GRUU-Net: Integrated convolutional and gated recurrent neural network for cell segmentation.

Cell segmentation in microscopy images is a common and challenging task. In recent years, deep neural networks achieved remarkable improvements in the field of computer vision. The dominant paradigm in segmentation is using convolutional neural networks, less common are recurrent neural networks. In this work, we propose a new deep learning method for cell segmentation, which integrates convolutional neural networks and gated recurrent neural networks over multiple image scales to exploit the strength of both types of networks. To increase the robustness of the training and improve segmentation, we introduce a novel focal loss function. We also present a distributed scheme for optimized training of the integrated neural network. We applied our proposed method to challenging data of glioblastoma cell nuclei and performed a quantitative comparison with state-of-the-art methods. Insights on how our extensions affect training and inference are also provided. Moreover, we benchmarked our method using a wide spectrum of all 22 real microscopy datasets of the Cell Tracking Challenge.

Sequence and analysis of chromosome 3 of the plant Arabidopsis thaliana.

Arabidopsis thaliana is an important model system for plant biologists. In 1996 an international collaboration (the Arabidopsis Genome Initiative) was formed to sequence the whole genome of Arabidopsis and in 1999 the sequence of the first two chromosomes was reported. The sequence of the last three chromosomes and an analysis of the whole genome are reported in this issue. Here we present the sequence of chromosome 3, organized into four sequence segments (contigs). The two largest (13.5 and 9.2 Mb) correspond to the top (long) and the bottom (short) arms of chromosome 3, and the two small contigs are located in the genetically defined centromere. This chromosome encodes 5,220 of the roughly 25,500 predicted protein-coding genes in the genome. About 20% of the predicted proteins have significant homology to proteins in eukaryotic genomes for which the complete sequence is available, pointing to important conserved cellular functions among eukaryotes.

Simultaneous loading of 200 sample lanes for DNA sequencing on vertical and horizontal, standard and ultrathin gels.

We have developed a simple and efficient technique for automated parallel loading of >/=200 lanes on a 30 cm-wide gel in automated DNA sequencing, using porous filter materials and an associated manual or robotic system. The samples are loaded onto the teeth of a comb made of the porous material. The comb, with samples, is inserted directly above the straight edge of the polymerized gel. The samples are driven from the comb into the gel by the applied electrical field. A particularly advantageous aspect of this method is the elimination of the thin gel walls separating the sample wells in the standard gel loading technique. The time for sample loading is significantly reduced to a few minutes. The loading technique is applicable to horizontal or vertical systems, with standard or ultrathin gels.

Quantitative trait loci for refractoriness of Anopheles gambiae to Plasmodium cynomolgi B.

The severity of the malaria pandemic in the tropics is aggravated by the ongoing spread of parasite resistance to antimalarial drugs and mosquito resistance to insecticides. A strain of Anopheles gambiae, normally a major vector for human malaria in Africa, can encapsulate and kill the malaria parasites within a melanin-rich capsule in the mosquito midgut. Genetic mapping revealed one major and two minor quantitative trait loci (QTLs) for this encapsulation reaction. Understanding such antiparasite mechanisms in mosquitoes may lead to new strategies for malaria control.

Spectrum of spontaneous mutations in liver tissue of lacI transgenic mice.

The advent of transgenic technology has greatly facilitated the study of mutation in animals in vivo. The Big Blue mouse system, transgenic for the lacI gene, permits not only the quantification of mutations in different tissues but also provides for the generation of in vivo-derived mutational spectra. This report details the sequence alterations of 348 spontaneous mutations recovered from the liver of 6-8-week-old male Big Blue mice. The spectra recovered from two strains of mice, C57BI/6 and B6C3F1, were compared and found to be very similar. The predominant mutations are G:C-->A:T transitions, with 75% of these occurring at 5'-CpG-3' sequences. This mutational bias is consistent with deamination-directed mutation at methylated cytosine bases. The second most common class of mutations is G:C-->T:A transversions. A significant clonal expansion of mutants was found in several animals, and this was used to make an approximate correction of the mutant frequency such that the most conservative estimate of mutation frequency is presented. The establishment of this substantial database of spontaneous mutations in the liver of Big Blue mice is intended to serve as a reference against which mutations recovered after treatment can be compared.

Molecular analysis of mutations in T-lymphocytes from experienced Soviet cosmonauts.

