Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Sequencing”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 379 records · Page 21

From sequence to protein structure and conformational dynamics with artificial intelligence/machine learning

The 2024 Nobel Prize in Chemistry was awarded in part for de novo protein structure prediction using AlphaFold2, an artificial intelligence/machine learning (AI/ML) model trained on vast amounts of sequence and three-dimensional structure data. AlphaFold2 and related models, including RoseTTAFold and ESMFold, employ specialized neural network architectures driven by attention mechanisms to infer relationships between sequence and structure. At a fundamental level, these AI/ML models operate on the long-standing hypothesis that the structure of a protein is determined by its amino acid sequence. More recently, AlphaFold2 has been adapted for the prediction of multiple protein conformations by subsampling multiple sequence alignments. Herein, we provide an overview of the deterministic relationship between sequence and structure, which was hypothesized over half a century ago with profound implications for the biological sciences ever since. We postulate that protein conformational dynamics are also determined, at least in part, by amino acid sequence and that this relationship may be leveraged for construction of AI/ML models dedicated to predicting protein conformational ensembles. Accordingly, we describe a conceptual model architecture, which may be trained on sequence data in combination with conformationally sensitive structural information, coming primarily from nuclear magnetic resonance (NMR) spectroscopy. Notwithstanding certain limitations in this context, NMR offers abundant structural heterogeneity conducive to conformational ensemble prediction. As NMR and other data continue to accumulate, sequence-informed prediction of protein structural dynamics with AI/ML has the potential to emerge as a transformative capability across the biological sciences.

Artificial intelligence↗

On the equivalence of some spectral sequences for Serre fibrations

Several different constructions of a spectral sequence for a Serre fibration {pi}:E{yields}B over a compact simply connected manifold B are considered in this paper. Namely, we consider the spectral sequence for the minimal model ({Lambda}Vx{Lambda}W,d) of the fibration, along with the spectral sequences arising from the Cech filtration in the complexes C*(U,A*{sub PL}({pi}{sup -1}(U))) and C*(U,S*({pi}{sup -1}(U))), where U=(U) is a covering of the base B. It is known that all these spectral sequences have the same terms E{sub 2}*{sup ,}*=H*(X)xH{sup *}(F) and converge to the cohomology of the total space E. A new natural isomorphism of these spectral sequences is constructed in every term E{sub r} with r{>=}2. It is also proved that in the case of a smooth locally trivial fibration these spectral sequences are isomorphic to the spectral sequences of the complex of smooth forms {Omega}*(E) and of the Cech-de Rham complex. It is therefore established that all these constructions give the same spectral sequence, starting from the E{sub 2} term. Bibliography: 9 titles.

97 MATHEMATICS AND COMPUTING↗

THE REST-FRAME OPTICAL LUMINOSITY FUNCTION OF CLUSTER GALAXIES AT z < 0.8 AND THE ASSEMBLY OF THE CLUSTER RED SEQUENCE

We present the rest-frame optical luminosity function (LF) of red-sequence galaxies in 16 clusters at 0.4 < z < 0.8 drawn from the ESO Distant Cluster Survey (EDisCS). We compare our clusters to an analogous sample from the Sloan Digital Sky Survey (SDSS) and match the EDisCS clusters to their most likely descendants. We measure all LFs down to M {approx} M * + (2.5-3.5). At z < 0.8, the bright end of the LF is consistent with passive evolution but there is a significant buildup of the faint end of the red sequence toward lower redshift. There is a weak dependence of the LF on cluster velocity dispersion for EDisCS but no such dependence for the SDSS clusters. We find tentative evidence that red-sequence galaxies brighter than a threshold magnitude are already in place, and that this threshold evolves to fainter magnitudes toward lower redshifts. We compare the EDisCS LFs with the LF of coeval red-sequence galaxies in the field and find that the bright end of the LFs agree. However, relative to the number of bright red galaxies, the field has more faint red galaxies than clusters at 0.6 < z < 0.8 but fewer at 0.4 < z < 0.6, implying differential evolution. We compare the total light in the EDisCS cluster red sequences to the total red-sequence light in our SDSS cluster sample. Clusters at 0.4 < z < 0.8 must increase their luminosity on the red sequence (and therefore stellar mass in red galaxies) by a factor of 1-3 by z = 0. The necessary processes that add mass to the red sequence in clusters predict local clusters that are overluminous as compared to those observed in the SDSS. The predicted cluster luminosities can be reconciled with observed local cluster luminosities by combining multiple previously known effects.

