Dataset of AlphaFold's internal representations of 4,581 proteins relevant for drug discovery

This dataset contains the outputs of the AlphaFold model for 4,581 proteins that are relevant targets in drug discovery.

More information on the dataset can be found at the following repository:

Dataset structure:

↓ data/* -> main data directory

↓ data/PID/* -> data of a single protein of length L

Filename	Description	Tensor shape	Lightweight
single.npy	( s i ) evoformer single representation	[L x 384]	✔️
structure.npy	( a i ) output of the last layer of structure module	[L x 384]	✔️
msa.npy***	( m s i ) processed MSA representation	[N x L x 256]
pair.npy***	( z i j ) evoformer pair representation	[L x L x 128]
PID.pdb	3D protein structure prediction		✔️
PID_unrelaxed.pdb	3D protein structure prediction w/o relaxation step (D)		✔️
confidence.npy*	confidence in structure prediction (0-100)	1	✔️
plldt.npy*	confidence in structure prediction per residue	[L]	✔️
PID.fasta	protein amino acid sequence and metadata		✔️
timings.json	Processing log		✔️

↓ data/PID2/* -> data of protein #2

...

*Note: L: sequence length, N: number of aligned sequences via MSA.