DNA Sequencing Data Analysis

Identifying cancer-related mutations accurately is a critical step in precision medicine. Today, we’ve published new research in Nature Biotechnology on 🧬DeepSomatic🧬, an AI-powered tool that uses machine learning to identify genetic variants, or mutations, in cancer cells more accurately than current methods. This work is aimed at helping researchers pinpoint what's driving a cancer and informing more effective treatment plans. Somatic variant detection is an integral part of cancer genomics analysis. While most methods have focused on short-read sequencing, long-read technologies offer potential advantages to discover variants in the hardest to sequence parts of the genome. 🧬 About the model:  DeepSomatic was rigorously trained on high-confidence data, a feat made possible by working with our partners at UC Santa Cruz. The model is capable of accurately differentiating actual genetic cancer variants from the technical artifacts introduced during sample preservation, addressing a critical hurdle in early detection. 🧬 Superior Accuracy and Clinical Impact:  DeepSomatic consistently outperformed other tools across all major sequencing platforms. It shows major improvements in identifying complex insertions and deletions (Indels). Furthermore, in a new study with partners at Children's Mercy, DeepSomatic successfully found ten small variants in pediatric leukemia cells that were missed by other tools. 🧬 Flexible and Broad Use:  The model is flexible, working across all major sequencing platforms, and can be applied to both tumor-normal and challenging tumor-only samples, extending its utility for complex cancer types. 🧬 Open Access:  We are making DeepSomatic and the CASTLE dataset openly available to the research community. DeepSomatic is the most recent addition to our 10-year journey developing open source methods for geneticists to study the genomes of humans, plants, and animals. We are excited to see how researchers and drug manufacturers will use these resources to develop more effective, personalized treatments for cancer patients. The ability to accurately identify these subtle genetic drivers is key to unlocking new therapies. More in our blog authored by Kishwar Shafin and Andrew Carroll: https://goo.gle/4n23gIB   Read the full article in Nature Biotechnology: https://lnkd.in/drxii8fz

High throughput sequencing metrics: Don't be a monster, review them before sending data to the triage team. One of the most important things to avoid when doing high throughput sequencing is 'bias.' Properly assessing your post-alignment and variant calling metrics is super important for ensuring that biased data is not used to generate a patient report. Here are some of my favorite metrics to keep an eye on: Percent High Quality (HQ) Reads Aligned - The percentage of HQ reads that actually align to the reference genome. Here HQ is defined as Q20 or better and this stat should be greater than 98%. GC Bias Plot - This is one of my favorites and for short-read data it should look like an upside down U with high AT and high GC regions showing slight bias (because of amplification) and for long-read methods this plot is usually flat. Any major deviations can indicate bias either from over amplification or if this is target capture data, bias in the capture process. Insert size - These plots show the size distribution of the sequenced inserts within your library. These should pretty closely mimic the distribution you see in fragment analysis. Percent Duplication - This is a measure of the number of perfectly duplicated reads in a dataset. The ideal here is less than 5% and usually if you see problems with read duplication you'll also see issues in the GCbias plot. Coverage - A measure of the average depth of coverage across the genome or your provided target capture probe set. Transition/Transversion Ratio (TiTv) - Transitions are A<>G or C<>T (substitutions within the purines and pyrimidines) and Transversions A<>C, A<>T, C<>G, and G<>T (conversion of a purine to a pyrimidine, etc). For genomes the expected TiTv is 2 and for exon capture panels it's 3. Major deviations from these values could indicate a bias during sequencing or sample degradation during storage. Strand Bias - A measure of the bias of the genotype calls made on the positive and negative strands. No bias means calls are the same on each of the complementary strands, high bias means the calls differ and high strand bias around variant calls could indicate an over-reporting of false positives. Target capture specific metrics: Fold80 Penalty - This is a measure of evenness or uniformity. The best captures are ones that have perfect uniformity. Fold80 penalty is defined as "fold over-coverage necessary to raise 80% of bases in targets to the mean coverage level." 1 is perfect, so any deviation from that indicates a bias in capture. The best captures are <1.5. Percent Reads On Target - This is a measure of how much sequencing is being wasted on non-specific binding or off target. This value can vary greatly depending on the size of your capture from 60-70% for an exome down to <20% for smaller capture panels. Deviations from the expected value can indicate bias.