99 GENERAL AND MISCELLANEOUS↗

A TALE OF DWARFS AND GIANTS: USING A z = 1.62 CLUSTER TO UNDERSTAND HOW THE RED SEQUENCE GREW OVER THE LAST 9.5 BILLION YEARS

We study the red sequence in a cluster of galaxies at z = 1.62 and follow its evolution over the intervening 9.5 Gyr to the present day. Using deep YJK{sub s} imaging with the HAWK-I instrument on the Very Large Telescope, we identify a tight red sequence and construct its rest-frame i-band luminosity function (LF). There is a marked deficit of faint red galaxies in the cluster that causes a turnover in the LF. We compare the red-sequence LF to that for clusters at z < 0.8, correcting the luminosities for passive evolution. The shape of the cluster red-sequence LF does not evolve between z = 1.62 and z = 0.6 but at z < 0.6 the faint population builds up significantly. Meanwhile, between z = 1.62 and 0.6 the inferred total light on the red sequence grows by a factor of {approx}2 and the bright end of the LF becomes more populated. We construct a simple model for red-sequence evolution that grows the red sequence in total luminosity and matches the constant LF shape at z > 0.6. In this model the cluster accretes blue galaxies from the field whose star formation is quenched and who are subsequently allowed to merge. We find that three to four mergers among cluster galaxies during the 4 Gyr between z = 1.62 and z = 0.6 match the observed LF evolution between the two redshifts. The inferred merger rate is consistent with other studies of this cluster. Our result supports the picture that galaxy merging during the major growth phase of massive clusters is an important process in shaping the red-sequence population at all luminosities.

79 ASTRONOMY AND ASTROPHYSICS↗

A REVISED PARALLEL-SEQUENCE MORPHOLOGICAL CLASSIFICATION OF GALAXIES: STRUCTURE AND FORMATION OF S0 AND SPHEROIDAL GALAXIES

We update van den Bergh's parallel-sequence galaxy classification in which S0 galaxies form a sequence S0a-S0b-S0c that parallels the sequence Sa-Sb-Sc of spiral galaxies. The ratio B/T of bulge-to-total light defines the position of a galaxy in this tuning-fork diagram. Our classification makes one major improvement. We extend the S0a-S0b-S0c sequence to spheroidal ('Sph') galaxies that are positioned in parallel to irregular galaxies in a similarly extended Sa-Sb-Sc-Im sequence. This provides a natural 'home' for spheroidals, which previously were omitted from galaxy classification schemes or inappropriately combined with ellipticals. To motivate our juxtaposition of Sph and Im galaxies, we present photometry and bulge-disk decompositions of four rare, late-type S0s that bridge the gap between the more common S0b and Sph galaxies. NGC 4762 is an edge-on SB0bc galaxy with a very small classical-bulge-to-total ratio of B/T = 0.13 {+-} 0.02. NGC 4452 is an edge-on SB0 galaxy with an even tinier pseudobulge-to-total ratio of PB/T = 0.017 {+-} 0.004. It is therefore an SB0c. VCC 2048, whose published classification is S0, contains an edge-on disk, but its 'bulge' plots in the structural parameter sequence of spheroidals. It is therefore a disky Sph. And NGC 4638 is similarly a 'missing link' between S0s and Sphs-it has a tiny bulge and an edge-on disk embedded in an Sph halo. In the Appendix, we present photometry and bulge-disk decompositions of all Hubble Space Telescope Advanced Camera for Surveys Virgo Cluster Survey S0s that do not have published decompositions. We use these data to update the structural parameter correlations of Sph, S+Im, and E galaxies. We show that Sph galaxies of increasing luminosity form a continuous sequence with the disks (but not bulges) of S0c-S0b-S0a galaxies. Remarkably, the Sph-S0-disk sequence is almost identical to that of Im galaxies and spiral galaxy disks. We review published observations for galaxy transformation processes, particularly ram-pressure stripping of cold gas. We suggest that Sph galaxies are transformed, 'red and dead' Scd-Im galaxies in the same way that many S0 galaxies are transformed, red and dead Sa-Sc spiral galaxies.

