Computational Biology Resources

We’re getting better at reading genes. Now we’re learning how to read them in 3D. A new study introduces a method to resolve signal overlap in spatial transcriptomics data, one of the biggest technical bottlenecks in mapping gene expression inside intact tissue. In dense biological samples, transcripts from neighboring cells often overlap, making it difficult to accurately assign signals to the correct cellular source. This blurring limits how precisely we can reconstruct tissue architecture. By improving how overlapping signals are separated computationally in three-dimensional space, researchers can generate far more accurate maps of how cells are organized in situ. This doesn’t just refine the data, it changes the reliability of downstream biological interpretation. For neuroscience, this is particularly significant. The brain is a tightly packed 3D network of gradients, microenvironments and dynamic cellular interactions. Circuit function, disease progression and developmental processes all depend on spatial context. If our spatial resolution is compromised, our models of brain function are incomplete. As biology moves from bulk averages toward high-resolution spatial systems, segmentation accuracy becomes foundational infrastructure, not a minor technical upgrade. Precision in three dimensions is what enables precision in understanding. Source: Nature Biotechnology, 2026 — “Identifying 3D signal overlaps in spatial transcriptomics data with ovrlpy.” #Neuroscience #SpatialTranscriptomics #SystemsBiology #Genomics #BrainResearch #Biotechnology #Innovation #Research

🧬 What if the spatial structure of a tissue could be reconstructed without microscopy or optical coordinates? The recent study by Dahlberg et al. (Nature Communications, 2025) marks a turning point in spatial transcriptomics. By reanalyzing data from the Slide-tags method, the authors discovered something remarkable: the diffusion of DNA barcodes generates a latent cell–bead network that preserves the tissue’s geometry. 🔹 Instead of treating diffusion as “noise,” the team reinterpreted it as a spatial topological signal, using the algorithm STRND (Spatio-Topological Reconstruction by Network Discovery). STRND models each cell based on its random walk visitation profile within the network, allowing tissue reconstruction purely from the structure of molecular connections, without prior coordinates.] 📈 The results in human tonsil tissue are striking: Global coherence (CPD) R² = 0.915, Local neighborhood coherence (KNN) ≈ 0.69, Biological spatial correlation between reconstructed and real positions: 0.996, Spatially variable gene correlation: 0.986. In other words, histological patterns (B zones, T zones, germinal centers) were preserved with a fidelity sufficient for high-resolution biological interpretation. 💡 Practically, this means: Lower experimental cost (no optical decoding required). Greater accessibility—spatial data from sequencing alone. Higher robustness against loss of coordinates or imaging artifacts. And a new opportunity to model tissue space as a graph, creating an ideal foundation for Graph Neural Networks (GNNs) in digital pathology and spatial biomarker discovery. In essence, topology becomes the new map. Space is no longer “seen,” it is reconstructed from molecular data. 🔍 This approach opens a new frontier: moving from optical microscopy to “imaging-by-sequencing,” where tissue architecture emerges from its own molecular network. 🤔 Could this be the dawn of a more democratic, accessible, and computationally integrated form of spatial transcriptomics? How might this transform spatial biomarker validation in oncology and digital pathology? #SpatialTranscriptomics #DigitalPathology #AIinMedicine #STRND #SlideTags #NetworkBiology #OmicsIntegration #NatureCommunications #TranslationalAI 📘 If you're just starting your journey into the world of digital pathology, here's my free ebook to help you along the way: https://lnkd.in/edcZbMxk Paper link : https://lnkd.in/eFfApFaN

