Artificial Neural Networks Thomas Nordahl Petersen & Morten Nielsen

Use of artificial neural networks A data-driven method to predict a feature, given a set of training data In biology input features could be amino acid sequence or nucleotides Secondary structure prediction Signal peptide prediction Surface accessibility Propeptide prediction C N Signal peptide Propeptide Mature/active protein

Neural network prediction methods http://www.cbs.dtu.dk/services/

Pattern recognition

Biological Neural network

Biological neuron structure

Diversity of interactions in a network enables complex calculations Similar in biological and artificial systems Excitatory (+) and inhibitory (-) relations between compute units 1 fire

Transfer of biological principles to artificial neural network algorithms Non-linear relation between input and output Massively parallel information processing Data-driven construction of algorithms Ability to generalize to new data items google translate (https://www.youtube.com/watch?v=0zKU7jDA2nc)

Sparse encoding of amino acid sequence windows

Sparse encoding Inp Neuron 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 AAcid A 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 R 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 N 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 D 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 C 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Q 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 E 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0

BLOSUM encoding (Blosum50 matrix) A R N D C Q E G H I L K M F P S T W Y V A 4 -1 -2 -2 0 -1 -1 0 -2 -1 -1 -1 -1 -2 -1 1 0 -3 -2 0 R -1 5 0 -2 -3 1 0 -2 0 -3 -2 2 -1 -3 -2 -1 -1 -3 -2 -3 N -2 0 6 1 -3 0 0 0 1 -3 -3 0 -2 -3 -2 1 0 -4 -2 -3 D -2 -2 1 6 -3 0 2 -1 -1 -3 -4 -1 -3 -3 -1 0 -1 -4 -3 -3 C 0 -3 -3 -3 9 -3 -4 -3 -3 -1 -1 -3 -1 -2 -3 -1 -1 -2 -2 -1 Q -1 1 0 0 -3 5 2 -2 0 -3 -2 1 0 -3 -1 0 -1 -2 -1 -2 E -1 0 0 2 -4 2 5 -2 0 -3 -3 1 -2 -3 -1 0 -1 -3 -2 -2 G 0 -2 0 -1 -3 -2 -2 6 -2 -4 -4 -2 -3 -3 -2 0 -2 -2 -3 -3 H -2 0 1 -1 -3 0 0 -2 8 -3 -3 -1 -2 -1 -2 -1 -2 -2 2 -3 I -1 -3 -3 -3 -1 -3 -3 -4 -3 4 2 -3 1 0 -3 -2 -1 -3 -1 3 L -1 -2 -3 -4 -1 -2 -3 -4 -3 2 4 -2 2 0 -3 -2 -1 -2 -1 1 K -1 2 0 -1 -3 1 1 -2 -1 -3 -2 5 -1 -3 -1 0 -1 -3 -2 -2 M -1 -1 -2 -3 -1 0 -2 -3 -2 1 2 -1 5 0 -2 -1 -1 -1 -1 1 F -2 -3 -3 -3 -2 -3 -3 -3 -1 0 0 -3 0 6 -4 -2 -2 1 3 -1 P -1 -2 -2 -1 -3 -1 -1 -2 -2 -3 -3 -1 -2 -4 7 -1 -1 -4 -3 -2 S 1 -1 1 0 -1 0 0 0 -1 -2 -2 0 -1 -2 -1 4 1 -3 -2 -2 T 0 -1 0 -1 -1 -1 -1 -2 -2 -1 -1 -1 -1 -2 -1 1 5 -2 -2 0 W -3 -3 -4 -4 -2 -2 -3 -2 -2 -3 -2 -3 -1 1 -4 -3 -2 11 2 -3 Y -2 -2 -2 -3 -2 -1 -2 -3 2 -1 -1 -2 -1 3 -3 -2 -2 2 7 -1 V 0 -3 -3 -3 -1 -2 -2 -3 -3 3 1 -2 1 -1 -2 -2 0 -3 -1 4

Sequence encoding (continued) Sparse encoding V:0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 L:0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 V.L=0 (unrelated) Blosum encoding V: 0 -3 -3 -3 -1 -2 -2 -3 -3 3 1 -2 1 -1 -2 -2 0 -3 -1 4 L:-1 -2 -3 -4 -1 -2 -3 -4 -3 2 4 -2 2 0 -3 -2 -1 -2 -1 1 V.L = 0.88 (highly related) V.R = -0.08 (close to unrelated)

Simplified feed-forward neural network no bias neurons are shown I1 I2 I3 Input w1,2 w3,1 w2,2 w2,1 w1,1 w3,2 h1 h2 hidden v1,1 v2,1 O1 output h1  h1  = 1/ (1+e-x) o=H1*v1,1 + H2*v2,1 O1 = (o) Error = O - True

Sigmodial or logistic function

Training and error reduction 

Training and error reduction 

Training and error reduction Size matters 

Secondary Structure Elements ß-strand Helix Bend Turn

Neural Network Architecture Weights Input Layer I K H E Output Layer E E C H V I I Q A E Hidden Layer Window IKEEHVIIQAEFYLNPDQSGEF…..

Predictions and reliability of a prediction Normally the best prediction is obtained by averaging results from several predictions - “wisdom of the crowd Two types of neural networks Prediction of features in classes/bins e.g. H, E or C (1,0,0) Values close to 1 or 0 are more accurate than values close to 1/2 Prediction of real values e.g. Surface accessibility (0.43) Reliability of a prediction is more difficult to estimate

Signal peptide http://www.cbs.dtu.dk/services/SignalP

Eukaryotic SP & TM Signal peptide cleavage 1523 seq C-terminal end of TM-regions 669 seq 23

Signal peptide is present Signal peptide prediction Signal pepdide likeness Cleavage site Combined information Signal peptide is present If D-score is above threshold

Protein Engineering, Design and Selection: 17: 107-112, 2004. Propeptide prediction Many secretory proteins and peptides are synthesized as inactive precursors that in addition to signal peptide cleavage undergo post-translational processing to become biologically active polypeptides. Precursors are usually cleaved at sites composed of single or paired basic amino acid residues by members of the subtilisin/kexin-like proprotein convertase (PC) family. In mammals, seven members have been identified, with furin being the one first discovered and best characterized. Recently, the involvement of furin in diseases ranging from Alzheimer's disease and cancer to anthrax and Ebola fever has created additional focus on proprotein processing. We have developed a method for prediction of cleavage sites for PCs based on artificial neural networks. Two different types of neural networks have been constructed: a furin-specific network based on experimental results derived from the literature, and a general PC-specific network trained on data from the Swiss-Prot protein database. The method predicts cleavage sites in independent sequences with a sensitivity of 95% for the furin neural network and 62% for the general PC network. Protein Engineering, Design and Selection: 17: 107-112, 2004. General cleavage: R/K-Xn-R/K , n=0, 2, 4, 6 Furin cleavage: R-X-R/K-R

Propeptide prediction http://www.cbs.dtu.dk/services/ProP/ Furin cleavage

NetSurfP http://www.cbs.dtu.dk/services/NetSurfP

http://www.cbs.dtu.dk/services/NetSurfP a b c a = a-helix b=b-strand c=coil