Yoan Martínez-López
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yoan Martínez-López has a total of 47 co-authors. Here are the top co-authors by number of shared publications:
- Juan Alberto Castillo Garit
Instituto Universitario de Investigación y Desarrollo Tecnológico (IDT), Universidad Tecnológica Metropolitana, Ignacio Valdivieso 2409, San Joaquín, Santiago 8940577, Chile
4 shared publications - Julio Madera
University of Camagüey “Ignacio Agramonte Loynaz
4 shared publications Oscar Martínez-Santiago
Alfa Vitamins Laboratories, Miami, FL, 33166, USA
4 shared publications- Francisco Torrens
Institut Universitari de Ciència Molecular, Universitat de Valencia, 46010 Valencia, Spain
3 shared publications Julio Madera
Universidad de Camaguey
3 shared publications- Stephen J. Barigye
Departamento de Química Física Aplicada, Facultad de Ciencias, Universidad Autónoma de Madrid (UAM), 28049 Madrid, Spain
3 shared publications
Latest Publications
- 9 May 2026
- Published by Springer Nature in Molecular Diversity
- Vol. 30 (6) , 9031-9056
Exploring anti-dengue activity with atomic-weighted vectors, class balancing and machine learning
Dengue is a major mosquito-borne viral disease with no effective antiviral treatment currently available. This work introduces a machine-learning framework to predict anti-dengue activity in small molecules using Atomic-Weighted Vector (AWV) descriptors and data-balancing techniques. Sixteen datasets, each containing 2118 molecules, were generated with MD-LOVIs (Molecular Descriptor from Local Vertex Invariants) and preprocessed with IMMAN (Information theory-based CheMoMetric ANalysis), with Shannon entropy applied for feature selection. To address class imbalance (imbalance ratio = 6.66), the ADASYN algorithm was employed. Thirty classifiers spanning six methodological families were evaluated under two validation schemes (tenfold cross-validation and percentage split) on both balanced and imbalanced datasets. Performance was assessed using accuracy (ACC). Nonparametric statistical tests (Friedman, Nemenyi, Wilcoxon) indicated that data balancing improved model robustness. Tree-based and function-based classifiers achieved the best predictive performance. Overall, the proposed workflow offers a reproducible, data-driven approach for virtual screening of anti-dengue compounds and is readily extensible to other antiviral drug discovery tasks.
- 9 May 2026
- Published by Springer Nature in Molecular Diversity
- Vol. 30 (6) , 9031-9056
Exploring anti-dengue activity with atomic-weighted vectors, class balancing and machine learning
Dengue is a major mosquito-borne viral disease with no effective antiviral treatment currently available. This work introduces a machine-learning framework to predict anti-dengue activity in small molecules using Atomic-Weighted Vector (AWV) descriptors and data-balancing techniques. Sixteen datasets, each containing 2118 molecules, were generated with MD-LOVIs (Molecular Descriptor from Local Vertex Invariants) and preprocessed with IMMAN (Information theory-based CheMoMetric ANalysis), with Shannon entropy applied for feature selection. To address class imbalance (imbalance ratio = 6.66), the ADASYN algorithm was employed. Thirty classifiers spanning six methodological families were evaluated under two validation schemes (tenfold cross-validation and percentage split) on both balanced and imbalanced datasets. Performance was assessed using accuracy (ACC). Nonparametric statistical tests (Friedman, Nemenyi, Wilcoxon) indicated that data balancing improved model robustness. Tree-based and function-based classifiers achieved the best predictive performance. Overall, the proposed workflow offers a reproducible, data-driven approach for virtual screening of anti-dengue compounds and is readily extensible to other antiviral drug discovery tasks.