𝗪𝗵𝘆 𝗙𝗼𝘂𝗿𝗶𝗲𝗿 𝗧𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺 𝗶𝘀 𝘁𝗵𝗲 𝗕𝗮𝗰𝗸𝗯𝗼𝗻𝗲 𝗼𝗳 𝗠𝗲𝗱𝗶𝗰𝗮𝗹 𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 The Fourier Transform is a powerful mathematical tool that breaks down complex signals into their frequency components. Its role in modern medical technology is nothing short of revolutionary — enabling the analysis, reconstruction, and interpretation of biological signals that are vital for diagnosis and treatment. 𝗠𝗲𝗱𝗶𝗰𝗮𝗹 𝗜𝗺𝗮𝗴𝗶𝗻𝗴 🧲 MRI (Magnetic Resonance Imaging): MRI scanners collect raw data in the spatial frequency domain (k-space). The Inverse Fourier Transform reconstructs this into high-resolution images, allowing physicians to visualize internal organs, tissues, and abnormalities. 🖥 CT (Computed Tomography): While CT uses the Radon Transform, the Fourier Slice Theorem links it to the Fourier domain. CT reconstruction algorithms like filtered back-projection enhance image quality using Fourier-based filtering. 🔊 Ultrasound: Doppler ultrasound depends on the Fourier Transform to measure blood flow by analyzing frequency shifts in sound waves. It’s also used in beamforming to sharpen ultrasound images. 𝗦𝗶𝗴𝗻𝗮𝗹 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 ❤️ ECG (Electrocardiography): Analyzing the frequency content of ECG signals helps detect arrhythmias and assess autonomic nervous system activity. Tools like Power Spectral Density (PSD) rely on the Fourier Transform. 🧠 EEG (Electroencephalography): Brain activity across sleep stages, seizures, and consciousness states is identified by decomposing EEG signals into frequency bands — made possible through the Fourier Transform. 💪 EMG (Electromyography): Used to monitor muscle activity and fatigue, the Fourier Transform reveals how muscle signal frequencies shift during contraction or neuromuscular disorders. Final Thought From imaging to biosignal analysis, the Fourier Transform is a silent enabler of breakthroughs in medical science — transforming raw data into life-saving insights.

Most single-cell analyses tell you what changes between cell types or within cell types across conditions. But biology is often driven by something harder to see: 👉 brief, transient events that push cells from one state to another. And those are exactly what we’ve been missing. In our newly accepted paper, we introduce scTransient — a method to detect short-lived bursts of gene or protein activity along biological trajectories. Instead of just ordering cells in pseudotime, we: • Convert expression into continuous signals • Apply wavelet-based signal processing • Quantify transient behavior with a new score (TES) • Systematically identify genes driving state transitions Why this matters: These fleeting signals often mark regulatory checkpoints in development, differentiation, and disease… but they’re buried in noise and easy to miss We show that we can now reliably detect them across: • Synthetic benchmarks • iPSC-derived neuron development bulk proteomics data • Single-cell proteomics cell cycle data But the real story here is the people behind it. This work was driven by Alexandre Hutton, who led the development end-to-end — from core method design to implementation and validation. And it wouldn’t have been possible without Jesús Muñoz-Estrada, Ph.D., who played a key role in the experimental and biological interpretation side. This was a true computational + wet lab biological collaboration. I love that my lab can produce both ends in works like this. Stepping back: We’re moving from: “Which genes change?” to: “Which genes briefly spike at the exact moment biology decides what happens next?” Paper link is in the comments 👇

Real-Time Feature Extraction, meet the Savitzky–Golay Filter. Accurate real-time feature extraction does not necessarily require complex or data-driven methods. The example shown uses a 3rd-order Savitzky–Golay derivative filter to extract QRS features from a biomedical ECG waveform directly in the time domain, using just a fixed FIR implementation and simple zero-crossing and extrema detection. Savitzky–Golay differentiation analytically differentiates a locally fitted polynomial, producing a robust, smoothed numerical derivative that accurately estimates slope and curvature even in the presence of noise. In the QRS complex, this yields a structured derivative response whose zero crossings, extrema and sign changes are directly related to the underlying Q, R and S features. The combination of the fixed FIR Savitzky–Golay derivative filter with simple zero-crossing and extrema detection is computationally efficient, consisting only of FIR convolution followed by lightweight decision logic, and is therefore well suited to implementation on many Arm Cortex-M processors. Although illustrated here using a biomedical ECG waveform, the same approach can also be applied to infrared gas concentration measurements, where the amplitude of the gas-concentration waveform must be determined, as well as to test and measurement applications involving sinusoidal signals, where reliable amplitude estimation and identification of key extrema or features are required. 👉🏽 This example is covered in The Real-Time Edge Intelligence Solutions Handbook as part of the chapter on FIR filtering methods, illustrating how compact DSP techniques deliver robust, low-power solutions for commercial AIoT systems. For more information and to obtain a copy, visit the book’s homepage: https://lnkd.in/eMtNSpCN #DSP #EdgeAI #physicalAI #SignalProcessing #RTEIbook Jayakumar S Thad Meyer Ricardo Abdoel Eric Sondhi Rajiv Biswas Dan Boschen