The paper "DeepDTA: deep drug–target binding affinity prediction" by Hakime Öztürk, Arzucan Özgür, and Elif Ozkirimli proposes a deep learning-based model to predict drug-target binding affinities using only sequence information of both targets and drugs. The authors address the challenge of predicting binding affinities, which is a continuous value representing the strength of interaction between a drug and a target. Traditional methods often focus on binary classification, while this study aims to predict the actual affinity values. The proposed model, DeepDTA, uses Convolutional Neural Networks (CNNs) to learn representations from the raw sequence data of proteins and drugs. The model consists of two separate CNN blocks, each designed to learn representations from SMILES strings and protein sequences, respectively. These representations are then combined into a fully connected layer block. The model is evaluated on two benchmark datasets: the Davis kinase dataset and the KIBA kinase inhibitor bioactivity dataset. The results show that DeepDTA outperforms state-of-the-art methods such as KronRLS and SimBoost, both in terms of Concordance Index (CI) and Mean Squared Error (MSE). Specifically, DeepDTA achieves the best CI performance on the larger KIBA dataset and significantly outperforms the baselines on both datasets. The model also demonstrates lower MSE values, indicating better predictive accuracy. The authors conclude that deep learning-based methodologies perform better than baseline methods, especially when the dataset size increases. They suggest that future work could focus on improving the representation of protein sequences and extending the methodology to predict affinities for novel drug-target pairs.The paper "DeepDTA: deep drug–target binding affinity prediction" by Hakime Öztürk, Arzucan Özgür, and Elif Ozkirimli proposes a deep learning-based model to predict drug-target binding affinities using only sequence information of both targets and drugs. The authors address the challenge of predicting binding affinities, which is a continuous value representing the strength of interaction between a drug and a target. Traditional methods often focus on binary classification, while this study aims to predict the actual affinity values. The proposed model, DeepDTA, uses Convolutional Neural Networks (CNNs) to learn representations from the raw sequence data of proteins and drugs. The model consists of two separate CNN blocks, each designed to learn representations from SMILES strings and protein sequences, respectively. These representations are then combined into a fully connected layer block. The model is evaluated on two benchmark datasets: the Davis kinase dataset and the KIBA kinase inhibitor bioactivity dataset. The results show that DeepDTA outperforms state-of-the-art methods such as KronRLS and SimBoost, both in terms of Concordance Index (CI) and Mean Squared Error (MSE). Specifically, DeepDTA achieves the best CI performance on the larger KIBA dataset and significantly outperforms the baselines on both datasets. The model also demonstrates lower MSE values, indicating better predictive accuracy. The authors conclude that deep learning-based methodologies perform better than baseline methods, especially when the dataset size increases. They suggest that future work could focus on improving the representation of protein sequences and extending the methodology to predict affinities for novel drug-target pairs.