- 3 February 2026
- Published by Springer Nature
- p. 192-203
Evaluation of Probabilistic Data Augmentation Models for Emotion Detection
- 3 February 2026
- Published by Springer Nature
- p. 192-203
Evaluation of Probabilistic Data Augmentation Models for Emotion Detection
- Open Access
- 1 January 2026
- Published by Wolters Kluwer Health in Journal of Vector Borne Diseases
- Vol. 63 (1) , 74-80
Exploring antimalarial activity of drugs using weighted atomic vectors and artificial intelligence
Background & objectives: Malaria is a global health issue, causing over two million deaths annually. The development of new and potent antimalarial drugs is essential to combat the disease. Machine learning has been increasingly applied to predict antimalarial activity of compounds, offering a promising approach for antimalarial pharmaceutical research. This study aims to predict the antimalarial activity of potential compounds using weighted atomic vectors and machine learning algorithms. Methods: The research employs several machine learning algorithms, such as Decision Tree, Bagging Regressor, and Ada Boost. The study uses weighted atomic vectors to represent compounds and employs machine learning algorithms for prediction. The performance of the models is assessed using metrics like R2, MAE, and RMSLE, statistical validation using Friedman and Wilcoxon Tests. Results: The results highlight the remarkable efficacy of Ada Boost in predicting antimalarial activity, consistently outperforming other algorithms across different datasets, achieving a maximum precision of 93%. Interpretation & conclusion: The combination of weighted atomic vectors and machine learning emerges as a promising approach for antimalarial pharmaceutical research, emphasizing the significance of artificial intelligence in this field.
- 1 January 2026
- Published by Wolters Kluwer Health in Journal of Vector Borne Diseases
- Vol. 63 (1) , 74-80
Open AccessExploring antimalarial activity of drugs using weighted atomic vectors and artificial intelligence
Background & objectives: Malaria is a global health issue, causing over two million deaths annually. The development of new and potent antimalarial drugs is essential to combat the disease. Machine learning has been increasingly applied to predict antimalarial activity of compounds, offering a promising approach for antimalarial pharmaceutical research. This study aims to predict the antimalarial activity of potential compounds using weighted atomic vectors and machine learning algorithms. Methods: The research employs several machine learning algorithms, such as Decision Tree, Bagging Regressor, and Ada Boost. The study uses weighted atomic vectors to represent compounds and employs machine learning algorithms for prediction. The performance of the models is assessed using metrics like R2, MAE, and RMSLE, statistical validation using Friedman and Wilcoxon Tests. Results: The results highlight the remarkable efficacy of Ada Boost in predicting antimalarial activity, consistently outperforming other algorithms across different datasets, achieving a maximum precision of 93%. Interpretation & conclusion: The combination of weighted atomic vectors and machine learning emerges as a promising approach for antimalarial pharmaceutical research, emphasizing the significance of artificial intelligence in this field.
- Open Access
- 9 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
UC-UCO-CICESE_UT3-Plenitas Team - Exploring in the PROFE 2025: Language Proficiency Evaluation
PROFE2025 is a challenge that tests automated systems on authentic Spanish reading-comprehension exams used by the Instituto Cervantes for human learners, without providing task-specific training data. This setup emphasizes transfer learning and large generative models. Participants can choose from three subtasks: Multiple Choice, Matching, and Fill-in-the-Gap, each evaluated using accuracy metrics that align with human assessment. Teams completing all subtasks receive an exam-level score based on the proportion of exams passed (60% accuracy), enabling direct comparison to human performance. The competition utilizes the IC-UNED-RC-ES corpus and Hugging Face's transformer model ecosystem. In internal evaluations, three Qwen-3 variants were tested: Qwen 3.0, a "think" prompting version, and a DeepSeek-distilled Qwen 2.5 7B. Both base and distilled models achieved 32.22% overall accuracy, while the "think" variant scored only 10.19%, indicating room for improvement in reasoning-focused prompts. Among 40 submissions, only two teams completed all three subtasks. UC-CICESE_UT3-Plenitas placed second with 32.22%, 4.89%, and 10.19%, respectively, underscoring both potential and current limitations of LLMs in this domain. PROFE2025 highlights that large language models can engage meaningfully with complex, human-like exam conditions, but significant improvements are possible through specialized adaptation, fine-tuning, and advanced prompting strategies, positioning it as a key benchmark for progress in automated language understanding..
