Scientific Computing Software Tools

Contemporary applications and advances of LLMs in bioinformatics. 1. DNA/RNA sequence analysis, functional and structure prediction: includes tools for sequence analysis (e.g. HvenaDNA, DNAGPT), functional prediction (e.g. BERT-enhancer, DNABERT), and structure-focused methods (e.g. RNABERT, GeoBoost2). 2. Protein sequence analysis, functional and structure prediction: covers protein sequence modeling (e.g. ESM, ProtGPT), post-translational modification prediction (e.g. EpiBERTope, TransPPMP), and structural analysis (e.g. ProteinBERT, MSA transformer). 3. Multi-omics data analysis: features tools for genomics (e.g. scGPT, iDNA-ABT), epigenomics (e.g. scELMo, Mul_an-methyl), and integrative omics approaches (e.g. DeepGene transformer, POOE). 4. Computational drug discovery and design: includes models for molecular design (e.g. MolGPT, ChemBERTa), drug–target interaction (e.g. DT-I-BERT, TransDTI), and pharmaceutical applications (e.g. PharmBERT). 5. Biomedical literature mining: lists NLP models for biomedical text analysis (e.g. BioBERT, ClinicalBERT, Galactica). Paper Link: https://lnkd.in/gPHSyKRe

Before you can design a drug, you need to find where it binds. Binding site prediction is often overlooked, but it's critical. 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀: → Cryptic pockets only visible in certain conformations → Allosteric sites far from active sites → Protein-protein interaction interfaces (flat, featureless) → Transient pockets that open/close dynamically 𝗧𝗼𝗼𝗹𝘀 𝗜'𝘃𝗲 𝗳𝗼𝘂𝗻𝗱 𝘂𝘀𝗲𝗳𝘂𝗹: 𝗣𝟮𝗥𝗮𝗻𝗸: ML-based, fast, good accuracy 𝗙𝗣𝗼𝗰𝗸𝗲𝘁: Geometry-based, well-validated 𝗦𝗶𝘁𝗲𝗠𝗮𝗽: Comprehensive druggability assessment 𝗗𝗼𝗚𝗦𝗶𝘁𝗲𝗦𝗰𝗼𝗿𝗲𝗿: Good for comparing multiple pockets 𝗣𝗿𝗼 𝘁𝗶𝗽: Run pocket detection on MULTIPLE conformations from MD. That cryptic site might only appear in 10% of frames. But it could be your most druggable option. What's your go-to tool for binding site prediction? #DrugDiscovery #BindingSite #P2Rank #Druggability #CADD #StructuralBiology #BioCogniz #UAE #Pharmacy #Bioinformatics

After 18 years of infertility, a couple turned to a Columbia University team who deployed something called S.T.A.R. - Sperm Tracking And Recovery: it’s AI scanning 8 million images in under 60 minutes, locating sperm cells technicians missed in two full days. Let that sink in: the same AI approach used for discovering stars is now spotting sperm. We’re talking about replacing laborious human scrutiny with image‑processing horsepower and robotics. In one case, STAR found 44 sperm in under an hour where manual methods found none. The end result? A healthy pregnancy and the first documented case of AI‑confirmed conception via this technique . 🧠 Why this matters to us in enterprise and tech: Tech/biology convergence: The fusion of astrophysics‑grade image analysis with micro‑robotics showcases how cross‑disciplinary algorithms can revolutionize established industries: fertility science today, your sector tomorrow. Operational leverage: STAR achieved in 60 minutes what took human teams two days. Multiply that by hundreds of tasks and imagine the scale. Ethics meets engineering: Deploying AI for reproductive medicine invites us to re-examine trust, regulation, and accountability: not just for boardrooms but for society at large. Competitive foresight: If your product involves edge‑case detection (fraud, anomalies, rare events), take note; STAR’s architecture is a template. My take (tongue half in cheek): If you thought AI’s “obvious” use‑cases were limited to chatbots and predictive marketing, meet STAR. It proves the real frontier is where domain experts declare “we’ve tried everything”. That’s your cue- find that frontier in your industry and ask: what could AI find in “haystacks” we’ve written off? Final thought: Entrepreneurs, view this as a road map. Identify the domain where humans have “tried it already,” then deploy cross‑disciplinary AI to reveal what we’ve overlooked. What other “haystacks” have you written off too soon? ↓ ↓ ↓ 👉 Stay ahead—Follow me on LinkedIn and subscribe to the newsletter: www.michaelhousman.com #AIInnovation #CrossDisciplinaryTech #OperationalExcellence #EthicalAI #ThoughtLeadership