Somatic mutation in five cosmonauts who have completed spaceflights of 7 to 365 days was analyzed using the clonal HPRT assay. The doses received in space by the cosmonauts ranged from 4 to 127 mGy. hprt mutant frequencies were 2.4-5.0-fold higher than age-corrected values established for healthy, unexposed subjects in western countries [Tates et al. (1991): Mutat Res 253: 199-213; Branda et al. (1993): Mutat Res 285: 267-279] and 2- to 3-fold higher than those determined for unexposed individuals residing in Russia [Jones et al. (1995): Mutat Res 338: 129-139]. A total of 107 collected mutant clones were analyzed by multiplex PCR. No excess of deletions was detected and their frequency did not correlate with either accumulated dose or the age of the cosmonauts. In 62 mutants cDNA was isolated by RT-PCR and sequenced. Those with splicing errors, as well as the mutants that did not produce cDNA, were further analyzed by the sequencing of exon(s)-containing fragments amplified from genomic DNA. The mutational spectrum recovered from the cosmonauts differed substantially from that of unexposed healthy subjects (P = 0.042), and exhibited an increased incidence of splicing errors, frameshifts, and complex mutations. Higher frequencies of contribution of AT-->GC transitions and GC-->TA transversions were also observed. The increased mutant frequencies and observed shifts in mutational spectra likely indicate a combination of potential influences, including environment, lifestyle, and occupational exposures. Further elucidation of these potential influences will require a more extensive study involving the general population sharing similar environment, cosmonauts in training and cosmonauts participating in space flights.

Spontaneous mutants recovered from liver and germ cell tissue of low copy number lacI transgenic rats.

The finding of a large discordance between animal species in their response to a carcinogenic challenge, has led to the realization that the useful extrapolation of animal test data to humans requires a better understanding of animal interspecies differences. With the development of transgenic shuttle vector based animal systems we are now able to study mutation of the same genetic target in both mice and rats. We have begun to analyze mutants recovered from rat lines carrying low copy numbers of the same lambda/lacI constructs carried by the Big Blue mouse. A large database on mutations in lacI transgenic mice is already available for comparison. The data indicate that the differences between the mutations recovered from rat liver and germ cell tissues are similar to those recovered from transgenic mice, but when compared with a large database of mutations available for mice, some site-to-site differences may exist. This study represents the first interspecies look into the molecular nature of mutations in the lacI transgenic rodents.

Sequencing and analysis of 51 kb on the right arm of chromosome XV from Saccharomyces cerevisiae reveals 30 open reading frames.

We have sequenced a region of 51 kb of the right arm from chromosome XV of Saccharomyces cerevisiae. The sequence contains 30 open reading frames (ORFs) of more than 100 amino acid residues. Thirteen new genes have been identified. Thirteen ORFs correspond to known yeast genes. One delta element and one tRNA gene were identified. Upstream of the RPO31 gene, encoding the largest subunit of RNA polymerase III, lies a Abf1p binding site. The nucleotide sequence data reported in this paper are available in the EMBL, GenBank and DDBJ nucleotide sequence databases under the Accession Number X90518.

An efficient laboratory protocol for the sequencing of large numbers of lacI mutants recovered from Big Blue transgenic animals.

Mutational spectra provide a powerful approach to investigate both the mutagenic potential and the mechanism of action of suspected mutagens and carcinogens. Recently, transgenic techniques have made it possible to generate mutational spectra in animals. Such a spectrum may consist of 50 to 200 mutants depending on the nature of the mutations, and many spectra can be generated depending on the design of the experiment. This report describes a practical approach for the processing and sequencing of large numbers of lacI mutants recovered from Big Blue animals.

Efficient low redundancy large-scale DNA sequencing at EMBL.

An efficient low redundancy DNA sequencing strategy should allow high accuracy determination of the consensus sequence on both strands of a DNA fragment from a minimal number of sequencing reactions with minimal overlap. At EMBL we developed a directed strategy for cosmid-scale sequencing based on primer walking, whereas most other sequencing projects of this scale rely on the random 'shotgun' strategy. In our strategy, highly accurate raw data are obtained from automated double-stranded Sanger dideoxy sequencing with inexpensive walking primers (8 to 10 $ per primer), T7 DNA polymerase and internal labelling by fluorescein-15- dATP on A.L.F. DNA sequencers (Pharmacia Biotech). The use of 60-cm long glass plates enables reading length of up to 1000 bases. Comparing various random and directed sequencing strategies in the course of the European Community yeast genome sequencing project on cosmids from chromosomes IX, XI and XV, primer walking was found to be the strategy resulting in the lowest possible redundancy of 2.6 to 2.8. Future development of the sequencing strategy is based on the new EMBL 2-dye sequencing device for simultaneous sequencing on both strands, and implementation of an initial limited random sequencing phase to reduce the number of walking primers required by a factor of 3, while still maintaining a low redundancy of approx. 3.

Complete DNA sequence of yeast chromosome XI.

The complete DNA sequence of the yeast Saccharomyces cerevisiae chromosome XI has been determined. In addition to a compact arrangement of potential protein coding sequences, the 666,448-base-pair sequence has revealed general chromosome patterns; in particular, alternating regional variations in average base composition correlate with variations in local gene density along the chromosome. Significant discrepancies with the previously published genetic map demonstrate the need for using independent physical mapping criteria.

Sequencing and analysis of 51.6 kilobases on the left arm of chromosome XI from Saccharomyces cerevisiae reveals 23 open reading frames including the FAS1 gene.