- 9 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessUC-UCO-CICESE_UT3-Plenitas Team - Exploring in the PROFE 2025: Language Proficiency Evaluation
PROFE2025 is a challenge that tests automated systems on authentic Spanish reading-comprehension exams used by the Instituto Cervantes for human learners, without providing task-specific training data. This setup emphasizes transfer learning and large generative models. Participants can choose from three subtasks: Multiple Choice, Matching, and Fill-in-the-Gap, each evaluated using accuracy metrics that align with human assessment. Teams completing all subtasks receive an exam-level score based on the proportion of exams passed (60% accuracy), enabling direct comparison to human performance. The competition utilizes the IC-UNED-RC-ES corpus and Hugging Face's transformer model ecosystem. In internal evaluations, three Qwen-3 variants were tested: Qwen 3.0, a "think" prompting version, and a DeepSeek-distilled Qwen 2.5 7B. Both base and distilled models achieved 32.22% overall accuracy, while the "think" variant scored only 10.19%, indicating room for improvement in reasoning-focused prompts. Among 40 submissions, only two teams completed all three subtasks. UC-CICESE_UT3-Plenitas placed second with 32.22%, 4.89%, and 10.19%, respectively, underscoring both potential and current limitations of LLMs in this domain. PROFE2025 highlights that large language models can engage meaningfully with complex, human-like exam conditions, but significant improvements are possible through specialized adaptation, fine-tuning, and advanced prompting strategies, positioning it as a key benchmark for progress in automated language understanding..
- Open Access
- 24 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
UC-UCO-CICESE_UT3-Plenitas Team at HOMO-LAT 2025: Exploring the Challenge of Determining the Opinion of Latin-American Speakers about LGBTQ+ Community
The HOMO-LAT25 shared task focuses on detecting sentiment polarity in online messages related to the LGBTQ+ community across Latin American Spanish dialects. The UC-UCO-CICESE_UT3-Plenitas team participated in two main subtasks: multi-dialect polarity detection-where the training and evaluation dialects match-and crossdialect polarity detection-where they differ. These subtasks challenge systems to recognize positive, negative, or neutral sentiment toward specific keywords across varying sociolinguistic contexts. The team employed transformer-based models from the Hugging Face ecosystem, particularly leveraging the Robertuito model from the pysentimiento library, which is pre-trained for Spanish sentiment classification. The overall methodology relied on deep learning techniques to manage unstructured social media text, emphasizing the adaptability of the transformers for multilingual sentiment analysis. Out of 31 teams and 67 total submissions, only 8 teams progressed to the final evaluation phase. The UC-UCO-CICESE_UT3-Plenitas team reached the finals and placed 7th in both subtasks. Their F1-scores were 0.2592 and 0.2588, respectively-demonstrating baseline performance but highlighting a clear gap from top-performing teams. Despite the modest scores, their progression to the final stage reflects the model's initial competence and provides a solid foundation for future enhancements. Key areas for improvement include better handling of dialectal variation, improved feature engineering, and optimization of classification strategies. Overall, the HOMO-LAT25 task represents a critical benchmark for developing inclusive, culturally aware NLP systems, supporting applications such as content moderation, public sentiment tracking, and digital policy efforts aimed at protecting LGBTQ+ communities in Latin America.
