Naslov
Predviđanje interakcije među proteinima korištenjem konvolucijske neuronske mreže
(Protein-protein interaction prediction using convolutional neural network)

Autori
Segarić, Lucija

Vrsta, podvrsta i kategorija rada
Ocjenski radovi, diplomski rad, diplomski

Fakultet
Fakultet elektrotehnike i računarstva

Mjesto
Zagreb

Datum
07.07

Godina
2020

Stranica
37

Mentor
Domazet-Lošo, Mirjana

Ključne riječi
protein-protein interaction ; deep learning ; convolutional neural network

Sažetak
The development of high-throughput experimental techniques has enabled discovery of a large amount of protein-protein interaction (PPI) data. Since those methods are expensive and time-consuming, many computational methods for PPI prediction have been developed and show promising results. In this thesis, a deep learning model based on Siamese convolutional neural network for prediction of PPIs from amino acid sequences was studied and built. The model was implemented in Lua programming language with Torch framework which has a wide support for machine learning and deep learning algorithms. I evaluated it on three different datasets and it shows good results on two of those datasets, with best performance on Multi-species dataset, where accuracy and F1 scores of 0.72 were achieved.

Izvorni jezik
Engleski

Znanstvena područja
Računarstvo

POVEZANOST RADA