ilp-datasets / protein /README.md
andrewcropper's picture
Add files using upload-large-folder tool
cb1dd83 verified
|
Raw
History Blame Contribute Delete
2.62 kB

Protein

Alpha-helix prediction in protein sequences, encoded as ILP problems. The task is to predict which positions in a protein sequence form part of an alpha-helix, based on the amino acid identity and physico-chemical properties at each position.

Tasks: 1 (single task in train/)

Head relation: alpha(Protein, Index) — position Index in protein Protein is part of an alpha-helix.

Background relations:

Relation Arity Description
position(Protein, Index, AminoAcid) 3 Position Index in protein Protein has amino acid AminoAcid
alpha_pair3(I1, I2) 2 Indices I1 and I2 are 3 positions apart (used to detect helix periodicity)
alpha_triplet(I1, I2, I3) 3 Indices I1, I2, I3 form a triplet pattern
octf(I1,...,I9) 9 Nine consecutive indices forming an octamer window
lth(A1, A2) 2 Amino acid A1 is less than A2 in hydrophobicity
ltv(A1, A2) 2 Amino acid A1 is less than A2 in volume
aliphatic(A) 1 Amino acid A is aliphatic
aromatic(A) 1 Amino acid A is aromatic
polar(A) 1 Amino acid A is polar
hydrophobic(A) 1 Amino acid A is hydrophobic
hydrophilic(A) 1 Amino acid A is hydrophilic
very_hydrophobic(A) 1 Amino acid A is very hydrophobic
small(A) 1 Amino acid A has a small side chain
tiny(A) 1 Amino acid A has a tiny side chain
large(A) 1 Amino acid A has a large side chain
positive(A) 1 Amino acid A has a positively charged side chain
negative(A) 1 Amino acid A has a negatively charged side chain
neutral(A) 1 Amino acid A is charge-neutral
hydro_b_acc(A) 1 Amino acid A is a hydrogen-bond acceptor
hydro_b_don(A) 1 Amino acid A is a hydrogen-bond donor
aromatic_or_very_hydrophobic(A) 1 Amino acid A is aromatic or very hydrophobic
ar_or_al_or_m(A) 1 Amino acid A is aromatic, aliphatic, or methionine
small_or_polar(A) 1 Amino acid A is small or polar
not_aromatic(A) 1 Amino acid A is not aromatic
not_<aa>(A) 1 Type-constant: A is not a specific amino acid (one relation per residue letter)

Task statistics

task_name num_bk_facts num_bk_relations num_pos_examples num_neg_examples
protein 4027 44 848 764

Source: IC London ILP Dataset Collection. Retrieved from https://www.doc.ic.ac.uk/~shm/Datasets/proteins/