Bioinformatics for Drug Discovery

Text understanding with #LLMs is useful but not enough for scientific understanding and discovery. In chemistry, in addition to text, chemical structure is essential to determine the properties of molecules. We have created the first multimodal text-chemical structure model: MoleculeSTM. It has an aligned latent space of both modalities. This allows the users to provide free-form text instructions to create molecules with arbitrary sets of properties. This enables zero-shot text-guided molecule editing (lead optimization) without the need to fine-tune the model for each new specification. Paper: bit.ly/4736BPH Code: bit.ly/4877YOS The core idea of MoleculeSTM is to align the chemical structure and textual description modalities using contrastive pretraining. The pivotal advantage of such alignment is its capacity to introduce a new paradigm of LLM for drug discovery: by fully utilizing the open vocabulary and compositionality attributes of natural language. To adapt it to a more concrete task, we focus on zero-shot text-guided molecule editing (aka lead optimization). Existing ML-based molecule editing methods suffer from data insufficiency issues. MoleculeSTM circumvents this by formulating molecule editing as a natural language understanding and interpolation problem, which is much easier to solve under the zero-shot setting. Such a novel paradigm is meaningful for addressing more practical drug discovery challenges. We will have more follow-up works along this LLM for the molecule/drug discovery research line. Please stay tuned! Shengchao Liu Chaowei Xiao Weili Nie Zhuoran Qiao Caltech

Google DeepMind just open-sourced TxGemma: the first efficient, generalist LLM suite for therapeutic development! Drug discovery has long been hindered by high failure rates, expensive experiments + the need for specialized AI models for each step of the pipeline. 𝗧𝘅𝗚𝗲𝗺𝗺𝗮 𝗶𝘀 𝗮 𝗳𝗮𝗺𝗶𝗹𝘆 𝗼𝗳 𝗹𝗮𝗿𝗴𝗲 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗺𝗼𝗱𝗲𝗹𝘀 𝘁𝗿𝗮𝗶𝗻𝗲𝗱 𝘁𝗼 𝗽𝗿𝗲𝗱𝗶𝗰𝘁 𝘁𝗵𝗲𝗿𝗮𝗽𝗲𝘂𝘁𝗶𝗰 𝗽𝗿𝗼𝗽𝗲𝗿𝘁𝗶𝗲𝘀, 𝗲𝗻𝗮𝗯𝗹𝗲 𝗶𝗻𝘁𝗲𝗿𝗮𝗰𝘁𝗶𝘃𝗲 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴, 𝗮𝗻𝗱 𝗼𝗿𝗰𝗵𝗲𝘀𝘁𝗿𝗮𝘁𝗲 𝗰𝗼𝗺𝗽𝗹𝗲𝘅 𝘀𝗰𝗶𝗲𝗻𝘁𝗶𝗳𝗶𝗰 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀. 1. Achieved superior or comparable results to state-of-the-art models on 64 out of 66 therapeutic tasks, surpassing specialist models on 26. 2. Reduced the need for large training datasets in fine-tuning, making it suitable for data-limited applications like clinical trial outcome prediction.   𝟯. 𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗲𝗱 𝗮𝗻 𝗲𝘅𝗽𝗹𝗮𝗶𝗻𝗮𝗯𝗹𝗲 𝗔𝗜 𝘁𝗵𝗮𝘁 𝗮𝗹𝗹𝗼𝘄𝘀 𝘀𝗰𝗶𝗲𝗻𝘁𝗶𝘀𝘁𝘀 𝘁𝗼 𝗶𝗻𝘁𝗲𝗿𝗮𝗰𝘁 𝗶𝗻 𝗻𝗮𝘁𝘂𝗿𝗮𝗹 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲, 𝗿𝗲𝗰𝗲𝗶𝘃𝗲 𝗺𝗲𝗰𝗵𝗮𝗻𝗶𝘀𝘁𝗶𝗰 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 𝗳𝗼𝗿 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗼𝗻𝘀, 𝗮𝗻𝗱 𝗲𝗻𝗴𝗮𝗴𝗲 𝗶𝗻 𝘀𝗰𝗶𝗲𝗻𝘁𝗶𝗳𝗶𝗰 𝗱𝗶𝘀𝗰𝘂𝘀𝘀𝗶𝗼𝗻𝘀.   4. Developed a therapeutic AI agent powered by Gemini 2.0, which surpassed leading models in complex chemistry and biology reasoning benchmarks (+9.8% on Humanity’s Last Exam, +5.6% on ChemBench-Preference) Since Evo2 by NVIDIA, I've been on the lookout for papers using mechanistic interpretability for explainability. It has obvious benefits for medicine. The use of a conversational variant that explains its reasoning here is a great for informing the user both the strengths and limitations of the model. I'd recommend looking at the example of this when the model is given a molecule’s SMILES string and asked if it can cross the blood‑brain barrier. I'm a firm believer that more researchers in the field should be incorporating explainability into their models. Will be highlighting research more that does so here. It is essential for our ability to iterate on the right things faster to improve the model and actually trust the models. Here's the awesome work: https://lnkd.in/gP--FXVU Congrats to Eric Wang, Samuel Schmidgall, Fan Zhang, Paul F. Jaeger, Rory P. and Tiffany Chen! I post my takes on the latest developments in health AI – 𝗰𝗼𝗻𝗻𝗲𝗰𝘁 𝘄𝗶𝘁𝗵 𝗺𝗲 𝘁𝗼 𝘀𝘁𝗮𝘆 𝘂𝗽𝗱𝗮𝘁𝗲𝗱! Also, check out my health AI blog here: https://lnkd.in/g3nrQFxW

