Biology subjects

Parimbelli, E.

Publications and source records attributed to Parimbelli, E..

3 recordsLinked to original sources

bioRxiv2025

Vir2vec: A Genome-Wide Viral Embedding

Genomic language models (gLMs) are powerful numerical surrogates for DNA, but existing architectures focus primarily on human DNA or limited viral sets, and no dedicated benchmark exists for viral genome understanding. Here we introduce Vir2vec (from 17M to 422M parameters), decoder-only gLMs continually pretrained on a curated pan-viral corpus of 565,747 complete genomes across 295 species. We also present the viral Genome Understanding Evaluation (vGUE), a benchmark probing viral representations across broad organism discrimination, evolutionary signatures, intra-genus separation, variant subtyping, and host tropism. Evaluated as frozen feature extractors under nested cross-validation, Vir2vec achieves top balanced accuracy across 5 of 7 vGUE tasks, outperforming human-trained (MistralDNA), viral-specific (ModernBERTvirus), and broad biological foundation (Evo1) model. Furthermore, Vir2vec displays robust zero-shot generalization to unseen viral families. Finally, we introduce an argmax-routed SHAP explainability framework that projects feature attributions directly onto single-codon coordinates, correctly isolating key receptor-binding domain mutations driving mouse host adaptation. Vir2vec and vGUE provide a foundation for viral surveillance and discovery pipelines.

bioinformatics↗

bioRxiv2024

SARITA: A Large Language Model for Generating the S1 Subunit of the SARS-CoV-2 Spike Protein

The COVID-19 pandemic has profoundly impacted global health, economics, and daily life, with over 776 million cases and 7 million deaths from December 2019 to November 2024. Since the original SARS-CoV-2 Wuhan strain emerged, the virus has evolved into variants such as Alpha, Beta, Gamma, Delta, and Omicron, all characterized by mutations in the Spike glycoprotein, critical for viral entry into human cells via its S1 and S2 subunits. The S1 subunit, binding to the ACE2 receptor and mutating frequently, affects infectivity and immune evasion; the more conserved S2, on the other hand, facilitates membrane fusion. Predicting future mutations is crucial for developing vaccines and treatments adaptable to emerging strains, enhancing preparedness and intervention design. Generative Large Language Models (LLMs) are becoming increasingly common in the field of genomics, given their ability to generate realistic synthetic biological sequences, including applications in protein design and engineering. Here we present SARITA, an LLM with up to 1.2 billion parameters, based on GPT-3 architecture, designed to generate high-quality synthetic SARS-CoV-2 Spike S1 sequences. SARITA is trained via continuous learning on the pre-existing protein model RITA. When trained on Alpha, Beta, and Gamma variants (data up to February 2021 included), SARITA correctly predicts the evolution of future S1 mutations, including characterized mutations of Delta, Omicron and Iota variants. Furthermore, we show how SARITA outperforms alternative approaches, including other LLMs, in terms of sequence quality, realism, and similarity with real-world S1 sequences. These results indicate the potential of SARITA to predict future SARS-CoV-2 S1 evolution, potentially aiding in the development of adaptable vaccines and treatments.

bioinformatics↗

bioRxiv2019

Atrous Convolution with transfer learning for Skin Lesions Classification

Skin cancer is a crucial public health issue and by far the most usual kind of cancer specifically in the region of North America. It is estimated that in 2019, only because of melanoma nearly 7,230 people will die, and 192,310 cases of malignant melanoma will be diagnosed. Nonetheless, nearly all types of skin lesions can be treatable if they can be diagnosed at an earlier stage. The accurate prediction of skin lesions is a critically challenging task even for vastly experienced clinicians and dermatologist due to a little distinction between surrounding skin and lesions, visual resemblance between melanoma and other skin lesions, fuddled lesion border, etc. A well-grounded automated computer-aided skin lesions detection system can help clinicians immensely to prognosis malignant skin lesion in the earliest possible time. From the past few years, the emergence of machine learning and deep learning in the medical imaging has produced several image-based classification systems in the medical field and these systems perform better than traditional image processing classification methods. In this paper, we proposed a popular deep learning technique namely atrous or, dilated convolution for skin lesions classification, which is known to have enhanced accuracy with the same amount of computational cost compared to traditional CNN. To implement atrous convolution we choose the transfer learning technique with several popular deep learning architectures such as VGG16, VGG19, MobileNet, and InceptionV3. To train, validate, and test our proposed models we utilize HAM10000 dataset which contains total 10015 dermoscopic images of seven different skin lesions (melanoma, melanocytic nevi, Basal cell carcinoma, Benign keratosis-like lesions, Dermatofibroma, Vascular lesions, and Actinic keratoses). Four of our proposed dilated convolutional frameworks show promising outcome on overall accuracy and per-class accuracy. For example, overall test accuracy achieved 87.42%, 85.02%, 88.22%, and 89.81% on dilated VGG16, dilated VGG19, dilated MobileNet, and dilated IncaptionV3 respectively. These dilated convolutional models outperformed existing networks in both overall accuracy and individual class accuracy. Among all the architectures dilated InceptionV3 shows superior classification accuracy and dilated MobileNet is also achieving almost impressive classification accuracy like dilated InceptionV3 with the lightest computational complexities than all other proposed model. Compared to previous works done on skin lesions classification we have experimented one of the most complicated open-source datasets with class imbalances and achieved better accuracy (dilated inceptionv3) than any known methods to the best of our knowledge.

cancer biology↗