Felix Kallenborn

Transfer RNAs (tRNAs) are essential for protein synthesis and are extensively modified to ensure their structure and function. Direct RNA sequencing with Oxford Nanopore Technologies enables positional modification analysis but is challenged by tRNAs’ short length, redundancy, and dense modifications. We present QutRNA2, a scalable workflow that includes GPU-accelerated local alignment, statistical filtering, pairwise error profile comparison, and customizable visualization. Achieving up to 25-fold speed gains over CPU methods, QutRNA2 identifies enzyme-dependent modifications in nuclear- and mitochondrial-encoded tRNAs, demonstrated in high-volume human and mouse samples. This open-source solution provides a comprehensive, multiplexing-compatible framework for tRNA analysis, addressing a key gap in current tools. QutRNA2 is released under Apache-2.0 license and is available at https://github.com/dieterich-lab/QutRNA2.

Background: The continually increasing volume of sequence data results in a growing demand for fast implementations of core algorithms. Computation of pairwise alignments based on dynamic programming is an important part in many bioinformatics pipelines and a major contributor to overall runtime due to the associated quadratic time complexity. This motivates the need for a library of efficient implementations on modern GPUs for a variety of alignment algorithms for different types of sequence data including DNA, RNA, and proteins. Results: Accelign is a library of accelerated pairwise sequence alignment algorithms for CUDA-enabled GPUs. Its parallelization strategy is based on a common wavefront design that can be adapted to support a variety of dynamic programming algorithms: local, global, and semi-global alignment of genomic and protein sequences with a variety of commonly used scoring schemes supporting one-to-one, one-to-many or all-to-all pairwise sequence alignments. This leads to a peak performance between 16.1 TCUPS and 9.1 TCUPS for computing optimal global alignment scores with linear gaps and affine gap penalties on a single RTX PRO 6000 Blackwell GPU, respectively. In addition, our library demonstrates significant speedups in several real-world case studies over prior CPU-based (SeqAn, Parasail, BSalign, EdLib, KSW2, WFA2, A*PA2) and GPU-based libraries (ADEPT, GASAL2), and can even outperform highly customized algorithms (WFA-GPU, CUDASW++4.0). Furthermore, the performance of our approach scales linearly with the number of employed GPUs, which makes it feasible to exploit multi-GPU nodes for increased processing speeds. Conclusion: Accelign provides significant speedups for commonly used pairwise alignment algorithms compared to prior implementations. It is freely available at https://github.com/fkallen/Accelign .

We introduce accelerations for deep learning inference with OpenFold and TensorRT that, combined with MMseqs2-GPU on an x86 system with one NVIDIA RTX PRO 6000 Blackwell Server Edition GPU, reach up to 131× faster inference compared to AlphaFold2. ARM-optimizations enable homology search on power efficient DGX Spark, and allow search beyond available GPU RAM on NVIDIA Grace Hopper Superchip at comparable overall folding speed. These accelerations enable high-throughput protein structure inference at no accuracy cost using familiar tools.

The continually increasing volume of DNA sequence data has resulted in a growing demand for fast implementations of core algorithms. Computation of pairwise alignments between candidate haplotypes and sequencing reads using Pair-HMMs is a key component in DNA variant calling tools such as the GATK HaplotypeCaller but can be highly time consuming due to its quadratic time complexity and the large number of pairs to be aligned. Unfortunately, previous approaches to accelerate this task using the massively parallel processing capabilities of modern GPUs are limited by inefficient memory access schemes. This established the need for significantly faster solutions. We address this need by presenting gpuPairHMM – a novel GPU-based parallelization scheme for the dynamic-programming based Pair-HMM forward algorithm based on wavefronts and warp-shuffles. It gains efficiency by minimizing both memory accesses and instructions. We show that our approach achieves close-to-peak performance on several generations of modern CUDA-enabled GPUs (Volta, Ampere, Ada, Hopper, Blackwell). It also outperforms prior implementations on GPUs, CPUs, and FPGAs by a factor of at least 11.7, 14.2, and 19.8, respectively. gpuPairHMM is publicly available at https://github.com/asbschmidt/gpuPairHMM.

Background: All-Food-Sequencing (AFS) is a method for untargeted metagenomic analysis that allows for the detection and quantification of food ingredients. While this approach avoids some of the shortcomings of targeted PCR-based methods, its performance depends on sequencing technologies, taxonomic classification tools, and genomic reference databases. Results: AFS-MetaCache2 implements an improved reference database construction mechanism compared to prior approaches. To demonstrate the effectiveness to AFS, we sequenced sausages composed of mammalian and avian species using both short-read (Illumina) and long-read (Oxford Nanopore Technologies) platforms. While both approaches reliably detect the main components, our comparison shows that long-read sequencing is superior in terms of both quantification accuracy and false positive rates. The evaluation of representative metagenomic tools (Kraken2+Bracken, KrakenUniq, AFS-MetaCache1) demonstrates that AFS-MetaCache2 yields the best accuracy and fastest database build times, while reducing peak main memory consumption. It thus allows for efficient scaling to large reference genome sets. Conclusion: Our study suggests that deep sequencing of total genomic DNA from samples with heterogeneous taxon composition, using 3rd generation sequencing technology followed by metagenomic analysis with AFS-MetaCache2, is a valuable approach for bio-surveillance of food ingredients. Our software is available at https://github.com/muellan/metacache.

