Kutatószeminárium

Classifying DNA sequences based on their protein binding profiles using Deep Learning has enjoyed considerable success in recent years. Meanwhile, adversarial attacks against machine learning models have revealed serious issues in the fields of image- and natural language processing. However, examining the effect of shifted sequences regarding Transcription Factor Binding Sites and deep learning models shows that even a simple lengthwise translation can affect performance. We evaluated two architectures using two data sets with three shifting strategies and trained robust models with adversarial data augmentation. In the seminar I will explore the examined approaches and discuss some of the benefits and issues that adversarial training might mean to Transcription Factor Binding Site classification.