Biology subjects

Paccanaro, A.

Publications and source records attributed to Paccanaro, A..

11 recordsLinked to original sources

bioRxiv2026

Stereochemistry-Aware Drug-Target Affinity Prediction

Drug-target affinity (DTA) prediction is a key task in drug discovery, enabling the estimation of the interaction strength between candidate compounds and biological targets. However, current models rely on connectivity-based molecular representations and do not explicitly account for the spatial organization, also known as stereochemistry. This limitation becomes evident when considering chirality, where a drug can exist as enantiomers, i.e., molecules that share the same atoms and bonds but differ in their three-dimensional arrangement. Despite their chemical similarity, they can interact differently with the same target, leading to variations in binding affinity and biological activity. In this paper, we propose a stereochemistry-aware DTA prediction framework that incorporates this information into molecular representations. Drug representations are learned from chemical structure using a directed-bond message passing graph neural network that captures enantiomers configurations, while protein targets are represented through sequence-based embeddings. Experiments on the Davis dataset demonstrate that our model can improve affinity prediction. Importantly, a case study on a manually curated dataset of enantiomers with different biological action shows that the model is able to distinguish the affinities in the two forms consistent with their experimentally observed biological activity. These findings support the relevance of stereochemistry-aware molecular representation for more accurate and chemically faithful DTA prediction.

bioinformatics↗

bioRxiv2026

Learning Chirality-Aware Representations to Predict Drug Side Effect Frequencies

Ab initio prediction of side effect frequencies is important for assessing the risk-benefit profile of drugs and for identifying potential adverse effects early in development. A key challenge is chirality: many drugs exist as enantiomers, pairs of molecules with the same atoms and bond connectivity but different three-dimensional arrangements. Although chemically similar, enantiomers can interact differently with biological targets and therefore exhibit distinct efficacy and adverse-effect profiles. Here we introduce F2S (Features to Signatures), a method to predict the frequencies of drug side effects while explicitly accounting for chirality. Drug representations are learned directly from chemical structure using a directed-bond message-passing graph neural network that captures stereochemical configurations. Side effect representations are derived from curated textual descriptions encoded with a frozen PubMedBERT model. Side effect frequencies are predicted from the dot product between drug and side effect signatures together with biases for drugs and side effects. We evaluated F2S extensively across multiple settings, including cold-start and warm-start prediction, prospective evaluation, and scenarios controlling for chemical similarity between training and test drugs. Across these evaluations, F2S achieves performance comparable to state-of-the-art methods for general side-effect frequency prediction while producing fewer false positives and substantially improves the prediction of frequency differences between enantiomer pairs. Finally, F2S learns compact 10-dimensional signatures that support interpretability: drug signatures reflect therapeutic class and shared targets, side-effect signatures capture phenotype similarity, and the learned bias terms correlate with the popularity of drugs and side effects.

bioinformatics↗

bioRxiv2025

Robust prediction of drug combination side effects in realistic settings

Side effects caused by drug combinations pose a major challenge in healthcare. Knowledge of these side effects is limited because often they are not detected in clinical trials, which typically involve a restricted number of participants and tested drug combinations. We introduce DCSE (Drug Combinations Side Effects), a novel machine learning method for predicting polypharmacy side effects. DCSE learns latent signatures for drugs, drug pairs, and side effects to predict the probability that a side effect occurs in a given drug combination. We first evaluate its performance in the commonly adopted experimental settings in the literature. Then, a key contribution of this paper is the introduction of more realistic experimental settings that incorporate warm-start and cold-start scenarios under a prospective evaluation. Here, we attempt to predict side effects reported between 2009 and 2014 after training only on data available prior to that period. Our results indicate that DCSE consistently outperforms state-of-the-art methods, demonstrating its robustness and efficacy in real-world applications. Author summaryUnderstanding the risks of combining medications is a central problem in healthcare. Clinical trials rarely capture the full range of adverse reactions, and many side effects are identified only after drugs reach the market, leaving many possible drug combinations insufficiently tested. Computational approaches can help fill this gap by predicting which combinations are likely to cause specific side effects. Our approach works by learning signatures for both drugs and side effects, which can be thought of as representations that describe how drugs act and what is required to trigger a side effect. A key challenge is how to combine the signatures of two drugs into one representation for the pair. We show that simple linear models fail to capture the complexity of drug interactions, so we use a deep neural network to combine them in a non-linear way. Our tests, which use reports collected at different points in time to mirror how information about drug pairs becomes available, show that our approach is useful for monitoring risks after drugs have been approved and for exploring new uses of pairs of existing drugs.

bioinformatics↗

bioRxiv2025

AI methods and biologically informed data curation enable accurate RNA m5C prediction