We’re thrilled to announce our latest publication, “Next-Gen Therapeutics: Pioneering Drug Discovery with iPSCs, Genomics, AI, and Clinical Trials in a Dish”, now published in Annual Reviews of Pharmacology and Toxicology! 🔗 Read the full article here: https://lnkd.in/gGSNhUeP Congratulations to first author Zehra Yildirim, Kyle Swanson Xuekun WU James Zou, and the entire team at Stanford Cardiovascular Institute, Stanford University Department of Computer Science as well as biotech startup Greenstone Biosciences! Why does this research matter? The traditional drug discovery pipeline faces a 92% failure rate, largely due to unforeseen toxicities and lack of efficacy in clinical trials. With the FDA Modernization Act 2.0, the field is shifting toward human-relevant models, leveraging iPSC-derived organoids, organ-on-a-chip systems, and AI-driven analytics to accelerate drug development. Broad Implications: This work explores how next-generation drug discovery approaches can improve the efficiency and accuracy of preclinical testing. Clinical trials in a dish provide a promising way to predict drug responses using human-relevant models. AI-powered drug discovery helps optimize candidate selection, potentially accelerating the identification of safer and more effective therapies. Additionally, next-gen preclinical screening may reduce reliance on animal models, contributing to more predictive and patient-specific treatments while advancing the field toward precision medicine. #DrugDiscovery #AI #Organoids #iPSC #ClinicalTrialsInADish #PrecisionMedicine

