Biosignal Processing Group’s Post

🚀 I recently read the TransMA paper used to predict mRNA delivery efficiency and it really shows the progress of AI in chemistry! 🧬 The initial COVID vaccine relied on lipid nanoparticles to deliver mRNA into our cells but finding the right lipid formulation is like finding a needle in a haystack. Labs spend months and millions testing thousands of combinations. 🤖 This is where TransMA comes: It uses an innovative approach combining a molecular 3D transformer, the Mamba architecture, and applies self-attention to SMILES predicting LNP performance computationally before expensive lab testing. ✔️ The transformer analyzes spatial relationships between atoms in 3D space. ✔️ Molecule Mamba processes long SMILES sequences. ✔️ Mol-Attention looks to identify which atoms matter most for transfection efficiency. 🎯 The paper reports an achieved 35-43% better predictions than existing methods! ❌ But, while the approach is promising, I have concerns: 🔻 Without proper validation against known pharmacophores, can we trust which atoms the model thinks are "important" and rely on the self attention? 🔻 15 data points for generalization testing? That's not enough to convince me this works in the real world. Despite these issues, I am excited to see Mamba making its way into molecular discovery! Relevant links are in the comments. Chemetrian #MachineLearning #DrugDiscovery #AI #Chemistry #Biotech #Innovation

How can AI accelerate insights in clinical decision-making? At #PGRNClinPGx25, Duponte will host a hands-on microsession on the use of AI in pharmacogenomic analysis. Participants are invited to bring their laptops and VCF files for a practical session designed to explore how AI can contribute to: 🔵The interpretation of genetic variation, 🔵Clinical decision-making, and 🔵Biomedical research. Far from a tool demo, this session is structured as a scientific discussion—examining both the opportunities and the current limitations of AI in PGx. Meet us on Thursday, September 11 at the University Center, Rooms 332–333 (3rd floor) | 1:20–1:50 PM | ClinPGx Clinical Pharmacogenetics Implementation Consortium PharmGKB University of Montana

Another example of an attractive Trio: AI based model + CryoEM + MD simulation. “… how combining AI-driven structure prediction with cryo-EM imaging data and molecular simulations can enable accurate modeling of protein-drug complexes with minimal system knowledge or structural biology expertise.”

✨ Thrilled to Share My Publication! ✨ I am incredibly excited to announce that our review article “AI and Machine Learning in Biotech: Drug Discovery, Epigenetics & Disease Prognosis” has been officially published in the Journal of Science Research International (JSRI) 🎉 This work, carried out under the invaluable guidance of Meenakshi Johri Ma'am and Mrs. Bindu Rajaguru (Assistant professor, Pillai college of Arts, commerce and science, New Panvel), explores how Artificial Intelligence (AI) and Machine Learning (ML) are revolutionizing the field of biotechnology. 🔬 The paper dives into three critical areas: Drug Discovery – how AI accelerates the identification of novel drug targets, molecular design, and clinical trial optimization. Epigenetics – decoding complex gene regulation patterns and identifying biomarkers with precision. Disease Prognosis – using predictive models for early detection, personalized treatment, and better healthcare outcomes. With AI reshaping research and healthcare globally, this article highlights the opportunities, challenges, and the exciting future of personalized medicine. 📖 You can read the full article here: https://lnkd.in/gYYfSCCd I am grateful to my mentors and co-authors for their constant support and guidance throughout this journey. This milestone motivates me to keep learning, exploring, and contributing to the growing intersection of AI and life sciences. #ArtificialIntelligence #MachineLearning #DrugDiscovery #Epigenetics #Biotechnology #HealthcareInnovation #ResearchPublication

We're proud to share new research from Constructor Knowledge Labs by our Principal Investigator Petr Popov's group, and now published in ”Quarterly Reviews of Biophysics”. Titled “Computational Methods for Binding Site Prediction on Macromolecules”, a new study presents a comprehensive review of state-of-the-art computational approaches for predicting binding sites on macromolecules—an essential step for drug discovery, functional annotation, and understanding molecular mechanisms. The paper emphasizes how deep learning architectures have started to outperform traditional approaches by capturing structural and evolutionary patterns. It also outlines emerging challenges, including handling flexibility, multi-specific binding, and the integration of experimental data. Looking forward, the authors highlight opportunities in combining physics-based modeling with AI, improving benchmarking standards, and expanding predictions beyond proteins targets. By synthesizing current knowledge and charting future directions, this work provides both newcomers and experts with a roadmap for advancing computational biology. The review underscores the transformative impact of integrating structural biology, computational modeling, and artificial intelligence in the search for novel therapeutics. Read the full article here: https://lnkd.in/dhD28N2s

