DNA Sequence Polishing Using Deep learning

Cilj ovog rada bio je implementirati model temeljen na dubokom uˇcenju koji bi pove´cao toˇcnost sastavljenog genoma. Konstruirani model temelji se na arhitekturi rezidualne neuronske mreže te postiže toˇcnost ve´cu od trenutno najboljih alata temeljenih na klasiˇcnom raˇcunarstvu, no zaostaje ne nadmašuje najbolje alate temeljene na dubokom ucˇenju. Med¯utim, zbog ogranicˇenja u vidu prostora za pohranu podataka i vremena treniranja modela, detaljnija analiza tek treba biti napravljena za više bakterijskih uzoraka i razliˇcite arhitekture mreže.Goal of this work was to implement a deep learning model in order to increase the accuracy of the assembled genome. This model is based on a residual network architecture and achieves higher accuracy than classical state-of-the-art tools, but falls behind the best deep-learning tools available. However, due to limitations in terms of storage and time for training the model, further analysis has to be conducted for different bacteria datasets and network architectures