Lab 3 — The matrix effect¶

Question for students: almonds and almond flour have the same ingredient list. Do they behave the same in the gut?

This lab has no gate in it at all. It is a pure bound story driven by physical form — the case where a nutrition-facts panel is least informative.

The pair¶

from biology_as_code.packets import get_packet

for packet_id in ("ex.almond.whole", "ex.almond.flour"):
    packet = get_packet(packet_id)
    print(packet.common_name)
    print("  cargo    :", packet.cargo_nutrients())
    print("  integrity:", packet.matrix_integrity)
    print("  notes    :", packet.matrix.get("notes"))

Identical cargo. The difference is recorded in matrix.integrity: intact for the whole nut, destroyed for the flour.

Audit both¶

from biology_as_code import Claim, audit_claim

claim = Claim(
    id="claim.matrix_form",
    surface_claim="Food form changes lipid accessibility",
    verb_class="bound_increase",
    nutrient="lipid",
)

for packet_id in ("ex.almond.whole", "ex.almond.flour"):
    result = audit_claim(claim, get_packet(packet_id))
    directions = [f.direction for f in result.bound_findings]
    print(f"{packet_id:20s} {result.verdict:10s} {directions} {result.law_refs}")

Expected:

ex.almond.whole      Plausible  ['NARROWS_BOUND'] ('LAW-024',)
ex.almond.flour      Plausible  ['EXPANDS_BOUND'] ('LAW-024',)

The teaching point¶

LAW-024 states that food form and fibre viscosity govern gastric emptying and the rate of absorption. The intact cell wall encapsulates intracellular lipid, so the accessible fraction is lower. Milling removes that barrier.

Nothing about the panel changes. Calories, fat grams, protein — all identical. The variable that moved is not in the panel at all.

Why this matters for scoring systems

Any nutrient-profiling model that reads only a panel cannot distinguish these two foods. Whatever score it emits will be the same for both. That is a structural limit of panel-based scoring, not a tuning problem.

Unknown is not intact¶

Most packets in the repository do not declare matrix integrity:

from biology_as_code.packets import iter_packets
from collections import Counter

print(Counter(p.matrix_integrity for p in iter_packets()))

Those unknown packets produce no matrix finding at all — the auditor does not assume an intact default. A missing declaration yields silence, not a guess.

Exercise¶

ex.orange.whole and ex.orange.juice are the same physical pair as the almonds. Both are stubs. Fill in their matrix blocks and re-run this lab.
ex.oats.porridge.plain and ex.oat.flour.gruel are a third such pair. Which law would you cite for a viscosity claim, and does the register support it?
LAW-024 covers rate. Does rate alone justify a claim about total absorbed lipid over a day? Where does that argument break?