🔬 Google’s DeepSomatic: AI for short- and long-read cancer genomics 🧬 Big step forward from Google Research — the new DeepSomatic model, just published in Nature Biotechnology, uses deep learning to identify somatic mutations (the DNA changes driving tumors) directly from sequencing data. 🧠 Why it matters: Accurately detecting tumor-specific mutations is key to precision oncology, but conventional tools often struggle across sequencing platforms and sample types. DeepSomatic applies the same AI principles behind DeepVariant to cancer genomes. 🚀 Highlights: - Works with short- and long-read sequencing (Illumina, PacBio, Nanopore) - Handles tumor–normal, tumor-only, and FFPE samples - Major improvement for indel detection — traditionally one of the hardest challenges - Released with a new benchmark dataset: CASTLE (Cancer Standards Long-read Evaluation) - Outperforms established tools like MuTect2, Strelka2, and ClairS 💡 While long-read sequencing is still rare in clinical oncology, tools like DeepSomatic signal a shift: AI + long-readscould soon deliver richer, more accurate tumor profiling for precision medicine. The model performs best with high-accuracy chemistries (PacBio HiFi or ONT duplex/Q20+), showing that modern long-read data can rival short-reads for small variant detection. 🔗 Links to the paper and blog are in the comments. #AI #Genomics #CancerResearch #DeepLearning #Bioinformatics #PrecisionOncology #LongReadSequencing #PacBio #OxfordNanopore #DeepSomatic #GoogleResearch #NatureBiotechnology

1/ You have a clear question: Is gene A and gene B co-expressed in my cell type of interest? You feel ready. You have single-cell data. 2/ The plan seems easy: Cluster the cells. Plot gene A vs gene B. Calculate a correlation. 3/ But... it's NOT that simple. Single-cell data are sparse. Meaning: lots of zeros. Zeros that can be technical noise, not real biology. 4/ Example: Gene A = 0 UMI Gene B = 100 UMI Are they uncorrelated? Or was gene A just not detected well? 5/ Complication #1: Sequencing depth varies A LOT between cells. In scRNAseq, a cell may have 400 UMIs or 20,000 UMIs. 6/ Complication #2: Measurement errors. Random dropout of genes makes "real" co-expression tricky to detect. 7/ Because of these issues, simple correlation on raw UMI counts can be misleading. You may find false positives everywhere. 8/ Why? High-depth cells have more "signal." Low-depth cells have more "noise." Mix them up, and correlations get distorted. 9/ One solution: Aggregate cells into "meta-cells" based on similarity (like k-nearest neighbors). Average UMIs across them. 10/ This process—pseudo-bulk or meta-cell analysis—reduces sparsity and depth variation. Giving you more reliable correlations. 11/ However, be careful: Oversmoothing artifacts happen. Normalization and imputation can create fake correlations. 12/ Methods like MAGIC or Scanorama may introduce spurious signals by over-smoothing sparse data. Be cautious using them. 13/ Better strategies: Use noise-aware statistical models that account for UMI noise and sequencing depth variation directly. 14/ CS-CORE: Explicitly models sequencing depth and measurement errors to estimate unbiased gene-gene correlations. 15/ BigSur: Analyzes raw counts without normalization. Avoids skewing distributions and reduces fake correlation signals. 16/ Want deeper insights? I wrote two detailed blog posts explaining step-by-step: Part 1: https://lnkd.in/eqknd8ZS 17/ Part 2 (with meta-cell tricks): https://lnkd.in/etNrBjJW 18/ Takeaways: * Know your data quirks * Watch for sparsity, sequencing depth artifacts * Be cautious with imputation methods 19/ Good bioinformatics is like good writing: Clear, careful, and grounded in the messy details of reality. Not just what should happen. I hope you've found this post helpful. Follow me for more. Subscribe to my FREE newsletter chatomics to learn bioinformatics https://lnkd.in/erw83Svn