79 ASTRONOMY AND ASTROPHYSICS↗

Combining protein sequences and structures with transformers and equivariant graph neural networks to predict protein function

Abstract Motivation Millions of protein sequences have been generated by numerous genome and transcriptome sequencing projects. However, experimentally determining the function of the proteins is still a time consuming, low-throughput, and expensive process, leading to a large protein sequence-function gap. Therefore, it is important to develop computational methods to accurately predict protein function to fill the gap. Even though many methods have been developed to use protein sequences as input to predict function, much fewer methods leverage protein structures in protein function prediction because there was lack of accurate protein structures for most proteins until recently. Results We developed TransFun—a method using a transformer-based protein language model and 3D-equivariant graph neural networks to distill information from both protein sequences and structures to predict protein function. It extracts feature embeddings from protein sequences using a pre-trained protein language model (ESM) via transfer learning and combines them with 3D structures of proteins predicted by AlphaFold2 through equivariant graph neural networks. Benchmarked on the CAFA3 test dataset and a new test dataset, TransFun outperforms several state-of-the-art methods, indicating that the language model and 3D-equivariant graph neural networks are effective methods to leverage protein sequences and structures to improve protein function prediction. Combining TransFun predictions and sequence similarity-based predictions can further increase prediction accuracy. Availability and implementation The source code of TransFun is available at https://github.com/jianlin-cheng/TransFun.

59 BASIC BIOLOGICAL SCIENCES↗

Evaluation of earthquake stress parameters and its scaling during the 2016-2017 Amatrice-Norcia-Visso sequence -- Part I

The Amatrice–Norcia–Visso sequence is distinguished by complex behaviour that is somewhat atypical of main-shock–aftershock sequences, as there were multiple large main shocks that continued for months. Here, we focus on the Amatrice sequence (main shock 2016 August 24, M w = 5.97) to evaluate the apparent stress values and magnitude-dependent scaling in order to improve our knowledge of processes that control small and large earthquakes within this active region of Italy. Apparent stress is proportional to the ratio of radiated seismic energy and seismic moment, and as such, these stress parameters play an important role in hazard prediction as they have a strong effect on the observed and predicted ground shaking. We analyse 83 events of the sequence from 2016 August 24 to October 16, within a radius of 20 km from the main shock and with an M w ranging between 5.97 and 2.72. Taking advantage of the averaging nature of coda waves, we analyse coda-envelope-based spectral ratios between neighbouring event pairs. We use equations proposed by Walter et al. to consider stable, low-frequency and high-frequency spectral ratio levels which provide measures of the corner frequency and apparent stress ratios of the events within the sequence. The results demonstrate non-self-similar behaviour within the sequence, indicating a change in dynamics between the largest events and the smaller aftershocks. The apparent stress and corner frequency estimates are compared to those obtained by Malagnini and Munafò who utilized hundreds of direct S-wave spectral ratio measurements to obtain their results. Although our analysis is based only on 83 events, our results are in very good agreement, demonstrating once more that the use of coda waves is very stable and provides lower variance measures than those using direct waves. A comparison with recent Central Apennines source scaling models derived from various seismic sequences (1997–1998 Colfiorito, 2002 San Giuliano di Puglia, 2009 L'Aquila) shows that the Amatrice sequence source scaling in this study is well represented by the models proposed by Pacor et al. and Malagnini and Mayeda.