5-methylcytosine (m5C) RNA modifications influence nearly every aspect of RNA metabolism, but their transcriptome wide detection is limited by costly, error-prone assays. To bridge this experimental gap, a wave of AI tools now predicts putative m5C sites in silico. However, most existing approaches prioritize architectural complexity while neglecting data quality, so their reported gains mainly reflect the artifacts inherited from noisy datasets. We inverted this paradigm by constructing a high-confidence, methyltransferase-specific catalog of m5C sites, removing artifacts that confound existing resources. Using this curated corpus, we trained (for the first time in a multiclass setting) three different models (Bi-GRU, CNN, Transformer) to distinguish writer-specific m5C sites from unmethylated cytosines. All AI models converged to similar, nearly optimal, performance (AUPRC > 0.97), and a biologically informed analysis revealed that most errors clustered in unmethylated sites mimicking true positives. By augmenting the training set with these hard-to-predict negatives, mined from millions of unmodified cytosines, the models were forced to exploit more nuanced features such as RNA secondary structure and subtle sequence cues, which sharply reduced transcriptome-wide false positive predictions, and predicted methylated transcripts exhibited strong concordance with known methyltransferase biology. Explainable AI techniques also showed that our AI models effectively capture how sequence mutations disrupt m5C sites, underscoring their potential to prioritize disease-relevant variants. The main findings of our study underscore that AI models can be decisive levers for reliable m5C identification only if fed with curated data and validated through biologically informed computational analysis.

bioinformatics↗

bioRxiv2024

Fine-tuning of conditional Transformers for the generation of functionally characterized enzymes

We introduce Finenzyme, a Protein Language Model (PLM) that employs a multifaceted learning strategy based on transfer learning from a decoder-based Transformer, conditional learning using specific functional keywords, and fine-tuning to model specific Enzyme Commission (EC) categories. Using Finenzyme, we investigate the conditions under which fine-tuning enhances the prediction and generation of EC categories, showing a two-fold perplexity improvement in EC-specific categories compared to a generalist model. Our extensive experimentation shows that Finenzyme generated sequences can be very different from natural ones while retaining similar tertiary structures, functions and chemical kinetics of their natural counterparts. Importantly, the embedded representations of the generated enzymes closely resemble those of natural ones, thus making them suitable for downstream tasks. Finally, we illustrate how Finenzyme can be used in practice to generate enzymes characterized by specific functions using in-silico directed evolution, a computationally inexpensive PLM fine-tuning procedure significantly enhancing and assisting targeted enzyme engineering tasks.

bioinformatics↗

bioRxiv2021

Unlike Chloroquine, mefloquine inhibits SARS-CoV-2 infection in physiologically relevant cells and does not induce viral variants.

Repositioning of clinical approved drugs could represent the fastest way to identify therapeutic options during public health emergencies, the majority of drugs explored for repurposing as antivirals for 2019 coronavirus disease (COVID-19) have failed to demonstrate clinical benefit. Without specific antivirals, the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic continues to cause major global mortality. Antimalarial drugs, such as chloroquine (CQ)/hydroxychloroquine (HCQ) and mefloquine have emerged as potential anti-SARS-CoV-2 antivirals. CQ/HCQ entered the Solidarity and RECOVERY clinical trials against COVID-19 and showed lack of efficacy. Importantly, mefloquine is not a 4-aminoquinoline like CQ and HCQ and has been previously repurposed for other respiratory diseases. Unlike the 4-aminoquinolines that accumulate in the high pH of intracellular lysosomes of the lung, the high respiratory tract penetration of mefloquine is driven by its high lipophilicity. While CQ and HCQ exhibit activity in Vero E6 cells, their activity is obviated in TMPRSS2-expressing cells, such as Calu-3 cells, which more accurately recapitulate in vivo entry mechanisms for SARS-CoV-2. Accordingly, here we report the anti-SARS-CoV-2 activity of mefloquine in Calu-3 type II pneumocytes and primary human monocytes. Mefloquine inhibited SARS-CoV-2 replication in Calu-3 cells with low cytotoxicity and EC50 and EC90 values of 1.2 and 5.3 {micro}M, respectively. In addition, mefloquine reduced up to 68% the SARS-CoV-2 RNA levels in infected monocytes, reducing viral-induced inflammation. Mefloquine blocked early steps of the SARS-CoV-2 replicative cycle and was less prone than CQ to induce drug-associated viral mutations and synergized with RNA polymerase inhibitor. The pharmacological parameters of mefloquine are consistent with its plasma exposure in humans and its tissue-to-plasma predicted coefficient points that this drug may accumulate in the lungs. These data indicate that mefloquine could represent an orally available clinically approved drug option against COVID-19 and should not be neglected on the basis of the failure of CQ and HCQ.

microbiology↗

bioRxiv2021

The Great Deceiver: miR-2392's Hidden Role in Driving SARS-CoV-2 Infection

MicroRNAs (miRNAs) are small non-coding RNAs involved in post-transcriptional gene regulation that have a major impact on many diseases and provides an exciting avenue towards antiviral therapeutics. From patient transcriptomic data, we have discovered a circulating miRNA, miR-2392, that is directly involved with SARS-CoV-2 machinery during host infection. Specifically, we show that miR-2392 is key in driving downstream suppression of mitochondrial gene expression, increasing inflammation, glycolysis, and hypoxia as well as promoting many symptoms associated with COVID-19 infection. We demonstrate miR-2392 is present in the blood and urine of COVID-19 positive patients, but not detected in COVID-19 negative patients. These findings indicate the potential for developing a novel, minimally invasive, COVID-19 detection method. Lastly, using in vitro human and in vivo hamster models, we have developed a novel miRNA-based antiviral therapeutic that targets miR-2392, significantly reduces SARS-CoV-2 viability in hamsters and may potentially inhibit a COVID-19 disease state in humans.