- 24 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessUC-UCO-CICESE_UT3-Plenitas Team at HOMO-LAT 2025: Exploring the Challenge of Determining the Opinion of Latin-American Speakers about LGBTQ+ Community
The HOMO-LAT25 shared task focuses on detecting sentiment polarity in online messages related to the LGBTQ+ community across Latin American Spanish dialects. The UC-UCO-CICESE_UT3-Plenitas team participated in two main subtasks: multi-dialect polarity detection-where the training and evaluation dialects match-and crossdialect polarity detection-where they differ. These subtasks challenge systems to recognize positive, negative, or neutral sentiment toward specific keywords across varying sociolinguistic contexts. The team employed transformer-based models from the Hugging Face ecosystem, particularly leveraging the Robertuito model from the pysentimiento library, which is pre-trained for Spanish sentiment classification. The overall methodology relied on deep learning techniques to manage unstructured social media text, emphasizing the adaptability of the transformers for multilingual sentiment analysis. Out of 31 teams and 67 total submissions, only 8 teams progressed to the final evaluation phase. The UC-UCO-CICESE_UT3-Plenitas team reached the finals and placed 7th in both subtasks. Their F1-scores were 0.2592 and 0.2588, respectively-demonstrating baseline performance but highlighting a clear gap from top-performing teams. Despite the modest scores, their progression to the final stage reflects the model's initial competence and provides a solid foundation for future enhancements. Key areas for improvement include better handling of dialectal variation, improved feature engineering, and optimization of classification strategies. Overall, the HOMO-LAT25 task represents a critical benchmark for developing inclusive, culturally aware NLP systems, supporting applications such as content moderation, public sentiment tracking, and digital policy efforts aimed at protecting LGBTQ+ communities in Latin America.
- 1 November 2024
- Published by Springer Nature in Journal of Molecular Modeling
- Vol. 30 (11) , 1-15
Exploring blood–brain barrier passage using atomic weighted vector and machine learning
Context This study investigates the potential of leveraging molecular properties, as determined by MD-LOVIs software and machine learning techniques, to predict the ability of compounds to cross the blood–brain barrier (BBB). Accurate prediction of BBB permeation is critical for the development of central nervous system (CNS) drugs. The study applies various machine learning models, including both classification and regression techniques, to predict BBB passage and molecular activity. Notably, classification models such as GBM-AWV (accuracy = 0.801), GLM-CN (accuracy = 0.808), SVMPoly-means (accuracy = 0.980), SVMPoly-AC (accuracy = 0.980), SVMPoly-MI_TI_SI (accuracy = 0.900), SVMPoly-GI (accuracy = 0.900), RF-means (accuracy = 0.870), and GLM-means (accuracy = 0.818) demonstrate high accuracy in predicting BBB passage. In contrast, regression models like ES-RLM-AG (R2 = 0.902), IB-IBK (R2 = 0.82), IB-Kstar (R2 = 0.834), IB-MLP (R2 = 0.843), and DRF-AWV (R2 = 0.810) exhibit strong performance in predicting molecular activity. The results show that classification models like GBM-AWV, GLM-CN, and SVMPoly variants, as well as regression models like ES-RLM-AG and IB-MLP, achieve high performance, demonstrating the effectiveness of machine learning in predicting BBB permeability. Methods The computational methods employed in this study include the MD-LOVIs software for generating molecular descriptors and several machine learning algorithms, including gradient boosting machines (GBM), generalized linear models (GLM), support vector machines (SVM) with polynomial kernels, random forests (RF), ensemble regression models, and instance-based learning algorithms. These models were trained and validated using various datasets to predict BBB passage and molecular activity, with the performance metrics reported for each model. Standard computational techniques were employed throughout, ensuring the reliability of the predictions. Graphical Abstract
- 1 November 2024
- Published by Springer Nature in Journal of Molecular Modeling
- Vol. 30 (11) , 1-15
Exploring blood–brain barrier passage using atomic weighted vector and machine learning
