 如果恢复半衰期已知,那么临界间隔就是上表。由于它未知,那些在流行说法中以「每三天一次」「每四天一次」形式给出的日程,不能从受体下调推出来。它们可能出于别的理由(经验、方便、减少风险),但不是出于这一段药理学。
每次下降,累计上升
现在是本章的第二个结果,而它与通常的推论方向相反。
推导
# Chapter 4. From receptor loss to effect loss, and the schedule that makes a
# difference. The effect of a dose is taken proportional to R * O(Ce)
# integrated over the dose, where R is the surface density from p06 and O the
# occupancy from p01. Two things then follow that are worth separating: (a)
# effect per dose declines as R ratchets down, and (b) the CUMULATIVE effect
# over a fixed calendar period can still be higher on a tighter schedule even
# though each dose does less. Popular accounts conflate these. Dimensionless.
import math
DT = 0.002
KINT, KE_CE = 2.0, math.log(2.0) / 0.25
def course(dose_fraction, interval_days, n_doses, tau_r=3.0, horizon=None):
kdeg = math.log(2.0) / tau_r
ksyn = kdeg
R, Ce = 1.0, 0.0
t_end = horizon if horizon is not None else interval_days * n_doses + 20.0
doses = [
i * interval_days for i in range(n_doses) if i * interval_days < t_end
]
per_dose = [0.0] * len(doses)
di, cur = 0, -1
total = 0.0
for i in range(int(t_end / DT)):
t = i * DT
if di < len(doses) and t >= doses[di] - DT / 2:
cur = di
Ce += dose_fraction
di += 1
O = Ce / (Ce + 1.0)
signal = R * O
total += signal * DT
if cur >= 0:
per_dose[cur] += signal * DT
R += DT * (ksyn - kdeg * R - KINT * O * R)
Ce += DT * (-KE_CE * Ce)
return per_dose, total
if __name__ == "__main__":
print("Effect signal = R x occupancy, integrated.")
print("Dose fraction 0.5, tau_R = 3 days.\n")
print(" interval (days) effect of dose 1 dose 5 dose 10"
" dose 10 / dose 1")
for iv in (1.0, 2.0, 3.0, 5.0, 7.0):
pd, _ = course(0.5, iv, 10)
print(" {:15.1f} {:17.4f} {:8.4f} {:9.4f} {:17.3f}"
.format(iv, pd[0], pd[4], pd[9], pd[9] / pd[0]))
print(
"\nEach dose does less than the one before, and the loss is larger on"
" tighter"
)
print(
"schedules: at a 1-day interval the tenth dose produces {:.0%} of the"
" first,".format(
course(0.5, 1.0, 10)[0][9] / course(0.5, 1.0, 10)[0][0]
)
)
print(
"at a 7-day interval {:.0%}. That is tolerance in the ordinary sense."
.format(course(0.5, 7.0, 10)[0][9] / course(0.5, 7.0, 10)[0][0])
)
print(
"\nBut now hold the CALENDAR PERIOD fixed at 42 days and vary the"
" schedule:"
)
print(" interval (days) doses in 42 d cumulative effect per dose")
rows = []
for iv in (1.0, 2.0, 3.0, 4.0, 6.0, 7.0, 14.0):
n = int(42.0 / iv)
pd, total = course(0.5, iv, n, horizon=42.0)
rows.append((iv, n, total, total / n))
print(
" {:15.1f} {:15d} {:19.3f} {:10.4f}".format(
iv, n, total, total / n
)
)
best = max(rows, key=lambda r: r[2])
print(
"\n Cumulative effect over the fixed period is greatest at the"
" {:.0f}-day".format(best[0])
)
print(
" interval ({:.3f}), even though the effect per dose there is the"
" lowest".format(best[2])
)
print(
" ({:.4f} against {:.4f} at the 14-day interval). So 'tolerance"
" builds up'".format(best[3], rows[-1][3])
)
print(
" and 'a tighter schedule delivers less in total' are different"
" claims,"
)
print(
" and in this model the second is false. Any argument for spacing"
" doses"
)
print(
" that rests only on receptor downregulation therefore does not go"
)
print(
" through: downregulation reduces the effect PER DOSE while"
" leaving the"
)
print(
" total over a period higher, and which of those matters depends on"
" what"
)
print(
" the practice is for - a question the pharmacology cannot answer."
)
print(
"\nSensitivity to the one parameter that is not measured, tau_R,"
" with the"
)
print("calendar period and total doses held fixed (14 doses in 42 days):")
print(" tau_R (days) dose 14 / dose 1 cumulative effect")
for tau in (0.5, 1.0, 3.0, 7.0, 14.0):
pd, total = course(0.5, 3.0, 14, tau_r=tau, horizon=42.0)
print(
" {:12.1f} {:18.3f} {:19.3f}".format(tau, pd[13] / pd[0], total)
)
print(
"\n Across plausible recovery half-lives the ratio of the last dose"
" to the"
)
print(
" first ranges from near 1 to well under a half. The schedules in"
" popular"
)
print(" accounts are stated as though tau_R were known; it is not.")
