Skip to main content

Table 2 Knowledge base

From: Joint probabilistic-logical refinement of multiple protein feature predictors

Rule name

Weight

Rule

Description

I1

per-protein

PredLoc(p,l,w) ∧ IsLoc(p,l)

Input rule for subcellular localization

I2

per-cysteine

PredDis(p,n,w) ∧ IsDis(p,n)

Input rule for disulfide bonding state

I3

per-cysteine

PredMet(p,n,w) ∧ IsMet(p,n)

Input rule for metal bonding state

I1P

per-protein

PredLoc(p,l,w) ∧ ProxyLoc(p,l)

Input rule for proxy subcellular localization

I2P

per-cysteine

PredDis(p,n,w) ∧ ProxyDis(p,n)

Input rule for proxy disulfide bonding state

I3P

per-cysteine

PredMet(p,n,w) ∧ ProxyMet(p,n)

Input rule for proxy metal bonding state

PX1

from data

ProxyLoc(p,l) ⇔ IsLoc(p,l)

Proxy rule for subcellular localization

PX2

from data

ProxyDis(p,n) ⇔ IsDis(p,n)

Proxy rule for disulfide bonding state

PX3

from data

ProxyMet(p,n) ⇔ IsMet(p,n)

Proxy rule for metal bonding state

DL1

from data

IsLoc(p,l) ⇒!IsDis(p,n)

Compartment lhinders the formation of disulf. bonds

DL2

from data

IsLoc(p,l) ⇒IsDis(p,n)

Compartment l favors the formation of disulf. bonds

DM

from data

!(IsDis(p,n) ∧IsMet(p,n))

A half-cysteine can’t bind a metal atom

L1

∞

∨ l IsLoc(p,l)

A protein must belong to at least onecompartment

L2

∞

∀l1IsLoc(p,l1)∧ ∧ l2 !IsLoc(p,l2)

A protein must belong to at most onecompartment