A new report “Beyond Legacy Tools: Defining Modern AI Drug Discovery for 2025 and Beyond,” is out! (link in the comments) The report by BiopharmaTrend (Disclaimer: I am a co-founder of the company) analyzes the AI platforms behind companies like Recursion, Insilico Medicine, Iambic Therapeutics, Schrödinger, Verge Genomics, NOETIK and several others — and shows that despite their different architectures and areas of focus, they share a set of category defining traits: ✔️ Modeling biology holistically, not just focusing on single targets or pathways ✔️ Building scalable, software-first platforms that integrate wet-lab and in silico workflows ✔️ Owning or generating massive, multimodal datasets (e.g. omics, imaging, patient data, and proprietary perturbation experiments) ✔️ Embedding AI at every stage of the pipeline, connecting the dots via gen AI. 👉 The report also introduces the concept of Holistic Drug Development (HDD), a vision where AI platforms integrate real-world patient data, systems biology, and generative chemistry into a continuous, learning-driven loop. Here is my take: “We’ve been using machine learning in biology for decades” is a common argument meant to downplay the idea that AI drug discovery (AIDD) is a new category. But IMO, this argument falls short. Yes, machine learning (ML) has been used in biology and chemistry for decades: QSAR models, clustering, PCA, support vector machines, and basic neural nets, etc. But those were point solutions: tools applied to narrow tasks (e.g., predicting solubility, docking ligands, or clustering gene expression data). What’s different now is that ML, particularly deep learning, generative modeling, and transformer architectures, is being used to rebuild the entire discovery workflows. Next, earlier ML approaches required handcrafted features (e.g., molecular descriptors). Today’s models can learn rich, abstract representations directly from raw data — from sequences, graphs, images, and text — and use them across tasks. That shift is foundational and category-defining. Also, traditional ML tools were modular and disconnected. Today’s AIDD platforms integrate multimodal data (omics, imaging, EHRs, chemical structures, etc). Modern AI drug discovery platforms, operate in closed feedback loops with wet-lab systems, offer full-stack software products with APIs, dashboards, and orchestration layers. That level of scope and systems integration is categorically different, IMO. Classical ML mostly focused on prediction and classification. AIDD platforms now generate novel chemistry, hypotheses, even trial designs, shifting from prediction tools to creative engines within the discovery process. Finally, the important aspect is production-grade software platforms, not just scripts and models. Using ML as a helper tool ≠ building AI-native, data-driven engines. I am pretty certain. Disagree? Image credit: BiopharmaTrend

🟨🟪 𝙏𝙖𝙧𝙜𝙚𝙩𝙞𝙣𝙜 𝙈𝙃𝘾-𝙄 𝙢𝙤𝙡𝙚𝙘𝙪𝙡𝙚𝙨 𝙛𝙤𝙧 𝙘𝙖𝙣𝙘𝙚𝙧: 𝙛𝙪𝙣𝙘𝙩𝙞𝙤𝙣, 𝙢𝙚𝙘𝙝𝙖𝙣𝙞𝙨𝙢, 𝙖𝙣𝙙 𝙩𝙝𝙚𝙧𝙖𝙥𝙚𝙪𝙩𝙞𝙘 𝙥𝙧𝙤𝙨𝙥𝙚𝙘𝙩𝙨 👉#MD_Immunol https://lnkd.in/dUGBPevC 💠 MHC-I molecules play a critical role in presenting antigens to T cells, particularly cytotoxic T cells, which recognize and eliminate infected or cancerous cells. 💠 MHC-I molecule-targeted therapies aim to restore immune control in the tumor microenvironment, which can lead to improved T cell-mediated immunity and better cancer treatment outcomes. 💠 Defects in the MHC-I antigen presentation pathway are a common mechanism by which tumors can evade the immune response. 💠 Reduced levels and complete loss of TAP1 and MHC-I expression have been observed in various cancers, leading to the evasion of immune surveillance and the promotion of tumor growth. 👉 mutations that cause loss of TAP1 or TAPASIN result in decreased MHC-I expression. 💠 promising MHC-I molecule-targeted therapies for cancer treatment include: ✅Cytokine therapy: Interferon  molecules are promising cytokines targeted at MHC-I because they prompt dendritic cells or T cells to eliminate cancer cells through upregulating MHC-I molecules. ✅ Combination therapy with immune checkpoint inhibitors (ICB): Combination therapy with ICB has been done in clinical trials, showing promising results in enhancing the immune response against cancer cells. ✅ Inhibition of negative regulators of MHC-I-inducing pathways: Dysregulation of factors affecting the downstream signaling of MHC-I-inducing pathways makes it difficult for MHC-I molecules to present antigens to T cells, leading to immune evasion of tumors. Inhibiting negative regulators of these pathways, possibly combined with stimulation of positive regulators of these pathways, may be a promising strategy to explore. ✅ Targeting oncogenic pathways regulating MHC-I in cancer immunotherapy: such as EGFR and STAT3, can increase MHC-I expression and antigen presentation components, promoting anti-tumor immune responses. ✅ NK-cell therapy strategy: NK-cells are capable of either killing or upregulating MHC-I, making them a potential target for cancer immunotherapy. 💠 These therapies differ from traditional cancer treatments in their immune-based approach, personalized medicine approach, combination therapy potential, and immunologic effects. 💠 Associated side effects include: ✅ Immune-related adverse events ✅ Tumor progression ✅ Therapeutic resistance ✅ Off-target effects 💠 Some examples of cancers that may be targeted by MHC-I molecule-targeted therapies include: ✅ Pancreatic ductal adenocarcinoma ✅ Neuroblastoma ✅ Solid tumors 💠 It is essential to note that the clinical development of MHC-I molecule-targeted therapies is still in its early stages, and more research is needed to determine the specific types of cancer that these therapies can effectively treatment. #immunology #cancer