AI-powered virtual screening that scores 10 trillion protein-ligand pairs in a single day Of ~20,000 human protein-coding genes, only about 10% have been successfully targeted by FDA-approved drugs or have documented small-molecule binders. The bottleneck isn't biology—it's computational scale. Traditional molecular docking takes seconds to minutes per protein-ligand pair, making genome-wide screening essentially impossible with current resources. Yinjun Jia and coauthors tackle this head-on with DrugCLIP, a contrastive learning framework that reframes virtual screening as a dense retrieval problem—similar to how modern search engines work. The key innovation: encode protein pockets and small molecules into a shared latent space using separate neural networks, then use cosine similarity for ultrafast ranking. The model is pretrained on 5.5 million synthetic pocket-ligand pairs extracted from protein structures, then fine-tuned on 40,000 experimentally determined complexes. The speed gains are staggering—up to 10 million times faster than docking. Combined with GenPack, a generative module that refines pocket detection on AlphaFold2-predicted structures, DrugCLIP enables screening at a scale previously unthinkable: 500 million compounds against ~10,000 human proteins, scoring more than 10 trillion pairs in under 24 hours on just 8 GPUs. The wet-lab validations are equally compelling. For norepinephrine transporter (NET), a 15% hit rate with two inhibitors structurally confirmed by cryo-EM. For TRIP12—a challenging E3 ubiquitin ligase with no known inhibitors or holo structures—a 17.5% hit rate using only AlphaFold2 predictions, with functional enzymatic inhibition confirmed. The resulting database, GenomeScreenDB, covers ~20,000 pockets from 10,000 proteins—nearly half the human genome—and is freely available at drugclip.com. The message is clear: by combining contrastive representation learning with generative pocket refinement and AlphaFold structures, we've entered an era where genome-wide drug discovery becomes computationally tractable, opening systematic exploration of the vast undrugged proteome. Paper: https://lnkd.in/e7aGUvAX #DrugDiscovery #ArtificialIntelligence #MachineLearning #DeepLearning #VirtualScreening #ComputationalBiology #AlphaFold #ProteinScience #Biotech #AIforScience #StructuralBiology #Bioinformatics #Pharmaceuticals #ComputationalChemistry #PrecisionMedicine

💊 AI just made drug discovery searchable Virtual screening at genome scale has been computationally prohibitive. Traditional molecular docking works well for one target at a time, but screening large compound libraries against thousands of human proteins simultaneously would take years, even on modern GPU clusters. A team at Tsinghua University just changed this. They screened 500 million compounds against 10,000 human proteins, scoring 10 trillion protein-ligand pairs in under 24 hours using just 8 GPUs (!). Their new paper in Science (https://lnkd.in/ek5-d9F7) introduces DrugCLIP, a contrastive learning approach that's 10 million times faster than traditional docking. 🔬 How it works Two neural networks encode protein pockets and drug molecules into a shared embedding space, trained so that binders cluster near their targets while non-binders are pushed apart. Both encoders are built on UniMol, a 3D transformer that processes atomic coordinates directly rather than chemical formulas. The training is clever: pretrained on 5.5 million synthetic protein-fragment pairs, then fine-tuned on 44,000 real crystal structures using random conformations rather than exact poses - forcing the model to learn chemical features, not memorize geometry. Once trained, screening becomes nearest-neighbor search. 🚀 Why it's so fast The speed comes from pre-computation. You encode your 500 million molecules once and store the vectors offline. Screening a new protein target then becomes vector similarity - no physics simulations, no pose sampling, no energy minimization per molecule. 📊 The validation The team validated hits in wet-lab experiments. Traditional virtual screens typically yield 1-5% hit rates. DrugCLIP achieved: → 15% hit rate for norepinephrine transporter (NET), with structurally novel inhibitors distinct from existing drugs - two confirmed by cryo-EM → 17.5% hit rate for TRIP12, a target with no previously known ligands, using only AlphaFold-predicted structures That second result is remarkable - they found the first functional inhibitors for an unexplored target implicated in cancer and Parkinson's. 🌐 The resource The team released GenomeScreenDB (https://drugclip.com), an open-access database containing candidate molecules for ~20,000 pockets across ~10,000 human proteins - more targets than have any known ligands in ChEMBL. I think this represents a shift in how drug discovery will work. When screening becomes this fast and cheap, the bottleneck moves from computation to ideas: which targets matter, which patient populations to prioritize, how to validate hits efficiently. Congratulations to co-first authors Yinjun Jia, Bowen Gao, Jiaxin Tan, Jiqing Zheng, Xin Hong and senior authors Yanyan Lan, Wei Zhang, Chuangye Yan, and Lei Liu.

