Computational biology and bioinformatics provide vast data gold-mines from protein sequences, ideal for Language Models (LMs) taken from Natural Language Processing (NLP). These LMs reach for new ...prediction frontiers at low inference costs. Here, we trained two auto-regressive models (Transformer-XL, XLNet) and four auto-encoder models (BERT, Albert, Electra, T5) on data from UniRef and BFD containing up to 393 billion amino acids. The protein LMs (pLMs) were trained on the Summit supercomputer using 5616 GPUs and TPU Pod up-to 1024 cores. Dimensionality reduction revealed that the raw pLM- embeddings from unlabeled data captured some biophysical features of protein sequences. We validated the advantage of using the embeddings as exclusive input for several subsequent tasks: (1) a per-residue (per-token) prediction of protein secondary structure (3-state accuracy Q3=81%-87%); (2) per-protein (pooling) predictions of protein sub-cellular location (ten-state accuracy: Q10=81%) and membrane versus water-soluble (2-state accuracy Q2=91%). For secondary structure, the most informative embeddings (ProtT5) for the first time outperformed the state-of-the-art without multiple sequence alignments (MSAs) or evolutionary information thereby bypassing expensive database searches. Taken together, the results implied that pLMs learned some of the grammar of the language of life . All our models are available through https://github.com/agemagician/ProtTrans .
The current generation of the Xeon Phi Knights Landing (KNL) processor provides a highly multi-threaded environment on which regular programming models such as MPIjopenMP can be used. Many factors ...impact the performance achieved by applications on these devices: one of the key points is the efficient exploitation of SIMD vector units, and one another is the memory access pattern. Works have been conducted to adapt a plasma turbulence application, namely Gysela, for this architecture. A set of different techniques have been used: standard vectorization techniques, auto-tuning of one computation kernel, switching to high-order scheme. As a result, KNL execution times have been reduced by up to a factor 3. This effort has also permitted to gain a speedup of 2x on Broadwell architecture and 3x on Skylake. Nice scalability curves up to a few thousands cores have been obtained on a strong scaling experiment. Incremental work meant a large payoff without resorting to using low-level intrinsics.
Bacterial infections can be treated with bacteriophages that show great specificity towards their bacterial host and can be genetically modified for different applications. However, whether and how ...bacteriophages can kill intracellular bacteria in human cells remains elusive. Here, using CRISPR/Cas selection, we have engineered a fluorescent bacteriophage specific for E. coli K1, a nosocomial pathogen responsible for urinary tract infections, neonatal meningitis and sepsis. By confocal and live microscopy, we show that engineered bacteriophages K1F-GFP and E. coli EV36-RFP bacteria displaying the K1 capsule, enter human cells via phagocytosis. Importantly, we show that bacteriophage K1F-GFP efficiently kills intracellular E. coli EV36-RFP in T24 human urinary bladder epithelial cells. Finally, we provide evidence that bacteria and bacteriophages are degraded by LC3-associated phagocytosis and xenophagy.
Computational biology and bioinformatics provide vast data gold-mines from protein sequences, ideal for Language Models taken from NLP. These LMs reach for new prediction frontiers at low inference ...costs. Here, we trained two auto-regressive models (Transformer-XL, XLNet) and four auto-encoder models (BERT, Albert, Electra, T5) on data from UniRef and BFD containing up to 393 billion amino acids. The LMs were trained on the Summit supercomputer using 5616 GPUs and TPU Pod up-to 1024 cores. Dimensionality reduction revealed that the raw protein LM-embeddings from unlabeled data captured some biophysical features of protein sequences. We validated the advantage of using the embeddings as exclusive input for several subsequent tasks. The first was a per-residue prediction of protein secondary structure (3-state accuracy Q3=81%-87%); the second were per-protein predictions of protein sub-cellular localization (ten-state accuracy: Q10=81%) and membrane vs. water soluble (2-state accuracy Q2=91%). For the per-residue predictions the transfer of the most informative embeddings (ProtT5) for the first time outperformed the state-of-the-art without using evolutionary information thereby bypassing expensive database searches. Taken together, the results implied that protein LMs learned some of the grammar of the language of life. To facilitate future work, we released our models at https://github.com/agemagician/ProtTrans.
Emergent Properties of Reduced-Genome Escherichia coli Pósfai, György; Plunkett, Guy III; Fehér, Tamás ...
Science (American Association for the Advancement of Science),
05/2006, Letnik:
312, Številka:
5776
Journal Article
Recenzirano
With the use of synthetic biology, we reduced the Escherichia coli K-12 genome by making planned, precise deletions. The multiple-deletion series (MDS) strains, with genome reductions up to 15%, were ...designed by identifying nonessential genes and sequences for elimination, including recombinogenic or mobile DNA and cryptic virulence genes, while preserving good growth profiles and protein production. Genome reduction also led to unanticipated beneficial properties: high electroporation efficiency and accurate propagation of recombinant genes and plasmids that were unstable in other strains. Eradication of stress-induced transposition evidently stabilized the MDS genomes and provided some of the new properties.