58 GEOSCIENCES↗

A minimal motif for sequence recognition by mitochondrial transcription factor A (TFAM)

Abstract Mitochondrial transcription factor A (TFAM) plays a critical role in mitochondrial transcription initiation and mitochondrial DNA (mtDNA) packaging. Both functions require DNA binding, but in one case TFAM must recognize a specific promoter sequence, while packaging requires coating of mtDNA by association with non sequence-specific regions. The mechanisms by which TFAM achieves both sequence-specific and non sequence-specific recognition have not yet been determined. Existing crystal structures of TFAM bound to DNA allowed us to identify two guanine-specific interactions that are established between TFAM and the bound DNA. These interactions are observed when TFAM is bound to both specific promoter sequences and non-sequence specific DNA. These interactions are established with two guanine bases separated by 10 random nucleotides (GN10G). Our biochemical results demonstrate that the GN10G consensus is essential for transcriptional initiation and contributes to facilitating TFAM binding to DNA substrates. Furthermore, we report a crystal structure of TFAM in complex with a non sequence-specific sequence containing a GN10G consensus. The structure reveals a unique arrangement in which TFAM bridges two DNA substrates while maintaining the GN10G interactions. We propose that the GN10G consensus is key to facilitate the interaction of TFAM with DNA.

59 BASIC BIOLOGICAL SCIENCES↗

Arrayed in vivo barcoding for multiplexed sequence verification of plasmid DNA and demultiplexing of pooled libraries

Sequence verification of plasmid DNA is critical for many cloning and molecular biology workflows. To leverage high-throughput sequencing, several methods have been developed that add a unique DNA barcode to individual samples prior to pooling and sequencing. However, these methods require an individual plasmid extraction and/or in vitro barcoding reaction for each sample processed, limiting throughput and adding cost. Here, we develop an arrayed in vivo plasmid barcoding platform that enables pooled plasmid extraction and library preparation for Oxford Nanopore sequencing. This method has a high accuracy and recovery rate, and greatly increases throughput and reduces cost relative to other plasmid barcoding methods or Sanger sequencing. We use in vivo barcoding to sequence verify >45 000 plasmids and show that the method can be used to transform error-containing dispersed plasmid pools into sequence-perfect arrays or well-balanced pools. In vivo barcoding does not require any specialized equipment beyond a low-overhead Oxford Nanopore sequencer, enabling most labs to flexibly process hundreds to thousands of plasmids in parallel.

59 BASIC BIOLOGICAL SCIENCES↗

Estimating Protistan Diversity Using High‐Throughput Sequencing

Abstract Sequencing hypervariable regions from the 18S rRNA gene is commonly employed to characterize protistan biodiversity, yet there are concerns that short reads do not provide the same taxonomic resolution as full‐length sequences. A total of 7,432 full‐length sequences were used to perform an in silico analysis of how sequences of various lengths and target regions impact downstream ecological interpretations. Sequences that were longer than 400 nucleotides and included the V4 hypervariable region generated results similar to those derived from full‐length 18S rRNA gene sequences. Present high‐throughput sequencing capabilities are approaching protistan diversity estimation comparable to whole gene sequences.