Context This study investigates the potential of leveraging molecular properties, as determined by MD-LOVIs software and machine learning techniques, to predict the ability of compounds to cross the blood–brain barrier (BBB). Accurate prediction of BBB permeation is critical for the development of central nervous system (CNS) drugs. The study applies various machine learning models, including both classification and regression techniques, to predict BBB passage and molecular activity. Notably, classification models such as GBM-AWV (accuracy = 0.801), GLM-CN (accuracy = 0.808), SVMPoly-means (accuracy = 0.980), SVMPoly-AC (accuracy = 0.980), SVMPoly-MI_TI_SI (accuracy = 0.900), SVMPoly-GI (accuracy = 0.900), RF-means (accuracy = 0.870), and GLM-means (accuracy = 0.818) demonstrate high accuracy in predicting BBB passage. In contrast, regression models like ES-RLM-AG (R2 = 0.902), IB-IBK (R2 = 0.82), IB-Kstar (R2 = 0.834), IB-MLP (R2 = 0.843), and DRF-AWV (R2 = 0.810) exhibit strong performance in predicting molecular activity. The results show that classification models like GBM-AWV, GLM-CN, and SVMPoly variants, as well as regression models like ES-RLM-AG and IB-MLP, achieve high performance, demonstrating the effectiveness of machine learning in predicting BBB permeability. Methods The computational methods employed in this study include the MD-LOVIs software for generating molecular descriptors and several machine learning algorithms, including gradient boosting machines (GBM), generalized linear models (GLM), support vector machines (SVM) with polynomial kernels, random forests (RF), ensemble regression models, and instance-based learning algorithms. These models were trained and validated using various datasets to predict BBB passage and molecular activity, with the performance metrics reported for each model. Standard computational techniques were employed throughout, ensuring the reliability of the predictions. Graphical Abstract
- Open Access
- 15 July 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Optimizing Energy Operation and Planning using Ring Cellular Encode-Decode Univariate Marginal Distribution Algorithm
- 15 July 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-2
Open AccessOptimizing Energy Operation and Planning using Ring Cellular Encode-Decode Univariate Marginal Distribution Algorithm
- 5 April 2023
- Published by Springer Nature in Molecular Diversity
- Vol. 28 (4) , 1983-1994
Exploring proteasome inhibition using atomic weighted vector indices and machine learning approaches
Ubiquitin–proteasome system (UPS) is a highly regulated mechanism of intracellular protein degradation and turnover. The UPS is involved in different biological activities, such as the regulation of gene transcription and cell cycle. Several researchers have applied cheminformatics and artificial intelligence methods to study the inhibition of proteasomes, including the prediction of UPP inhibitors. Following this idea, we applied a new tool for obtaining molecular descriptors (MDs) for modeling proteasome Inhibition in terms of EC50 (µmol/L), in which a set of new MDs called atomic weighted vectors (AWV) and several prediction algorithms were used in cheminformatics studies. In the manuscript, a set of descriptors based on AWV are presented as datasets for training different machine learning techniques, such as linear regression, multiple linear regression (MLR), random forest (RF), K-nearest neighbors (IBK), multi-layer perceptron, best-first search, and genetic algorithm. The results suggest that these atomic descriptors allow adequate modeling of proteasome inhibitors despite artificial intelligence techniques, as a variant to build efficient models for the prediction of inhibitory activity. Graphical abstract
- 5 April 2023
- Published by Springer Nature in Molecular Diversity
- Vol. 28 (4) , 1983-1994
Exploring proteasome inhibition using atomic weighted vector indices and machine learning approaches
Ubiquitin–proteasome system (UPS) is a highly regulated mechanism of intracellular protein degradation and turnover. The UPS is involved in different biological activities, such as the regulation of gene transcription and cell cycle. Several researchers have applied cheminformatics and artificial intelligence methods to study the inhibition of proteasomes, including the prediction of UPP inhibitors. Following this idea, we applied a new tool for obtaining molecular descriptors (MDs) for modeling proteasome Inhibition in terms of EC50 (µmol/L), in which a set of new MDs called atomic weighted vectors (AWV) and several prediction algorithms were used in cheminformatics studies. In the manuscript, a set of descriptors based on AWV are presented as datasets for training different machine learning techniques, such as linear regression, multiple linear regression (MLR), random forest (RF), K-nearest neighbors (IBK), multi-layer perceptron, best-first search, and genetic algorithm. The results suggest that these atomic descriptors allow adequate modeling of proteasome inhibitors despite artificial intelligence techniques, as a variant to build efficient models for the prediction of inhibitory activity. Graphical abstract
