SEQ2DOMML BINARY AMINO ACID IDENTITY FEATURE MANIFEST

Purpose

-------

This file documents the 20 binary amino acid identity features generated

directly within the Seq2DomML feature extraction script.

The features use one hot encoding to represent the identity of each residue.

Each residue is encoded as a vector containing exactly one value of 1 and

nineteen values of 0.

All binary amino acid identity features are stored as 32 bit floating point

values.

Accepted Amino Acids

--------------------

Only the 20 standard amino acids listed below are accepted by this encoding:

A  Alanine

C  Cysteine

D  Aspartic acid

E  Glutamic acid

F  Phenylalanine

G  Glycine

H  Histidine

I  Isoleucine

K  Lysine

L  Leucine

M  Methionine

N  Asparagine

P  Proline

Q  Glutamine

R  Arginine

S  Serine

T  Threonine

V  Valine

W  Tryptophan

Y  Tyrosine

Sequences containing any residue symbol outside this set are excluded before

feature encoding.

Feature Order

-------------

The 20 features are stored in the following exact order:

1.  bin_A

2.  bin_C

3.  bin_D

4.  bin_E

5.  bin_F

6.  bin_G

7.  bin_H

8.  bin_I

9.  bin_K

10. bin_L

11. bin_M

12. bin_N

13. bin_P

14. bin_Q

15. bin_R

16. bin_S

17. bin_T

18. bin_V

19. bin_W

20. bin_Y




Encoding Rule

-------------

For a residue with amino acid identity r, each binary feature is encoded as:

    bin_X(r) = 1 if r = X

    bin_X(r) = 0 if r does not equal X

where X is one of the 20 standard amino acid symbols.

Therefore, every valid residue has:

    exactly one active binary feature with a value of 1

    exactly nineteen inactive binary features with values of 0


Examples

--------

A residue with identity A is encoded as:

    bin_A = 1

and all other binary amino acid identity features are 0.

A residue with identity K is encoded as:

    bin_K = 1

and all other binary amino acid identity features are 0.

A residue with identity Y is encoded as:

    bin_Y = 1

and all other binary amino acid identity features are 0.

Scaling and Storage

-------------------

The 20 binary amino acid identity features are not standardized or otherwise

scaled after one hot encoding.

For a sequence of length L, the binary amino acid identity feature block has

dimensions:

    L rows x 20 columns

The feature columns are concatenated into the final residue feature matrix

immediately after the three positional features.