Deuterium (D), the second most abundant isotope of hydrogen is present in natural waters at an approximate concentration of 145-155 ppm (ca. 1.5E-4 atom/atom). D is known to influence various ...biological processes due to its physical and chemical properties, which significantly differ from those of hydrogen. For example, increasing D-concentration to >1000-fold above its natural abundance has been shown to increase the frequency of genetic mutations in several species. An interesting deterministic hypothesis, formulated with the intent of explaining the mechanism of D-mutagenicity is based on the calculation that the theoretical probability of base pairs to comprise two adjacent D-bridges instead of H-bridges is 2.3E-8, which is equal to the mutation rate of certain species. To experimentally challenge this hypothesis, and to infer the mutagenicity of D present at natural concentrations, we investigated the effect of a nearly 100-fold reduction of D concentration on the bacterial mutation rate. Using fluctuation tests, we measured the mutation rate of three Escherichia coli genes (cycA, ackA and galK) in media containing D at either <2 ppm or 150 ppm concentrations. Out of 15 pair-wise fluctuation analyses, nine indicated a significant decrease, while three marked the significant increase of the mutation/culture value upon D-depletion. Overall, growth in D-depleted minimal medium led to a geometric mean of 0.663-fold (95% confidence interval: 0.483-0.911) change in the mutation rate. This falls nowhere near the expected 10,000-fold reduction, indicating that in our bacterial systems, the effect of D abundance on the formation of point mutations is not deterministic. In addition, the combined results did not display a statistically significant change in the mutation/culture value, the mutation rate or the mutant frequency upon D-depletion. The potential mutagenic effect of D present at natural concentrations on E. coli is therefore below the limit of detection using the indicated methods.
Abstract
Objectives
We sought to characterize the carbapenem resistance mechanism of Bacteroides xylanisolvens 14880, an imipenem-resistant strain from Germany, and assess its prevalence.
Methods
...Antimicrobial susceptibilities were determined using agar dilution or Etest methodology and specific imipenemase activity was detected. The genomic sequence of B. xylanisolvens 14880 was determined and analysed for antibiotic resistance genes and genomic islands. We also used gene transfer to a carbapenem susceptible host, along with 5′-RACE, conventional PCR with capillary sequencing and RT–PCR-based screening.
Results
B. xylanisolvens 14880 displayed resistance to carbapenems and produced high specific imipenemase activity. Its genomic sequence was 6.1 Mbp and a class B1 β-lactamase gene (termed crxA) was detected in it. crxA was carried on a putative genomic island with insertion sequence (IS) elements and a putative GNAT (Gcn5-like acetyltransferase) toxin gene. Promoter localization by 5′-RACE and gene targeting to an imipenem-susceptible Bacteroides host indicated that it is activated by an IS1380-like IS element and it can confer carbapenem resistance. The PCR screening of Bacteroides strains showed that crxA was specific to B. xylanisolvens with a carriage rate of 16.7%.
Conclusions
B. xylanisolvens strains can harbour a carbapenem resistance gene, which has many similarities to the ‘cfiA system’: metallo-β-lactamase (MBL), IS element activation, carriage of a GNAT toxin gene, specific for a unique Bacteroides species with a significant prevalence.
The past decade has seen the emergence of multidrug resistant pathogens as a leading cause of death worldwide, reigniting interest in the field of phage therapy. Modern advances in the genetic ...engineering of bacteriophages have enabled several useful results including host range alterations, constitutive lytic growth, and control over phage replication. However, the slow licensing process of genetically modified organisms clearly inhibits the rapid therapeutic application of novel engineered variants necessary to fight mutant pathogens that emerge throughout the course of a pandemic. As a solution to this problem, we propose the SpyPhage system where a “scaffold” bacteriophage is engineered to incorporate a SpyTag moiety on its capsid head to enable rapid postsynthetic modification of their surfaces with SpyCatcher-fused therapeutic proteins. As a proof of concept, through CRISPR/Cas-facilitated phage engineering and whole genome assembly, we targeted a SpyTag capsid fusion to K1F, a phage targeting the pathogenic strain Escherichia coli K1. We demonstrate for the first time the cell-free assembly and decoration of the phage surface with two alternative fusion proteins, SpyCatcher-mCherry-EGF and SpyCatcher-mCherry-Rck, both of which facilitate the endocytotic uptake of the phages by a urinary bladder epithelial cell line. Overall, our work presents a cell-free phage production pipeline for the generation of multiple phenotypically distinct phages with a single underlying “scaffold” genotype. These phages could become the basis of next-generation phage therapies where the knowledge-based engineering of numerous phage variants would be quickly achievable without the use of live bacteria or the need to repeatedly license novel genetic alterations.
Background:
In patients with inflammatory bowel disease (IBD), Crohn’s disease (CD), and ulcerative colitis (UC), numerous cases of exacerbations could be observed after colonoscopy, raising the ...possible pathogenetic effect of colonic microbiota alterations in IBD flare.
Objectives:
We aimed to investigate the changes in the fecal microbiota composition in IBD patients influenced by the bowel preparation with sodium picosulfate.
Design:
We enrolled patients with IBD undergoing bowel preparation for colonoscopy in the prospective cohort study. The control group (Con) comprised non-IBD patients who underwent colonoscopy. Clinical data, blood, and stool samples were collected before colonoscopy (timepoint A), 3 days later (timepoint B), and 4 weeks later (timepoint C).
Methods:
Disease activity and gut microbiota changes were assessed at each timepoint. Fecal microbiota structure – at family level – was determined by sequencing the V4 region of the 16S rRNA gene. Statistical analysis included differential abundance analysis and Mann–Whitney tests.
Results:
Forty-one patients (9 CD, 13 UC, and 19 Con) were included. After bowel preparation, alpha diversity was lower in the CD group than in the UC (p = 0.01) and Con (p = 0.02) groups at timepoint B. Alpha diversity was significantly higher in the UC group than in the CD and Con (p = 0.03) groups at timepoint C. Beta diversity difference differed between the IBD and Con (p = 0.001) groups. Based on the differential abundance analysis, the Clostridiales family was increased, whereas the Bifidobacteriaceae family was decreased in CD patients compared to the Con at timepoint B.
Conclusions:
Bowel preparation may change the fecal microbial composition in IBD patients, which may have a potential role in disease exacerbation after bowel cleansing.