Hu, Sarah K.↗

SU-E-J-216: A Sequence Independent Approach for Quantification of MR Image Deformations From Brachytherapy Applicators

Purpose: MRI is increasingly used as a single imaging modality for brachytherapy treatment planning. The presence of a brachytherapy applicator may cause distortions in the images, especially at higher field strengths. Our aim is to develop a procedure to quantify these distortions theoretically for any MR-sequence and to verify the estimated deformations for clinical sequences. Methods: Image distortions due to perturbation of the B0-field are proportional to the ratio of the induced frequency shift and the read-out bandwidth of the applied sequence. By reconstructing a frequency-shift map from the phase data from a multi-echo sequence, distortions can be calculated for any MR-sequence. Verification of this method for estimating distortions was performed by acquiring images with opposing read-out directions and consequently opposing distortions. The applicator shift can be determined by rigidly matching these images. Clinically, T2W-TSE-images are used for this purpose. For pre-clinical tests, EPI-sequences with narrow read-out bandwidth (19.5–47.5Hz), consequently large distortions, were added to the set of clinical MRsequences. To quantify deformations of the Utrecht Interstitial CT/MR applicator (Elekta Brachytherapy) on a Philips Ingenia 3T MRI, pre-clinical tests were performed in a phantom with the applicator in water, followed by clinical validation. Results: Deformations observed in the narrow bandwidth EPI-images were well predicted using the frequency-shift, the latter giving an overestimation up to 30%/up to 1 voxel. For clinically applied MR-sequences distortions were well below the voxel size. In patient setup distortions determined from the frequency-shift map were at sub-voxel level (<0.7mm). Using T2W-images larger distortions were found (1–2mm). This discrepancy was caused by patient movement between/during acquisition of the T2W-images with opposing read-out directions. Conclusion: Phantom experiments demonstrated the feasibility of a clinical procedure for quantification of MR-image distortions for any MR-sequence. In a clinical set-up the distortions from a Utrecht interstitial CT/MR applicator are sub-voxel level. This work was partially funded by Elekta Brachytherpy.

60 APPLIED LIFE SCIENCES↗

SU-E-J-231: Comparison of Delineation Variability of Soft Tissue Volume and Position in Head-And-Neck Between Two T1-Weighted Pulse Sequences Using An MR-Simulator with Immobilization

Purpose: There is growing interests in applying MR-simulator(MR-sim) in radiotherapy but MR images subject to hardware, patient and pulse sequence dependent geometric distortion that may potentially influence target definition. This study aimed to evaluate the influence on head-and-neck tissue delineation, in terms of positional and volumetric variability, of two T1-weighted(T1w) MR sequences on a 1.5T MR-sim Methods: Four healthy volunteers were scanned (4 scans for each on different days) using both spin-echo (3DCUBE, TR/TE=500/14ms, TA=183s) and gradient-echo sequences (3DFSPGR, TE/TR=7/4ms, TA=173s) with identical coverage, voxel-size(0.8×0.8×1.0mm3), receiver-bandwidth(62.5kHz/pix) and geometric correction on a 1.5T MR-sim immobilized with personalized thermoplastic cast and head-rest. Under this setting, similar T1w contrast and signal-to-noise ratio were obtained, and factors other than sequence that might bias image distortion and tissue delineation were minimized. VOIs of parotid gland(PGR, PGL), pituitary gland(PIT) and eyeballs(EyeL, EyeR) were carefully drawn, and inter-scan coefficient-of-variation(CV) of VOI centroid position and volume were calculated for each subject. Mean and standard deviation(SD) of the CVs for four subjects were compared between sequences using Wilcoxon ranksum test. Results: The mean positional(<4%) and volumetric(<7%) CVs varied between tissues, majorly dependent on tissue inherent properties like volume, location, mobility and deformability. Smaller mean volumetric CV was found in 3DCUBE, probably due to its less proneness to tissue susceptibility, but only PGL showed significant difference(P<0.05). Positional CVs had no significant differences for all VOIs(P>0.05) between sequences, suggesting volumetric variation might be more sensitive to sequence-dependent delineation difference. Conclusion: Although 3DCUBE is considered less prone to tissue susceptibility-induced artifact and distortion, our preliminary data showed that both sequences had insignificant differences on positional and volumetric CV in most head-and-neck tissues except for PGL. This study is majorly limited in its small sample size. Influences of image contrasts(T1w v.s. T2w) and inter-observer difference have to be further investigated.