Before you blame #biology, check your #sequencing. 🧬 One of the most common traps in multi-sample #single-cell work: you see a difference between samples, assume it's biological, and build your whole story around it — when part of it was baked in at the sequencing and alignment stage. Different mean reads per cell. Different sequencing saturation. Different mapping rates. These technical fingerprints can quietly masquerade as biology if you never look at them side by side. This is where #Sam Marsh's #scCustomize package earns its place in the workflow. It has a set of functions built specifically for plotting Cell Ranger sequencing/alignment metrics across all your libraries at once: https://lnkd.in/e4KaPv85 → Read_Metrics_10X() pulls every metrics_summary.csv from your Cell Ranger output into a single tidy data frame — and it handles the multi pipeline too, returning GEX and VDJ metrics separately. → Join it with your sample metadata (sequencing batch, experiment batch, etc.) and the plots suddenly answer a real question: is this variation tracking with biology, or with batch? → Seq_QC_Plot_Basic_Combined() and Seq_QC_Plot_Alignment_Combined() give you patchwork summaries of the key metrics in one call, with optional significance testing via ggpubr. → And for the eternal "which barcodes are real cells" question, Iterate_Barcode_Rank_Plot() produces clean, ggplot-based knee plots straight from your raw feature-barcode matrices. lesson: QC isn't a box you check before the "real" analysis. Looking at your sequencing metrics as a group, grouped by batch is often what saves you from chasing an artifact for three weeks. Plot your alignment metrics before you interpret your clusters. Future-you will be grateful. What’s your go-to visualization for catching batch effects before diving into clustering? Let's talk in the comments. #SingleCell #Bioinformatics #scRNAseq #ComputationalBiology

Spatial biology is moving quickly and offers much to be excited about. But there is still a big learning curve. Especially with analysis. If you are a scientist new to spatial techniques, how do you actually understand your data? We will be walking through a concrete lifecycle and best practices with scientists + engineers from Takara Bio. 1/ Turning sequencing data into counts with graphical Nextflow interfaces 2/ Ingesting + visualizing H5AD data 3/ Quality Control 4/ Normalization (especially understanding different methods) 5/ Feature Selection 6/ Spatially Variable Gene Identification 7/ Spatial Neighborhood Analysis 8/ Spatial Differential Expression 9/ Visualizing interesting genes over your tissue map We are watching the very exciting early days of spatial development and adoption in industry. This is such a natural + fun way to do science. Just look at your tissue of interest and let visual intuition guide biological questions.

🧬 K-mer Analysis: The Building Blocks of Genome Assembly 🔹 What is a K-mer? A k-mer is a substring of length 'k' from a DNA sequence. Think of them as puzzle pieces that help us reconstruct the complete genome picture! 🔹 Why K-mers Matter: 📊 Quality Control 📊 Genome Size Estimation 📊 Repeat Detection 📊 Assembly Graph Construction 🔹 Choosing the Right K-mer Size: Smaller K (15-21): ✅ Better handling of errors ✅ Higher coverage ❌ More repeat issues Larger K (31-127): ✅ Better repeat resolution ✅ More specific matches ❌ Requires higher coverage 🔹 K-mer Frequency Distribution: 📈 Single peak: Homozygous genome 📈 Two peaks: Heterozygous genome 📈 Multiple peaks: Potential contamination/repeats 🔹 Essential K-mer Tools: 🛠️ Jellyfish: Fast k-mer counting 🛠️ KMC: Memory-efficient counting 🛠️ GenomeScope: Genome characteristics 🛠️ BBTools: K-mer analysis suite 🔹 Common Applications: 1️⃣ Error Correction: Low-frequency k-mers → likely errors 2️⃣ Coverage Estimation: K-mer depth = read depth × (L-K+1)/L 3️⃣ Genome Size Estimation: Total bases ÷ average k-mer depth 🔹 Best Practices: ✅ Start with k=21 for most applications ✅ Use odd k values to avoid reverse complements ✅ Consider multiple k values for complex genomes ✅ Monitor memory usage for large datasets 🔹 Troubleshooting Tips: ❌ Issue: High memory usage ✅ Solution: Use disk-based tools like KMC ❌ Issue: Strange k-mer distribution ✅ Solution: Check for contamination/quality 🔹 Advanced Applications: 🧪 Metagenome Analysis 🧪 Variant Detection 🧪 Species Identification 🧪 Assembly Quality Assessment #Bioinformatics #GenomeAssembly #SequencingData #DataAnalysis

Your sequencer finished. Your geneticist hasn't. Sequencing costs have dropped below $1,000 per genome. The sequencer is no longer the bottleneck. Interpretation is! Variant filtering. ACMG classification. Database cross-referencing. Clinical correlation. All manual. All queued behind one overloaded geneticist. We've built this platform before, and we know exactly where it breaks down. A VCF goes in. A clinical-grade report comes out. Annotated, ACMG-classified, cross-referenced across 10+ databases. All done in minutes. If you're running into this wall, it's worth knowing how we've approached it 👉 https://lnkd.in/dMXpCCsf #clinicalgenomics #variantinterpretation #bioinformatics #genomics #ngsworkflows #nonstopio

