Promoter Classification in Human Genome via DNA2Vec and UNK-Aware Deep Neural Networks

Aleyna Mengen; Emre Delibaş

Makale detayları

Promoter Classification in Human Genome via DNA2Vec and UNK-Aware Deep Neural Networks

Yazarlar:

Aleyna Mengen | Emre Delibaş

Yayıncı:

Sivas Cumhuriyet University

Ana Başlıklar:

Veri Mühendisliği ve Veri Bilimi

Dergi:

Sivas Cumhuriyet Üniversitesi Mühendislik Fakültesi Dergisi Vol.3 Issue 1, pp.92-99

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001353122

pISSN:

3023-7203

Anahtar kelimeler:

Özet:

 This study proposes a new hybrid model combining DNA2Vec-based embedded representations with UNK character support and a deep neural network DNN architecture for the classification of promoter and non-promoter DNA sequences belonging to the Homo sapiens genome. The models objective is twofold: first, to minimize the loss of contextual information, and second, to enhance the generalization performance by representing unknown or low-confidence k-mer sequences with an UNK vector. The model, which was structured with a GELU activation function and an AdamW optimization algorithm, achieved strong and balanced results, including 85.03 accuracy, 0.9376 ROC-AUC, and 0.8444 F1 score, when evaluated using a stratified 5-fold cross-validation method. The findings indicate that the proposed structure provides a more straightforward yet effective approach in comparison to the more complex models documented in the extant literature. Furthermore, this architecture provides pragmatic and comprehensible solutions in bioinformatics applications, particularly since it facilitates motif-independent learning. Future work should address the generalization capacity be increased across species and that the integration with Transformer-based models be evaluated in future studies.

 This study proposes a new hybrid model combining DNA2Vec-based embedded representations with UNK character support and a deep neural network DNN architecture for the classification of promoter and non-promoter DNA sequences belonging to the Homo sapiens genome. The models objective is twofold: first, to minimize the loss of contextual information, and second, to enhance the generalization performance by representing unknown or low-confidence k-mer sequences with an UNK vector. The model, which was structured with a GELU activation function and an AdamW optimization algorithm, achieved strong and balanced results, including 85.03 accuracy, 0.9376 ROC-AUC, and 0.8444 F1 score, when evaluated using a stratified 5-fold cross-validation method. The findings indicate that the proposed structure provides a more straightforward yet effective approach in comparison to the more complex models documented in the extant literature. Furthermore, this architecture provides pragmatic and comprehensible solutions in bioinformatics applications, particularly since it facilitates motif-independent learning. Future work should address the generalization capacity be increased across species and that the integration with Transformer-based models be evaluated in future studies.

 This study proposes a new hybrid model combining DNA2Vec-based embedded representations with UNK character support and a deep neural network DNN architecture for the classification of promoter and non-promoter DNA sequences belonging to the Homo sapiens genome. The models objective is twofold: first, to minimize the loss of contextual information, and second, to enhance the generalization performance by representing unknown or low-confidence k-mer sequences with an UNK vector. The model, which was structured with a GELU activation function and an AdamW optimization algorithm, achieved strong and balanced results, including 85.03 accuracy, 0.9376 ROC-AUC, and 0.8444 F1 score, when evaluated using a stratified 5-fold cross-validation method. The findings indicate that the proposed structure provides a more straightforward yet effective approach in comparison to the more complex models documented in the extant literature. Furthermore, this architecture provides pragmatic and comprehensible solutions in bioinformatics applications, particularly since it facilitates motif-independent learning. Future work should address the generalization capacity be increased across species and that the integration with Transformer-based models be evaluated in future studies.

 Bu çalışma, Homo sapiens genomunda promoter ve non-promoter DNA dizilerinin ayrımını sağlamak amacıyla DNA2Vec tabanlı gömülü temsiller ve UNK karakter duyarlılığı ile güçlendirilmiş derin sinir ağı DNN mimarisini bir araya getiren hibrit bir sınıflandırma yaklaşımı önermektedir. Model, bilinmeyen veya düşük güvenilirlikteki k-merleri özel olarak başlatılan UNK vektörü ile temsil ederek bağlamsal bilgi kaybını önlemekte ve genelleme kapasitesini artırmaktadır. Veri seti, eşit sayıda promoter ve non-promoter diziden oluşturulmuş, değerlendirmede stratified 5-fold çapraz doğrulama uygulanmıştır. Optimize edilen model; test setinde 85.03 doğruluk, 0.8786 kesinlik, 0.8128 duyarlılık, 0.8444 F1 skoru ve 0.9376 ROC-AUC başarısı elde etmiş ve insan genomu üzerinde yapılan çalışmalarda literatürdeki pek çok karmaşık modele kıyasla daha iyi veya benzer sonuçlar göstermiştir. Sonuçlar, önerilen mimarinin güçlü, yorumlanabilir ve hesaplama açısından verimli bir alternatif sunduğunu ve motif-bağımsız öğrenme yeteneğiyle biyoinformatik uygulamalarda pratik olarak kullanılabileceğini göstermektedir. Gelecek çalışmalarda türler arası genelleme ve Transformer gibi dikkat tabanlı modellerle entegrasyonun araştırılması önerilmektedir.

Makale detayları

Promoter Classification in Human Genome via DNA2Vec and UNK-Aware Deep Neural Networks

Yazarlar:

Aleyna Mengen | Emre Delibaş

Yayıncı:

Sivas Cumhuriyet University

Ana Başlıklar:

Veri Mühendisliği ve Veri Bilimi

Dergi:

Sivas Cumhuriyet Üniversitesi Mühendislik Fakültesi Dergisi Vol.3 Issue 1, pp.92-99

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001353122

pISSN:

3023-7203

Promoter Classification

DNA2Vec

Deep Neural Network

UNK Vector

BioinformaticsPromoter Sınıflandırması

DNA2Vec

Derin Sinir Ağı

UNK Vektörü

Biyoinformatik

Sayı

Makale detayları

Yazarlar: Aleyna Mengen | Emre Delibaş

Yayıncı: Sivas Cumhuriyet University

Ana Başlıklar: Veri Mühendisliği ve Veri Bilimi

Dergi: Sivas Cumhuriyet Üniversitesi Mühendislik Fakültesi Dergisi Vol.3 Issue 1, pp.92-99

Belge Tipi: Makale

Yayın Tarihi: 6/2025

Belge Dili: İngilizce

Document ID (Org No.): 2001353122

pISSN: 3023-7203

Sayı

Yazarlar:

Aleyna Mengen | Emre Delibaş

Yayıncı:

Sivas Cumhuriyet University

Ana Başlıklar:

Veri Mühendisliği ve Veri Bilimi

Dergi:

Sivas Cumhuriyet Üniversitesi Mühendislik Fakültesi Dergisi Vol.3 Issue 1, pp.92-99

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001353122

pISSN:

3023-7203