60 APPLIED LIFE SCIENCES↗

SU-G-IeP1-08: MR Geometric Distortion Dependency On Imaging Sequence, Acquisition Orientation and Receiver Bandwidth of a Dedicated 1.5T MR-Simulator

Purpose: To investigate the 3D geometric distortion of four potential MR sequences for radiotheraptic applications, and its dependency on sequence-type, acquisition-orientation and receiver-bandwidth from a dedicated 1.5T 700mm-wide bore MR-simulator (Magnetom-Aera, Sienmens Healthcare, Erlangen, Germany), using a large customized geometric accuracy phantom. Methods: This work studied 3D gradient-echo (VIBE) and spin-echo (SPACE) sequences for anatomical imaging; a specific ultra-short-TE sequence (PETRA) potentially for bone imaging and MR-based dosimetry; and a motion-insensitive sequence (BLADE) for dynamic applications like 4D-MRI. Integrated geometric-correction was employed, three orthogonal acquisition-orientations and up to three receiver-bandwidths were used, yielding 27 acquisitions for testing (Table 1a).A customized geometric accuracy phantom (polyurethane, MR/CT invisible, W×L×H:55×55×32.5cm3) was constructed and filled with 3892 spherical markers (6mm diameter, MR/CT visible) arranged on a 25mm-interval 3D isotropic-grid (Fig.1). The marker positions in MR images were quantitatively calculated and compared against those in the CT-reference using customized MatLab scripts. Results: The average distortion within various diameter-of-spherical-volumes (DSVs) and the usable DSVs under various distortion limits were measured (Tables 1b-c). It was observed that distortions fluctuated when sequence-type, acquisition-orientation or receiver-bandwidth changed (e.g. within 300mm-DSV, the lowest/highest average distortions of VIBE were 0.40mm/0.59mm, a 47.5% difference). According to AAPM-TG66 (<1mm distortion, left-most column of Table 1c), PETRA (Largest-DSV:253.9mm) has the potential on brain treatment, while BLADE (Largest-DSV:207.2mm) may need improvement for thoracic/abdominal applications. The results of VIBE (Largest-DSVs:294.3mm, the best among tested acquisitions) and SPACE (Largest-DSVs:267.7mm) suggests their potentials on head and neck applications. These Largest-DSVs were attained on different acquisition-orientations and receiver-bandwidths. Conclusion: Geometric distortion was shown to be dependent on sequence-type, acquisition-orientation and receiver-bandwidth. In the experiment, no configuration in any one of these factors could consistently reduce distortion while the others were varying. The distortion analysis result is a valuable guideline for sequence selection and optimization for MR-aided radiotherapy applications.

60 APPLIED LIFE SCIENCES↗

Single-Genome Sequencing of Hepatitis C Virus in Donor-Recipient Pairs Distinguishes Modes and Models of Virus Transmission and Early Diversification

Despite the recent development of highly effective anti-hepatitis C virus (HCV) drugs, the global burden of this pathogen remains immense. Control or eradication of HCV will likely require the broad application of antiviral drugs and development of an effective vaccine. A precise molecular identification of transmitted/founder (T/F) HCV genomes that lead to productive clinical infection could play a critical role in vaccine research, as it has for HIV-1. However, the replication schema of these two RNA viruses differ substantially, as do viral responses to innate and adaptive host defenses. These differences raise questions as to the certainty of T/F HCV genome inferences, particularly in cases where multiple closely related sequence lineages have been observed. To clarify these issues and distinguish between competing models of early HCV diversification, we examined seven cases of acute HCV infection in humans and chimpanzees, including three examples of virus transmission between linked donors and recipients. Using single-genome sequencing (SGS) of plasma vRNA, we found that inferred T/F sequences in recipients were identical to viral sequences in their respective donors. Early in infection, HCV genomes generally evolved according to a simple model of random evolution where the coalescent corresponded to the T/F sequence. Closely related sequence lineages could be explained by high multiplicity infection from a donor whose viral sequences had undergone a pretransmission bottleneck due to treatment, immune selection, or recent infection. These findings validate SGS, together with mathematical modeling and phylogenetic analysis, as a novel strategy to infer T/F HCV genome sequences.