UC San Diego may have cracked one of oncology’s hardest problems - treatment resistance. Scientists engineered a new antibody that targets integrin αvβ3, a protein found in aggressive cancers but absent in healthy tissue. → Activates macrophages (not NK cells) to kill tumor cells → Boosts iNOS and nitric oxide to trigger cancer cell death → Worked in both mouse models and patient-derived tumors That’s a new playbook: reprogram the tumor’s immune environment instead of fighting it. If clinical trials confirm this, we could be looking at a precision immunotherapy that turns tumors’ own defenses against them - and a real shot at taming resistance itself.

AI tools are significantly advancing drug discovery by improving the prediction of protein-ligand interactions at several key stages: ⏩Enhanced Prediction of Protein-Ligand Interactions: AI-driven methods are proving highly effective in predicting how proteins and ligands interact, encompassing tasks like pose prediction (how a ligand binds to a protein), scoring (assessing binding strength), and virtual screening (identifying potential drug candidates from large libraries). ⏩Improved Ligand Pose and Scoring with Advanced Models: Cutting-edge AI models, specifically diffusion and geometric deep learning models, are leading to more accurate predictions of ligand binding poses and more effective scoring functions. This means a better understanding of how a drug molecule will physically fit into and interact with its cellular target, both externally and internally. ⏩Refined Ligand Binding Site Identification: Hybrid AI approaches, which combine information from both protein sequence and structural embeddings, are improving the identification of specific regions on proteins where ligands are likely to bind. This is crucial for pinpointing druggable targets and peptide ligand-mediated targeted drug delivery of RNA therapeutics, especially for peptide-oligonucleotide conjugates (POC). ⏩Superior Virtual Screening Accuracy: AI-based scoring functions are outperforming traditional docking methods in virtual screening by providing more accurate assessments of potential drug candidates, leading to a higher success rate in identifying promising compounds. ⏩Addressing Generalizability Challenges: To further enhance the utility and applicability of AI in drug discovery, it is essential to incorporate protein flexibility (the natural movement of proteins) and utilize more diverse datasets. This helps to overcome limitations in how well AI models generalize to new and varied protein-ligand systems. #AIdrugdiscoverydevelopment