systems biology↗

bioRxiv2021

Interpretable Drug Target Predictions using Self-Expressiveness

The identification of missing drug targets is critical for the development of treatments and for the molecular elucidation of drug side effects. Drug targets have been predicted by exploiting molecular, biological or pharmacological features of drugs and protein targets. Yet, developing integrative and interpretable machine learning models for predicting drug targets remains a challenging task. We present Inception, an integrative and interpretable matrix completion model for predicting drug targets. Inception is a self-expressive model that learns two similarity matrices: one for drugs and another for protein targets. These learned similarity matrices are key for our models interpretability: they can explain how a predicted drug-target interaction can be explain in terms of a linear combination of chemical, biological and pharmacological similarities. We develop a novel objective function with efficient closed-form solution. To demonstrate the ability of Inception at recovering missing drug-target interactions (DTIs), we perform cross-validation experiments with stringent controls of data imbalance, chemical similarities between drugs and sequence similarities between targets. We also assess the performance of our model using a simulated prospective approach. Having trained our model with DTIs from a snapshot 2011 of the DrugBank database, we test whether we could predict DTIs from a 2020 snapshot of DrugBank. Inception outperforms two state-of-the-art drug target prediction models in all the scenarios. This suggests that Inception could be useful for predicting missing drug target interactions while providing interpretable predictions.

bioinformatics↗

bioRxiv2019

The Geometric Sparse Matrix Completion Model for Predicting Drug Side effects

Pair-input associations for drug-side effects are obtained through expensive placebo-controlled experiments in human clinical trials. An important challenge in computational pharmacology is to predict missing associations given a few entries in the drug-side effect matrix, as these predictions can be used to direct further clinical trials. Here we introduce the Geometric Sparse Matrix Completion (GSMC) model for predicting drug side effects. Our high-rank matrix completion model learns non-negative sparse matrices of coefficients for drugs and side effects by imposing smoothness priors that exploit a set of pharmacological side information graphs, including information about drug chemical structures, drug interactions, molecular targets, and disease indications. Our learning algorithm is based on the diagonally rescaled gradient descend principle of non-negative matrix factorization. We prove that it converges to a globally optimal solution with a first-order rate of convergence. Experiments on large-scale side effect data from human clinical trials show that our method achieves better prediction performance than six state-of-the-art methods for side effect prediction while offering biological interpretability and favouring explainable predictions.

bioinformatics↗

bioRxiv2019

Predicting the Frequency of Drug Side effects

Drug side effects are a leading cause of morbidity and mortality. Currently, the frequency of drug side effects is determined experimentally during human clinical trials through placebo-controlled studies. Here we present a novel framework to computationally predict the frequency of drug side effects. Our algorithm is based on learning a latent variable model for drugs and side effects by matrix decomposition. Extensive evaluations on held out test sets show that the frequency class is predicted with 67.8% to 94% accuracy in the neighborhood of the correct class. Evaluations on prospective data confirm the commonly held hypothesis that most post-marketing side effects are very rare in the population, with occurrences of less than 1 in a 10,000. Importantly, our model provides explanations of the biology underlying drug side effect relationships. We show that the drug latent representations in our model are related to distinct anatomical drug activities and that the similarity between these representations are predictive of the drug clinical activity as well as drug targets.\n\nOne summary sentencenovel explainable machine learning algorithm predicts the frequency of drug side effects in the population

bioinformatics↗

bioRxiv2017

Tracking Subclonal Mutation Frequencies Throughout Lymphomagenesis Identifies Cancer Drivers in Mouse Models of Lymphoma.

Determining whether recurrent but rare cancer mutations are bona fide driver mutations remains a bottleneck in cancer research. Here we present the most comprehensive analysis of retrovirus driven lymphomagenesis produced to date, sequencing 700,000 mutations from >500 malignancies collected at time points throughout tumor development. This enabled identification of positively selected events, and the first demonstration of negative selection of mutations that may be deleterious to tumor development indicating novel avenues for therapy. Customized sequencing and bioinformatics methodologies were developed to quantify subclonal mutations in both premalignant and malignant tissue, greatly expanding the statistical power for identifying driver mutations and yielding a high-resolution, genome wide map of the selective forces surrounding cancer gene loci. Screening two BCL2 transgenic models confirms known drivers of human B-cell non-Hodgkin lymphoma, and implicates novel candidates including modifiers of immunosurveillance such as co-stimulatory molecules and MHC loci. Correlating mutations with genotypic and phenotypic features also gives robust identification of known cancer genes independently of local variance in mutation density. An online resource http://mulv.lms.mrc.ac.uk allows customized queries of the entire dataset.

cancer biology↗