Background: The continually increasing volume of sequence data results in a growing demand for fast implementations of core algorithms. Computation of pairwise alignments based on dynamic programming is an important part in many bioinformatics pipelines and a major contributor to overall runtime due to the associated quadratic time complexity. This motivates the need for a library of efficient implementations on modern GPUs for a variety of alignment algorithms for different types of sequence data including DNA, RNA, and proteins. Results: Accelign is a library of accelerated pairwise sequence alignment algo- rithms for CUDA-enabled GPUs. Its parallelization strategy is based on a common wavefront design that can be adapted to support a variety of dynamic programming algorithms: local, global, and semi-global alignment of genomic and protein sequences with a variety of commonly used scoring schemes supporting one-to-one, one-to-many or all-to-all pairwise sequence alignments. This leads to a peak performance between 16.1 TCUPS and 9.1 TCUPS for computing optimal global alignment scores with linear gaps and affine gap penalties on a single RTX PRO 6000 Blackwell GPU, respectively. In addition, our library demonstrates sig- nificant speedups in several real-world case studies over prior CPU-based (SeqAn, Parasail, BSalign) and GPU-based libraries (ADEPT, GASAL2), and can even outperform highly customized algorithms (WFA-GPU, CUDASW++4.0). Fur- thermore, the performance of our approach scales linearly with the number of employed GPUs, which makes it feasible to exploit multi-GPU nodes for increased processing speeds. Conclusion: Accelign provides significant speedups for commonly used pairwise alignment algorithms compared to prior implementations. It is freely available at https://github.com/fkallen/Accelign.

Transfer RNAs (tRNAs) are essential for protein synthesis and are extensively modified to ensure their structure and function. Direct RNA sequencing with Oxford Nanopore Technologies enables positional modification analysis but is challenged by tRNAs’ short length, redundancy, and dense modifications. We present QutRNA2, a scalable workflow featuring GPU-accelerated local alignment, statistical filtering, pairwise error profile comparison, and customizable visualization. Achieving up to 25-fold speed gains over CPU methods, QutRNA2 identifies enzyme-dependent modifications in nuclear- and mitochondrial-encoded tRNAs, demonstrated in human and mouse samples. This open-source solution provides a comprehensive, multiplexing-compatible framework for tRNA analysis, addressing a key gap in current tools.

Rapidly growing protein databases demand faster sensitive search tools. Here the graphics processing unit (GPU)-accelerated MMseqs2 delivers 6× faster single-protein searches than CPU methods on 2 × 64 cores, speeds previously requiring large protein batches. For larger query batches, it is the most cost-effective solution, outperforming the fastest alternative method by 2.4-fold with eight GPUs. It accelerates protein structure prediction with ColabFold 31.8× over the standard AlphaFold2 pipeline and protein structure search with Foldseek by 4–27×. MMseqs2-GPU is available under an open-source license at https://mmseqs.com/ .

Summary: Nucleotide conversion sequencing techniques are frequently used for the detection of various types of chemical modifications at nucleotide level. However, mapping of chemically treated reads to large reference sequences that contain only three nucleotides can be highly compute-intensive. We present RMapAlign3N—an efficient yet accurate tool for mapping of 3 N-reads to reference genomes or transcriptomes that leverages the power of modern multi-core CPUs. Our performance evaluation using real and simulated data shows that RMapAlign3N is faster and more scalable than prior CPU-based approaches including HISAT-3N, BSMAP, Bismark, and SLAM-DUNK for BS-seq and SLAM-seq data at competitive accuracy. Availability and Implementation: RMapAlign3N is open source software written in C++ and can be downloaded at http://github.com/muellan/rmapalign3n.

The continually increasing volume of DNA sequence data has resulted in a growing demand for fast implementations of core algorithms. Computation of pairwise alignments between candidate haplotypes and sequencing reads using Pair-HMMs is a key component in DNA variant calling tools such as the GATK HaplotypeCaller but can be highly time consuming due to its quadratic time complexity and the large number of pairs to be aligned. Unfortunately, previous approaches to accelerate this task using the massively parallel processing capabilities of modern GPUs are limited by inefficient memory access schemes. This established the need for significantly faster solutions. We address this need by presenting gpuPairHMM -- a novel GPU-based parallelization scheme for the dynamic-programming based Pair-HMM forward algorithm based on wavefronts and warp-shuffles. It gains efficiency by minimizing both memory accesses and instructions. We show that our approach achieves close-to-peak performance on several generations of modern CUDA-enabled GPUs (Volta, Ampere, Ada, Hopper). It also outperforms prior implementations on GPUs, CPUs, and FPGAs by a factor of at least 8.6, 10.4, and 14.5, respectively. gpuPairHMM is publicly available at https://github.com/asbschmidt/gpuPairHMM.