Summary: This research paper introduces a computational approach for predicting the biological activity of drug compounds aimed at treating type 2 diabetes. By leveraging machine learning (ML) techniques, the authors developed a predictive model that estimates a drug compound's activity based on its molecular structure. Three ML methods were evaluated: support vector regression (SVR), neural networks (NN), and extreme gradient boosting (XGBoost). Among these, XGBoost emerged as the most accurate method for predicting biological activity. Methodology: The study utilized a dataset of drug compounds tested for activity against type 2 diabetes. Molecular descriptors—numerical features representing the structure of each molecule—were extracted and served as input data for the ML models. The authors trained and tested the models using these descriptors to predict biological activity as the output. Results and Discussion: XGBoost outperformed the other ML methods, achieving an R² value of 0.94 on the training set and 0.91 on the test set, indicating high predictive accuracy. Additionally, the authors identified the "element_count" feature as the most significant factor influencing biological activity predictions. Conclusion: This research highlights a promising computational method for predicting the biological activity of drug compounds, with XGBoost proving to be the most effective ML technique for this purpose. The approach has the potential to accelerate drug discovery for type 2 diabetes and could be adapted for other diseases. Key Terms and Concepts: Type 2 diabetes Drug discovery Machine learning (ML) Support vector regression (SVR) Neural networks (NN) Extreme gradient boosting (XGBoost) Molecular descriptors R² value Additional Notes: The dataset was sourced from PubChem, a publicly available chemical information repository. The authors built their ML models using the Python library scikit-learn. The presented method could be applied to predict the biological activity of drug compounds for other conditions beyond type 2 diabetes.

Most drug discovery efforts still chase single targets. But complex diseases don’t work that way, they’re driven by networks, redundancy, and eerie adaptability. What if we could use AI, not to find “a hit,” but to shift entire cellular states from disease back toward health? A new Science paper (DeMeo et al., Oct 2025) takes that paradigm seriously. The team (Cellarity/MIT/Helmholtz) built DrugReflector, a deep-learning model trained on more than 1.2 million human cells and 88 chemical perturbations, using single-cell transcriptomics as its foundation. Instead of asking, “Will this molecule bind my target?”, they ask, “Will this molecule rewire the system toward a healthy phenotype?” What’s new here? DrugReflector predicts not just which compounds bind, but which will actually shift the transcriptomic signature of, say, a blood stem cell, into paths leading to functional megakaryocytes or erythrocytes, the cells we need for treating anemia or platelet disorders. They validated against brute-force screening (the industry standard): Random selection: ~1% hit rate Their model: up to 17%: a 13–17x improvement (and robust across several donors and pathways). The method uses closed-loop reinforcement learning. Initial predictions guide the first round of screening, then transcriptomic and phenotypic readouts from the real experiments refine the model in “lab-in-the-loop” cycles. With each iteration, the hit rate and biological insight both get sharper. It recovers known standards of care and highlights new targets. Notably, it identified both established kinase inhibitors and a new class of molecules modulating cholesterol synthesis to drive megakaryocyte commitment, finding druggable nodes unseen by classic screens. Why does this matter? Phenotypic drug discovery, with deep, biology-aware AI, can leapfrog the “screen everything” mentality, bringing tractability and true systems-level correction to disease treatment. Every cycle isn’t just screening, it’s learning: about cell fate, about target redundancy, and about network rewiring as therapy. The future lies in AI that understands and actively learns from biology, the cell as its own target, not just a test tube for single-protein hits. Are we ready to reimagine drug discovery workflows around this? The tools, and now, the evidence, are here!

Promising target for CAR T-cell therapy leads to potent antitumor responses against cutaneous and rare melanomas UCLA researchers identified the protein TYRP1 as a potential target for CAR T-cell immunotherapy The new approach, described in the journal Nature Communications, uses an engineered CAR T-cell that is designed to recognize and attack cells with high levels of TYRP1, a protein found on the surface of melanoma cells. The team found these engineered CAR T-cells can effectively eliminate cancer cells in preclinical tests without causing severe side effects. “One of the biggest challenges in CAR T-cell therapies is the scarcity of suitable tumor targets,” said Cristina Puig-Saus, PhD, assistant professor of medicine at the David Geffen School of Medicine at UCLA and senior author of the study. “While TYRP1 has previously been targeted in clinical trials using monoclonal antibodies, this new approach harnesses the power of CAR T-cell therapy and has led to very good anti-tumor responses, improving the treatment's overall effectiveness.” https://lnkd.in/eRU8AvPQ https://lnkd.in/e6KGZc_8