Imagine we could map every cell in the human body, revealing its precise location and molecular identity. This tantalizing possibility is at the heart of our latest perspective piece published in Nature Methods, where we explore a groundbreaking approach to understanding biological systems at unprecedented depth and scale: Deep 3D Histology. In this perspective article, we discuss three key pillars of this emerging field: Advanced Tissue Clearing and Imaging: -Cutting-edge tissue clearing techniques for intact specimen visualization -High-resolution light-sheet microscopy pushing the boundaries of 3D imaging -Applications ranging from mouse embryos to entire human organs Spatial Omics Technologies: -Integration of single-cell omics data with 3D spatial context -Creation of comprehensive molecular atlases of entire organisms -Bridging the gap between molecular profiles and tissue architecture Artificial Intelligence in Image Analysis: -Deep learning revolutionizing 3D histology data processing -Automation of tasks from image enhancement to cell segmentation -Unveiling information invisible to the human eye through "virtual staining" The potential impact of combining these technologies is staggering. By accelerating our understanding of diseases and drug discovery, we could compress centuries of insights into just a few years of research. Challenges remain, including improving resolution, increasing imaging speed for large samples, and developing user-friendly AI tools. But as we overcome these hurdles, Deep 3D Histology could become a routine tool in both research and clinical settings. The future of biomedical research is three-dimensional, molecularly detailed, and AI-enhanced. This new era of 3D omics has the potential to revolutionize medicine and our understanding of life itself. You can read the full perspective and join the discussion on this exciting frontier of science: https://rdcu.be/dNBe8 More technical details are here as tweetorial: https://lnkd.in/d48cTXDE #AI #DeepLearning #Clearing #3D #Imaging #Omics #Deep3DHistology

Scientists at Tsinghua University and Westlake University introduced #Dynaformer, a revolutionary graph-based Deep Learning model for predicting protein-ligand binding affinities. Unlike previous methods, Dynaformer leverages molecular dynamics simulations to capture the dynamic nature of protein-ligand interactions. 🎯 Dynaformer demonstrates state-of-the-art performance on the CASF-2016 benchmark, outperforming existing methods. The model learns from a curated dataset of 3,218 protein-ligand complexes, offering unprecedented accuracy in binding affinity prediction. 🔬 In a real-world test, Dynaformer identified 12 hit compounds (including 2 submicromolar hits) for HSP90 through virtual screening. This success, coupled with novel scaffold discoveries, showcases Dynaformer's potential to accelerate early-stage drug discovery. Quick Read: https://lnkd.in/gRWKt-V8 #Bioinformatics #MolecularDynamics #DeepLearning #StructuralBIology #AIinDrugDiscovery #ComputationalChemistry #DrugDesign #ScienceNews

In the past month, new tools have improved protein design, benchmarking, and refinement, showing both progress and areas for improvement. ▪️ Proteína by NVIDIA is a flow-based generative model for de novo protein backbone design, trained on 21 million AlphaFold structures. Its 400M-parameter transformer supports C.A.T.H. conditioning, classifier-free guidance, and autoguidance for precise structure control. Proteína generates proteins up to 800 residues, outperforming RFdiffusion and Genie2 in scale and accuracy. ▪️ MotifBench is a comprehensive benchmark for evaluating motif-scaffolding methods in protein design. Featuring 30 challenging test cases from the Protein Data Bank, it provides a standardized evaluation pipeline and identifies key limitations in methods like RFdiffusion, highlighting the need for more advanced approaches. ▪️ ROCKET enhances AlphaFold2’s protein structure predictions by integrating experimental data from X-ray, cryo-EM, and cryo-ET without requiring retraining. It refines large structural changes, improves accuracy at low resolution, and allows automated, experiment-guided model building. At RECEPTOR.AI, we're closely following these innovations while recognizing their dependence on high-quality data. Rather than relying solely on model adjustments, we emphasize integrating experimental insights and physics-informed approaches with AI. This combined strategy helps overcome current limitations and accelerates meaningful advances in protein design and drug discovery. GIF credit: Ian Haydon / Institute for Protein Design (via MIT News) #ai #ml #artificialintelligence #biotech #drugdiscovery

Tried something interesting with automated annotations on histology images. Using a simple text prompt, Grounding DINO was able to detect glomeruli in a kidney H&E image and generate bounding boxes. I then passed those boxes to SAM 2, which converted them into clean, pixel-level segmentation masks. So essentially: text prompt → object detection → precise segmentation. What stood out to me is how well this worked on scientific imagery, not just natural images. Annotation is often a bottleneck in biomedical workflows, and this kind of pipeline could significantly speed things up. Would be interesting to see how robust this is across different stains, tissues, and imaging conditions. If there’s interest, I can put together a short video walkthrough. #microscopy #computervision #deeplearning #imageanalysis #digitalpathology

➡️ X-ray diffraction (XRD) and electron microscopy (EM) are among the most important techniques in materials science and many other fields. Yet, there has historically been no direct physics-based method to go from an XRD pattern or STEM image to the actual atomic structure. ➡️ Last week, Prof. Sergei Kalinin group demonstrated the use of our MicroscopyGPT model on a real STEM image, helping to determine the underlying structure. Here’s an example of obtaining atomic structure from a powder diffraction pattern using DiffractGPT. ➡️ You can now literally go to atomgpt.org, upload your 2θ-intensity data, and get candidate atomic structures or try it on the Google Colab notebook. ➡️ Compared to other AI approaches in materials science, this seems to be one of the most promising areas of research, with exciting potential for large-scale commercial applications. ➡️Google Colab: https://lnkd.in/eeHhTRYS ➡️YouTube Demo: https://lnkd.in/ee2fHxKp #MaterialsScience #XRD #STEM #AI #MachineLearning #AtomisticModeling #MicroscopyGPT #DiffractGPT