59 BASIC BIOLOGICAL SCIENCES↗

Moleculo Long-Read Sequencing Facilitates Assembly and Genomic Binning from Complex Soil Metagenomes

ABSTRACT Soil metagenomics has been touted as the “grand challenge” for metagenomics, as the high microbial diversity and spatial heterogeneity of soils make them unamenable to current assembly platforms. Here, we aimed to improve soil metagenomic sequence assembly by applying the Moleculo synthetic long-read sequencing technology. In total, we obtained 267 Gbp of raw sequence data from a native prairie soil; these data included 109.7 Gbp of short-read data (~100 bp) from the Joint Genome Institute (JGI), an additional 87.7 Gbp of rapid-mode read data (~250 bp), plus 69.6 Gbp (>1.5 kbp) from Moleculo sequencing. The Moleculo data alone yielded over 5,600 reads of >10 kbp in length, and over 95% of the unassembled reads mapped to contigs of >1.5 kbp. Hybrid assembly of all data resulted in more than 10,000 contigs over 10 kbp in length. We mapped three replicate metatranscriptomes derived from the same parent soil to the Moleculo subassembly and found that 95% of the predicted genes, based on their assignments to Enzyme Commission (EC) numbers, were expressed. The Moleculo subassembly also enabled binning of >100 microbial genome bins. We obtained via direct binning the first complete genome, that of “ Candidatus Pseudomonas sp. strain JKJ-1” from a native soil metagenome. By mapping metatranscriptome sequence reads back to the bins, we found that several bins corresponding to low-relative-abundance Acidobacteria were highly transcriptionally active, whereas bins corresponding to high-relative-abundance Verrucomicrobia were not. These results demonstrate that Moleculo sequencing provides a significant advance for resolving complex soil microbial communities. IMPORTANCE Soil microorganisms carry out key processes for life on our planet, including cycling of carbon and other nutrients and supporting growth of plants. However, there is poor molecular-level understanding of their functional roles in ecosystem stability and responses to environmental perturbations. This knowledge gap is largely due to the difficulty in culturing the majority of soil microbes. Thus, use of culture-independent approaches, such as metagenomics, promises the direct assessment of the functional potential of soil microbiomes. Soil is, however, a challenge for metagenomic assembly due to its high microbial diversity and variable evenness, resulting in low coverage and uneven sampling of microbial genomes. Despite increasingly large soil metagenome data volumes (>200 Gbp), the majority of the data do not assemble. Here, we used the cutting-edge approach of synthetic long-read sequencing technology (Moleculo) to assemble soil metagenome sequence data into long contigs and used the assemblies for binning of genomes. Author Video : An author video summary of this article is available.

De Novo assembly↗

Sequence length scaling in vision transformers for scientific images on frontier

Vision Transformers (ViTs) are pivotal for foundational models in scientific imagery, including Earth science applications, due to their capability to process large sequence lengths. While transformers for text have inspired scaling sequence lengths in ViTs, adapting these for ViTs introduces unique challenges. We develop distributed sequence parallelism for ViTs, enabling them to handle up to 1M tokens. Our approach, leveraging DeepSpeed-Ulysses and Long-Sequence-Segmentation with model sharding, is the first to apply sequence parallelism in ViT training, achieving a 94% batch scaling efficiency on 2,048 AMD-MI250X GPUs. Evaluating sequence parallelism in ViTs, particularly in models up to 10B parameters, highlighted substantial bottlenecks. We countered these with hybrid sequence, pipeline, and flash attention strategies, to scale beyond single GPU memory limits. Our method significantly enhances climate modeling accuracy by 20% in temperature predictions, marking the first training of a vision transformer model to convergence with a sequence length of 188K tokens, using full self-attention.