Just published in Nature Communications (Nature Portfolio); we show how long-read sequencing (LRS) improves rare disease diagnosis, detecting variants missed by short-read sequencing. Key findings: - Optimized a wet bench protocol and a bioinformatics pipeline for haplotype-specific SNV, indel, CNV, and SV detection, annotation, filtration and clinical prioritization. - Developed “EpiMarker”, a tool that extracts DNA methylation profiles from long-read data, adding an epigenetic layer to genetic diagnostics. - Validated both genomic and epigenomic modules using 76 patient samples with known pathogenic variants across different diseases and mutation types. - Applied these tools to 51 undiagnosed rare disease patients—and uncovered 5 additional diagnoses (10% diagnostic yield increase!) beyond what short-read sequencing could detect. New findings were mainly due to CNVs, SVs, and methylation (phasing helped in a case with SNV + CNV). - Used methylation tagging to unambiguously diagnose spinal muscular atrophy (SMA), confirmed through paralog-specific variants (PSVs) deconvolution of SMN1/SMN2. 📄 Read more: [https://lnkd.in/dWEHRY9K] Work spearheaded by Shruti Sinha, PhD, with great support from Fatma Rabea and SathishKumar Ramaswamy Ph.D. Along with contributions from the team at Dubai Health and Mohammed Bin Rashid University of Medicine and Health Sciences (MBRU): Ikram Chekroun Maha El Naofal, CG (ASCP), MB (ASCP) Ruchi Jain, Ph.D Roudha Alfalasi Nour Halabi Sawsan Alyafei Massy Sh Hassani, PhD Shruti Shenbagam Alan Taylor Mohammed Uddin Dafil Mohamed Almarri Stefan Du Plessis Alawi Alsheikh-Ali Grateful for all the support by Oxford Nanopore Technologies and their teams. #Genomics #RareDisease #LongReadSequencing #Epigenetics #Diagnostics

I spent 3 months building what I couldn't find: a bioinformatics library that doesn't make me switch tools 10 times per analysis. Introducing Seqcore: A Modern Alternative to Biopython for High-Performance Bioinformatics Here's the thing about computational biology in 2025: - Your single-cell dataset has 500K cells - Your GPU sits idle during sequence analysis - You're importing Biopython, pandas, NumPy, scanpy, MDAnalysis, RDKit... all in one script I asked: What if ONE library handled genomics, proteomics, structures, AND molecules? Why another bioinformatics library? Biopython has been the gold standard for 20+ years. But modern bioinformatics demands more: - Single-cell datasets with millions of cells - GPU acceleration for deep learning pipelines - Unified API instead of juggling 10 different libraries - NumPy-native operations for seamless integration Real-World Benchmark: PBMC 3k Single-Cell Analysis: I just ran a complete scRNA-seq pipeline on the 10x Genomics PBMC dataset: Dataset: 15.2M molecules, 2,853 cells, 13,754 genes Pipeline: Load -> Count Matrix -> QC -> Normalize -> PCA -> Analysis Total runtime: ~30 seconds (pure Python, no specialized dependencies) [1/7] Loading Data... 15,185,957 molecules [2/7] Creating Count Matrix... 2,853 cells x 13,754 genes [3/7] Quality Control... 84% cells passing [4/7] Normalizing... [5/7] PCA... 26% variance in 50 components [6/7] Marker Gene Analysis... 11 immune markers [7/7] Cell Type Classification... Done. What makes it different from Biopython? 1. Vectorized operations - No more Python loops over sequences 2. GPU-ready - with sc.device("cuda") and you're accelerated 3. Memory efficient - 2-bit DNA encoding uses 4x less RAM 4. Modern Python - Type hints, context managers, clean API Is it faster than Biopython? For batch operations on large datasets: 2-10x faster For single sequences: About the same The real win isn't raw speed it's not having to context-switch between libraries. Key Features: - 2-bit DNA encoding (4x memory reduction) - Batch operations on sequence arrays - Interoperability with NumPy, pandas, Biopython, RDKit - GPU support via CuPy Try it: pip install seqcore GitHub: https://lnkd.in/gx8GuvRT What's the one bioinformatics tool you wish existed? #Bioinformatics #Python #Genomics #SingleCell #OpenSource #DataScience #SingleCell #DataScience #MachineLearning #ComputationalBiology #Stanford #Research #Science #Biotech