- 9 September 2020
- Published by Springer Nature
- p. 207-215
SMOTE-Cov: A New Oversampling Method Based on the Covariance Matrix
- 9 September 2020
- Published by Springer Nature
- p. 207-215
SMOTE-Cov: A New Oversampling Method Based on the Covariance Matrix
- Open Access
- 20 November 2018
- Published by MDPI AG
Predicting Proteasome Inhibition using Atomic Weighted Vector and Machine Learning
Ubiquitin/Proteasome System (UPS) is a highly regulated mechanism of intracellular protein degradation and turnover. Through the concerted actions of a series of enzymes, proteins are marked for proteasomal degradation by being linked to the polypeptide co-factor, ubiquitin. The UPS participates in a wide array of biological functions such as antigen presentation, regulation of gene transcription and the cell cycle, and activation of NF-κB. Some researchers have applied QSAR method and machine learning in the study of proteasome inhibition (EC50(µmol/L)), such as: the analysis of proteasome inhibition prediction, in the prediction of multi-target inhibitors of UPP and in the prediction of protein contact map. Following this idea, we applied the new tool for obtaining molecular descriptor for modeling of proteasome Inhibition EC50 (µmol/L), in which has used this novel molecular descriptors (MDs) and different classification algorithms for these quantitative structure-activity relationship (QSAR) studies. In the present research, we use the Atomic Weighted Vector (AWV) as attributes with the objective to develop the QSAR modeling of this datasets and also compare a set of different machine learning (ML) techniques to solve this problem, such as: Linear Regression (LR), Multiple linear regression (MLR), Decision tree(DT), Regression Tree(RT), Random Forest(RF), M5P, K-nearest neighbors (IBK or kNN), Multi-Layer perceptron (MLP), Best-first search (BF) and Genetic Algorithm (GA). The figure shows the results of R2 of the ML-QSAR using ten- folds cross validation for 258 compounds. The results indicate that AWVs are very important tool for modeling the proteasome inhibitory regardless of the ML algorithm used. It can be suggested that the MD-AWV are suitable for codifying important structural information of the molecules and, thus, constitute an interesting alternative to building effective models for the prediction of the values of EC50 (µmol/L).
- 20 November 2018
- Published by MDPI AG
Open AccessPredicting Proteasome Inhibition using Atomic Weighted Vector and Machine Learning
Ubiquitin/Proteasome System (UPS) is a highly regulated mechanism of intracellular protein degradation and turnover. Through the concerted actions of a series of enzymes, proteins are marked for proteasomal degradation by being linked to the polypeptide co-factor, ubiquitin. The UPS participates in a wide array of biological functions such as antigen presentation, regulation of gene transcription and the cell cycle, and activation of NF-κB. Some researchers have applied QSAR method and machine learning in the study of proteasome inhibition (EC50(µmol/L)), such as: the analysis of proteasome inhibition prediction, in the prediction of multi-target inhibitors of UPP and in the prediction of protein contact map. Following this idea, we applied the new tool for obtaining molecular descriptor for modeling of proteasome Inhibition EC50 (µmol/L), in which has used this novel molecular descriptors (MDs) and different classification algorithms for these quantitative structure-activity relationship (QSAR) studies. In the present research, we use the Atomic Weighted Vector (AWV) as attributes with the objective to develop the QSAR modeling of this datasets and also compare a set of different machine learning (ML) techniques to solve this problem, such as: Linear Regression (LR), Multiple linear regression (MLR), Decision tree(DT), Regression Tree(RT), Random Forest(RF), M5P, K-nearest neighbors (IBK or kNN), Multi-Layer perceptron (MLP), Best-first search (BF) and Genetic Algorithm (GA). The figure shows the results of R2 of the ML-QSAR using ten- folds cross validation for 258 compounds. The results indicate that AWVs are very important tool for modeling the proteasome inhibitory regardless of the ML algorithm used. It can be suggested that the MD-AWV are suitable for codifying important structural information of the molecules and, thus, constitute an interesting alternative to building effective models for the prediction of the values of EC50 (µmol/L).