Tsaris, Aristeidis (aris) [ORNL] (ORCID:0000000277↗

SEPATH: benchmarking the search for pathogens in human tissue whole genome sequence data leads to template pipelines

Background: Human tissue is increasingly being whole genome sequenced as we transition into an era of genomic medicine. With this arises the potential to detect sequences originating from microorganisms, including pathogens amid the plethora of human sequencing reads. In cancer research, the tumorigenic ability of pathogens is being recognized, for example, Helicobacter pylori and human papillomavirus in the cases of gastric non-cardia and cervical carcinomas, respectively. As of yet, no benchmark has been carried out on the performance of computational approaches for bacterial and viral detection within host-dominated sequence data. Results: We present the results of benchmarking over 70 distinct combinations of tools and parameters on 100 simulated cancer datasets spiked with realistic proportions of bacteria. mOTUs2 and Kraken are the highest performing individual tools achieving median genus-level F1 scores of 0.90 and 0.91, respectively. mOTUs2 demonstrates a high performance in estimating bacterial proportions. Employing Kraken on unassembled sequencing reads produces a good but variable performance depending on post-classification filtering parameters. These approaches are investigated on a selection of cervical and gastric cancer whole genome sequences where Alphapapillomavirus and Helicobacter are detected in addition to a variety of other interesting genera. Conclusions: We provide the top-performing pipelines from this benchmark in a unifying tool called SEPATH, which is amenable to high throughput sequencing studies across a range of high-performance computing clusters. SEPATH provides a benchmarked and convenient approach to detect pathogens in tissue sequence data helping to determine the relationship between metagenomics and disease.

Viral↗

Complete genome sequence of the phenanthrene-degrading soil bacterium Delftia acidovorans Cs1-4

Polycyclic aromatic hydrocarbons (PAH) are ubiquitous environmental pollutants and microbial biodegradation is an important means of remediation of PAH-contaminated soil. Delftia acidovorans Cs1-4 (formerly Delftia sp. Cs1-4) was isolated by using phenanthrene as the sole carbon source from PAH contaminated soil in Wisconsin. Its full genome sequence was determined to gain insights into a mechanisms underlying biodegradation of PAH. Three genomic libraries were constructed and sequenced: an Illumina GAii shotgun library (916,416,493 reads), a 454 Titanium standard library (770,171 reads) and one paired-end 454 library (average insert size of 8 kb, 508,092 reads). The initial assembly contained 40 contigs in two scaffolds. The 454 Titanium standard data and the 454 paired end data were assembled together and the consensus sequences were computationally shredded into 2 kb overlapping shreds. Illumina sequencing data was assembled, and the consensus sequence was computationally shredded into 1.5 kb overlapping shreds. Gaps between contigs were closed by editing in Consed, by PCR and by Bubble PCR primer walks. A total of 182 additional reactions were needed to close gaps and to raise the quality of the finished sequence. The final assembly is based on 253.3 Mb of 454 draft data (averaging 38.4 X coverage) and 590.2 Mb of Illumina draft data (averaging 89.4 X coverage). The genome of strain Cs1-4 consists of a single circular chromosome of 6,685,842 bp (66.7 %G+C) containing 6,028 predicted genes; 5,931 of these genes were protein-encoding and 4,425 gene products were assigned to a putative function. Genes encoding phenanthrene degradation were localized to a 232 kb genomic island (termed the phn island), which contained near its 3’ end a bacteriophage P4-like integrase, an enzyme often associated with chromosomal integration of mobile genetic elements. Other biodegradation pathways reconstructed from the genome sequence included: benzoate (by the acetyl-CoA pathway), styrene, nicotinic acid (by the maleamate pathway) and the pesticides Dicamba and Fenitrothion. Lastly, determination of the complete genome sequence of D. acidovorans Cs1-4 has provided new insights the microbial mechanisms of PAH biodegradation that may shape the process in the environment.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