We have sequenced two segments containing a total of 51.6 kb of the left arm from chromosome XI of Saccharomyces cerevisiae. The first segment of 38.5 kb contains 18 open reading frames (ORFs) of more than 100 amino acid residues. Five ORFs encode known yeast genes, including the fatty acid synthase gene (FAS1). Three new yeast genes were discovered with homologies to non-yeast genes and ten new genes without homologies to any known sequences. The second segment of 13 kb contains five ORFs with two known yeast genes and three unknown genes. The sequences from cosmid pUKG041 were obtained entirely with the walking primer strategy resulting in a very low overall sequence redundancy of 2.8 and an average reading length of 443 bases.

Automated low-redundancy large-scale DNA sequencing by primer walking.

Low-redundancy automated DNA sequencing by primer walking is described. T7 DNA polymerase is used together with computer-selected walking primers and fluorescein-dATP as internal label to sequence large plasmids or cosmids directly on a standard DNA sequencer with an error rate below 1% up to 500 bases (in the unedited raw data). The low error rate allows efficient sequencing with low (2-3 times) redundancy. Plasmid subclones covering 20 kb of a cosmid insert were sequenced with an overall redundancy of 2.7 in the course of the European community Saccharomyces cerevisiae genome sequencing project. Neighboring plasmid subclones were linked by direct cosmid sequencing. Sets of ten walking primers are synthesized on the EMBL multiple segmental DNA synthesizer at low costs and used for sequencing with greater than 95% efficiency. The accuracy of the directed approach is improved by simultaneous walking on both strands by designing two primers in opposite directions in the same starting region. One primer is used to confirm sequence data on the opposite strand, and the other primer to obtain new sequence data.

Random mutagenesis of the human immunodeficiency virus type-1 trans-activator of transcription (HIV-1 Tat).

A new method is described for the direct construction of randomly mutagenized genes by applying the polymerase chain reaction (PCR) to an oligonucleotide synthesized using doped nucleotide reservoirs. We have demonstrated the utility of this method by generating a library of mutant HIV-1 tat genes. Several arbitrarily selected, inactive tat clones were sequenced to evaluate the extent of the mutagenesis. Moreover, fourteen recombinants encoding varying levels of transcriptional trans-activator activity were isolated by transient transfection of sub-library pools into a HeLa cell line bearing an HIV-LTR-chloramphenicol acetyltransferase (CAT) reporter gene. Sequence data revealed a spectrum of alterations including nucleotide substitutions, insertions, and deletions, suggesting that mutations arose from both the doped DNA synthesis and the subsequent PCR 'rescue' of full-length product. Sequence comparison between inactive and active Tat clones revealed a selection pressure against amino-acid substitutions within the N-terminal domains of Tat, indicating the importance of this region to trans-activation competence. In addition, single and double missense mutations within the basic-rich, TAR RNA-binding domain were seen to be tolerated within active Tat clones.

High-throughput automated DNA sequencing facility with fluorescent labels at the European Molecular Biology Laboratory.

One of the aims of the facility is to develop and push the automated on-line DNA sequencing gel technology to its limit in sequence throughput, which may be somewhere around 100 kilobases of sequence per device per day. Key new developments were initiated and applied in operation on the European Molecular Biology Laboratory (EMBL) automated sequencer and its commercial version A.L.F. (Pharmacia). Sequencing speed was increased by a factor of 10-20, up to 1500 bases per hour per clone on ultrathin (about 100 microns) gels, while the resolution and reading length were extended to 1000 bases on gels with 50 cm separation length, using fluorescein-15-*dATP as the internal label. With our sequencing strategy, closing about 40% of the sequence with "walking" primers and F-15-*dATP as internal label, we sequenced both strands of a cosmid insert of 38.5 kb in length, each strand twice, in only 430 sequencing reactions and with average reading of 380 bases per reaction.

Assembly and expression of a synthetic gene encoding the bovine glycoprotein hormone alpha-subunit.

The glycoprotein hormones are a family of alpha beta heterodimeric proteins which are responsible for gonadal and thyroid function. In previous studies we employed chimeric glycoprotein hormone beta-subunits to identify amino acid residues critical for binding to receptors and antibodies. To facilitate similar studies of the alpha-subunit of these hormones, we assembled a 406 bp synthetic gene which encodes the human alpha-subunit leader sequence and the secreted portion of the bovine alpha-subunit. It contains unique restriction sites that can be used for cassette mutagenesis or for making human/bovine alpha-subunit chimeras. The gene was assembled from eight long oligodeoxynucleotides in a single ligation and its structure verified by DNA sequencing. Co-transfection of COS-7 cells with the synthetic gene and the cDNA for human chorionic gonadotropin (hCG) beta-subunit resulted in the secretion of a functional alpha beta heterodimer which bound to luteinizing hormone receptors. The protein was recognized by several monoclonal antibodies including B109, an antibody to a conformational epitope which binds hCG but not the free bovine alpha-, human alpha-, or hCG beta-subunits. This suggests that the binding site for B109 may be formed by residues located primarily within the hCG beta-subunit and that formation of this epitope requires a change in conformation of the beta-subunit when it combines with the alpha-subunit.