RESUMEN
Automatic pollen images recognition is crucial for pollinosis symptoms prevention and treatment. The problem of pollen recognition can be efficiently solved using deep learning, however neural networks require tens of thousands of images to generalize. At the same time, the existing open pollen images datasets are very small. In this paper, we present a novel open pollen dataset annotated for both detection and classification tasks. Based on our dataset we study learning from a small data using different state-of-the-art approaches. For the detection task we propose to use our new Bayesian RetinaNet network, which models aleatoric uncertainty. We compare it with the baseline RetinaNet and demonstrate that our model allows for higher detection precision. For the classification task we compare the impact of pre-training on the synthetic images from generative adversarial networks (GANs) and metric-based few-shot learning. Namely, we pre-trained our small convolutional neural network and Siamese neural network classifiers on synthetic pollen images generated by two GANs: StyleGAN and Self-attention GAN. The best classifier is the convolutional neural network pre-trained on StyleGAN images. Our best models achieved 96.3% of mean average precision for the detection task and 97.7% of F1 measure for the classification task on 13 pollen plant species. We have implemented the best models in our pollen recognition web service, which is available for palynologists on request.
RESUMEN
T-cell acute lymphoblastic leukemia (T-ALL) is a rare disease usually treated with intensive, high-dose consolidation chemotherapy followed by an allotransplant in a substantial number of patients. The data of the RALL-2009 study on 125 adult T-ALL patients suggest that similar total chemotherapy doses given less intensively over a longer interval without interruptions and with an auto- rather than an allotransplant produce outcomes like current more intensive protocols and an allotransplant: 9-year cumulative incidence of relapse (CIR), leukemia-free survival (LFS), and survival were 24% (95% CI 16-33%), 70% (95% CI 59-79%) and 62% (95% CI 51-72%). In a landmark analysis, subjects achieving a complete remission and receiving an autotransplant had a lower 9-year CIR (9% [95% CI 2-22%] vs. 29% [95% CI 16-43%]; p = 0.0076) and better LFS (91% [95% CI 79-98%] vs. 58% [95% CI 41-74%]; p = 0.0009) and survival (92% [95% CI 77-99%] vs. 60% [95% CI 44-77%]; p = 0.001) compared with subjects not receiving an autotransplant. In a multivariate analysis, white blood cells ≥100 × 109/L at study entry were significantly associated with worse LFS (HR = 2.842 [95% CI 1.131-7.143]; p = 0.0263) and survival (HR = 6.085 [95% CI 1.918-19.3]; p = 0.0022) because of more early deaths (HR = 2.42 [95% CI 1.04-5.67]; p = 0.041). Receiving an autotransplant correlated with a lower CIR (HR = 0.23 [95% CI 0.07-0.73]; p = 0.0136) and better LFS (HR = 0.27 [95% CI 0.08-0.85]; p = 0.0256) and survival (HR = 0.158 [95% CI 0.045-0.550]; p = 0.0037).
Asunto(s)
Protocolos de Quimioterapia Combinada Antineoplásica/uso terapéutico , Trasplante de Células Madre Hematopoyéticas , Leucemia-Linfoma Linfoblástico de Células Precursoras/tratamiento farmacológico , Adolescente , Adulto , Supervivencia sin Enfermedad , Femenino , Humanos , Masculino , Persona de Mediana Edad , Análisis Multivariante , Leucemia-Linfoma Linfoblástico de Células Precursoras/mortalidad , Leucemia-Linfoma Linfoblástico de Células Precursoras/terapia , Modelos de Riesgos Proporcionales , Recurrencia , Inducción de Remisión , Tasa de Supervivencia , Trasplante Autólogo , Resultado del Tratamiento , Adulto JovenRESUMEN
PykF is one of two pyruvate kinases in Escherichia coli K-12. lambdaP(L) was convergently integrated into the chromosome of the MG1655 strain, downstream of pykF, face-to-face with its native promoter. In the presence of lambdacIts857, efficient pykF ts-silencing was achieved when the 5'-terminus of the P(L)-originated antisense RNA (asRNA), consisting of the rrnG-AT sequence, converted elongation complexes of RNA polymerase to a form resistant to Rho-dependent transcription termination. pykF silencing was detected by the following features: (a) impaired growth of the strain when pykA was also disrupted and when using ribose as a non-phosphotransferase system-transporting carbon source; (b) a pattern of reduced synthesis of the full-sized pykF mRNA, mediated by reverse transcription PCR, and (c) a significant decrease in PykF activity. The advantages of anti-terminated convergent transcription were clearly manifested in the strains where the rho_a-terminator was inserted specifically to interrupt asRNA synthesis. Most likely, the target gene was silenced by transcriptional interference due to collisions between converging RNA polymerases, although, strictly, the role of cis-asRNA effects could not be excluded. While details of the mechanisms have yet to be determined, anti-terminated convergent transcription is a promising new technique for silencing other target genes.
Asunto(s)
Escherichia coli K12/genética , Proteínas de Escherichia coli/biosíntesis , Silenciador del Gen , Piruvato Quinasa/biosíntesis , Factor Rho/metabolismo , Transcripción Genética , Bacteriófago lambda/genética , ADN Bacteriano/química , ADN Bacteriano/genética , Datos de Secuencia Molecular , Profagos/genética , Análisis de Secuencia de ADN , Integración ViralRESUMEN
BACKGROUND: The development of modern producer strains with metabolically engineered pathways poses special problems that often require manipulating many genes and expressing them individually at different levels or under separate regulatory controls. The construction of plasmid-less marker-less strains has many advantages for the further practical exploitation of these bacteria in industry. Such producer strains are usually constructed by sequential chromosome modifications including deletions and integration of genetic material. For these purposes complex methods based on in vitro and in vivo recombination processes have been developed. RESULTS: Here, we describe the new scheme of insertion of the foreign DNA for step-by-step construction of plasmid-less marker-less recombinant E. coli strains with chromosome structure designed in advance. This strategy, entitled as Dual-In/Out, based on the initial Red-driven insertion of artificial phi80-attB sites into desired points of the chromosome followed by two site-specific recombination processes: first, the phi80 system is used for integration of the recombinant DNA based on selective marker-carrier conditionally-replicated plasmid with phi80-attP-site, and second, the lambda system is used for excision of inserted vector part, including the plasmid ori-replication and the marker, flanked by lambda-attL/R-sites. CONCLUSION: The developed Dual-In/Out strategy is a rather straightforward, but convenient combination of previously developed recombination methods: phages site-specific and general Red/ET-mediated. This new approach allows us to detail the design of future recombinant marker-less strains, carrying, in particular, rather large artificial insertions that could be difficult to introduce by usually used PCR-based Recombineering procedure. The developed strategy is simple and could be particularly useful for construction of strains for the biotechnological industry.