MIT Course announcement: Machine Learning for Computational Biology #MLCB25 Fall'24 Lecture Videos: https://lnkd.in/efSvp7hY Fall'24 Lecture Notes: https://lnkd.in/eWBAxQHk (a) Genomes: Statistical genomics, gene regulation, genome language models, chromatin structure, 3D genome topology, epigenomics, regulatory networks. (b) Proteins: Protein language models, structure and folding, protein design, cryo-EM, AlphaFold2, transformers, multimodal joint representation learning. (c) Therapeutics: Chemical landscapes, small-molecule representation, docking, structure-function embeddings, agentic drug discovery, disease circuitry, and target identification. (d) Patients: Electronic health records, medical genomics, genetic variation, comparative genomics, evolutionary evidence, patient latent representation, AI-driven systems biology. Foundations and frontiers of computational biology, combining theory with practice. Generative AI, foundation models, machine learning, algorithm design, influential problems and techniques, analysis of large-scale biological datasets, applications to human disease and drug discovery. First Lecture: Thu Sept 4 at 1pm in 32-144 With: Prof. Manolis Kellis, Prof. Eric Alm, TAs: Ananth Shyamal, Shitong Luo Course website: https://lnkd.in/eemavz6J #MLCB25 #MIT #MITBiology #MITCSAIL #MachineLearning #DeepLearning #ArtificialIntelligence #GenerativeAI #FoundationModels #Transformers #ComputationalBiology #Genomics #Proteomics #Epigenomics #SystemsBiology #GenomeEngineering #PrecisionMedicine #PersonalizedMedicine #MedicalGenomics #GeneticVariation #EvolutionaryBiology #Bioinformatics #MedicalAI #AIinHealthcare #BiomedicalAI #HealthTech #DrugDiscovery #Therapeutics #Biotech #PharmaTech #Cheminformatics #ProteinDesign #ProteinFolding #AlphaFold2 #CryoEM #StructuralBiology #MolecularBiology #SyntheticBiology #NeuroscienceAI #ClinicalAI #BigDataBiology #DataScience #AlgorithmDesign #FoundationAI #Omics #Transcriptomics #Epigenetics #EHR #DiseaseModeling #TargetIdentification #ComputationalGenomics #BiomedicalResearch #HealthcareInnovation

One of the things I'm most hopeful for is AI will finally allow us to develop the clinical infrastructure to escape tyranny of simplicity in medicine. The prevalence of the single-biomarker approach in the clinic is not just a reflection of outdated science, but also a systemic failure in the translation of research into practice. The world of biomedical research is currently experiencing an explosion of data from "omics" technologies—genomics, proteomics, metabolomics, amazing new measurement technology that gives us data we never had before. We have new understanding on networks of interacting molecules can developing complex, multi-marker signatures to describe disease states.   However, this complexity often dies in the "valley of death" of translational medicine, that gap between laboratory discovery and clinical implementation. There is immense pressure from regulatory bodies, healthcare systems, and the constraints of a busy clinical practice to simplify these complex discoveries into something actionable. A single test with a single cut-off is easy to order, easy to interpret, easy to bill for, and easy to build a clinical guideline around. Validating a multi-marker panel is a far more complex, expensive, and time-consuming regulatory and scientific endeavor. As a result, the rich, high-dimensional understanding of disease being generated in research is often filtered and flattened down into the one-dimensional, low-information tools that are actually used at the patient's bedside. The clinical reality lags decades behind the scientific frontier, trapped by the practical tyranny of simplicity

Range Biotechnologies exists to solve this problem by productizing cutting edge research so it can actually make it to patients in a clinical setting. Not an easy task, but someone has to do it or medicine will remain in the past forever.

A new artificial intelligence (AI) tool developed at The University of Melbourne is helping transform how skin cancer is detected. The technology has the potential to save lives, reduce unnecessary biopsies and cut healthcare costs, while addressing long-standing equity gaps in diagnosis. Research Fellow Dr Noor E Karishma Shaik from the Faculty of Engineering and Information Technology is integrating AI with thermal multimodal imaging for point of care diagnosis to identify abnormal skin lesions in real time. Read more: https://lnkd.in/drcazMrq

BioASQ 2025 advances biomedical AI with six innovative tasks. 💡 These advancements can streamline drug discovery and personalized medicine, potentially saving billions in R&D costs. 🇬🇷 🇮🇹 Arxiv paper: 📄 https://lnkd.in/ekwxYUyN Author: Anastasios Nentidis, Georgios Katsimpras, Anastasia Krithara, Martin Krallinger, Miguel Rodríguez-Ortega, Eduard Rodriguez-López, Natalia Loukachevitch, Andrey Sakhovskiy, Elena Tutubalina, Dimitris Dimitriadis, Grigorios Tsoumakas, George Giannakoulas, Alexandra Bekiaridou, Athanasios Samaras, Giorgio Maria Di Nunzio, Nicola Ferro, Stefano Marchesin, Marco Martinelli, Gianmaria Silvello, Georgios Paliouras University of Padua #ComputerScience #ComputationandLanguage #arXiv