# Effect signal = R x occupancy, integrated.
# Dose fraction 0.5, tau_R = 3 days.
#
# interval (days) effect of dose 1 dose 5 dose 10 dose 10 / dose 1
# 1.0 0.1194 0.0718 0.0703 0.589
# 2.0 0.1277 0.0911 0.0897 0.702
# 3.0 0.1283 0.1026 0.1021 0.796
# 5.0 0.1283 0.1148 0.1148 0.895
# 7.0 0.1283 0.1207 0.1206 0.940
#
# Each dose does less than the one before, and the loss is larger on tighter
# schedules: at a 1-day interval the tenth dose produces 59% of the first,
# at a 7-day interval 94%. That is tolerance in the ordinary sense.
#
# But now hold the CALENDAR PERIOD fixed at 42 days and vary the schedule:
# interval (days) doses in 42 d cumulative effect per dose
# 1.0 42 2.826 0.0673
# 2.0 21 1.946 0.0927
# 3.0 14 1.471 0.1051
# 4.0 10 1.124 0.1124
# 6.0 7 0.840 0.1200
# 7.0 6 0.733 0.1221
# 14.0 3 0.382 0.1274
#
# Cumulative effect over the fixed period is greatest at the 1-day
# interval (2.826), even though the effect per dose there is the lowest
# (0.0673 against 0.1274 at the 14-day interval). So 'tolerance builds up'
# and 'a tighter schedule delivers less in total' are different claims,
# and in this model the second is false. Any argument for spacing doses
# that rests only on receptor downregulation therefore does not go
# through: downregulation reduces the effect PER DOSE while leaving the
# total over a period higher, and which of those matters depends on what
# the practice is for - a question the pharmacology cannot answer.
#
# Sensitivity to the one parameter that is not measured, tau_R, with the
# calendar period and total doses held fixed (14 doses in 42 days):
# tau_R (days) dose 14 / dose 1 cumulative effect
# 0.5 0.995 1.857
# 1.0 0.962 1.766
# 3.0 0.796 1.471
# 7.0 0.577 1.150
# 14.0 0.389 0.907
#
# Across plausible recovery half-lives the ratio of the last dose to the
# first ranges from near 1 to well under a half. The schedules in popular
# accounts are stated as though tau_R were known; it is not.第一张表是通常的推论所依据的部分:每次给药的效应都比前一次小,而且在越密的日程上损失越大。这就是耐受。
第二张表把日历期固定在 42 天,让日程变化。结果是:累计效应在最密的日程(1 天间隔)上最大(2.826),而那里每次的效应最小(0.0673,对比 14 天间隔的 0.1274)。
所以「耐受会积累」与「越密的日程总共给得越少」是两个不同的说法,而在这个模型里第二个是假的。于是
仅凭受体下调,无法论证「应当拉开给药间隔」。下调降低的是每次的效应,而累计效应反而更高;哪一个重要,取决于这个做法的目的是什么——而药理学不能回答这个问题。
注
这不是在推荐密集的日程。 本书不推荐任何日程。本节说的是一个论证不成立:从受体下调出发不能得到「拉开间隔更好」。拉开间隔可能因为别的理由更好——例如降低不良事件的累积风险、降低依赖的可能、或者让第六章所说的「练习输入」有时间发生。那些是不同的论证,需要各自的证据。
程序的最后一段给出对 的敏感性:在固定的日历期与固定的给药次数下,最后一次对第一次的效应比从接近 1( 天)变到 0.389( 天)。这个范围覆盖了「几乎没有耐受」到「明显耐受」,而区分它们需要那个未被测量的参数。
假说 · 尚未验证
作者的假说(尚未确立): 在亚知觉暴露下重复给药时,5-HT2A 受体密度的下降幅度小于高暴露下按占据比例外推的预测,因为内化速率在低占据下不是占据的线性函数。什么样的结果能证明它是错的: 在取得监管批准的受体显像研究中(例如 5-HT2A 放射配体的正电子发射断层显像),若低暴露重复给药后受体结合的下降幅度与按占据线性外推的预测之差的 95% 置信区间包含零,则本假说没有得到支持;若下降幅度显著大于该预测,则本假说错。
本章小结
反复高暴露使大鼠脑内 5-HT2 受体下调,行为耐受与皮层信号降低相关,人体耐受在 1956 年就有观察。但这些暴露远高于本书讨论的范围,而人体在亚知觉暴露下的恢复半衰期未被测量。棘轮的数学表明,决定结果的是恢复半衰期与给药间隔之比:临界间隔在 为 1、3、7 天时分别是 3.0、8.0、18.2 天。由于 未知,流行说法中的具体日程不能从受体下调推出。本章的第二个结果与通常的推论相反:在固定的日历期内,较密的日程使每次的效应下降而使累计效应上升,所以仅凭受体下调不能论证应当拉开间隔;哪一个量重要取决于目的,而药理学不回答这个问题。最后,在固定日历期与次数下,最后一次对第一次的效应比随 从接近 1 变到 0.389。
参考文献
Fan, Y. (2026). Biomathematical Models of Sub-Perceptual Tryptamine Pharmacokinetics, 5-HT2A Receptor Desensitisation and Neuroplastic Window Dynamics: From Michaelis–Menten Kinetics of Prodrug Deacetylation and Whole-Brain Connectivity Entropy to the Microdose Tolerance Ratchet, with a Textual Examination of the Dose Boundary in the Classical Materia Medica [亚知觉色胺药代动力学、5-HT2A 受体脱敏与神经可塑性窗口的生物数学模型] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24019347 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).