Many gene sequences in eukaryotic genomes encode entire proteins or large segments of proteins that lack a well-structured three-dimensional fold. Disordered regions can be highly conserved between species in both composition and sequence and, contrary to the traditional view that protein function equates with a stable three-dimensional structure, disordered regions are often functional, in ways that we are only beginning to discover. Many disordered segments fold on binding to their biological targets (coupled folding and binding), whereas others constitute flexible linkers that have a role in the assembly of macromolecular arrays.

Intrinsically unstructured proteins and their functions.

Deep learning describes a class of machine learning algorithms that are capable of combining raw inputs into layers of intermediate features. These algorithms have recently shown impressive results across a variety of domains. Biology and medicine are data-rich disciplines, but the data are complex and often ill-understood. Hence, deep learning techniques may be particularly well suited to solve problems of these fields. We examine applications of deep learning to a variety of biomedical problems-patient classification, fundamental biological processes and treatment of patients-and discuss whether deep learning will be able to transform these tasks or if the biomedical sphere poses unique challenges. Following from an extensive literature review, we find that deep learning has yet to revolutionize biomedicine or definitively resolve any of the most pressing challenges in the field, but promising advances have been made on the prior state of the art. Even though improvements over previous baselines have been modest in general, the recent progress indicates that deep learning methods will provide valuable means for speeding up or aiding human investigation. Though progress has been made linking a specific neural network's prediction to input features, understanding how users should interpret these models to make testable hypotheses about the system under study remains an open challenge. Furthermore, the limited amount of labelled data for training presents problems in some domains, as do legal and privacy constraints on work with sensitive health records. Nonetheless, we foresee deep learning enabling changes at both bench and bedside with the potential to transform several areas of biology and medicine.

https://royalsocietypublishing.org/doi/pdf/10.1098/rsif.2017.0387

Opportunities and obstacles for deep learning in biology and medicine.

The 3,308,274-bp sequence of the chromosome of Lactobacillus plantarum strain WCFS1, a single colony isolate of strain NCIMB8826 that was originally isolated from human saliva, has been determined, and contains 3,052 predicted protein-encoding genes. Putative biological functions could be assigned to 2,120 (70%) of the predicted proteins. Consistent with the classification of L. plantarum as a facultative heterofermentative lactic acid bacterium, the genome encodes all enzymes required for the glycolysis and phosphoketolase pathways, all of which appear to belong to the class of potentially highly expressed genes in this organism, as was evident from the codon-adaptation index of individual genes. Moreover, L. plantarum encodes a large pyruvate-dissipating potential, leading to various end-products of fermentation. L. plantarum is a species that is encountered in many different environmental niches, and this flexible and adaptive behavior is reflected by the relatively large number of regulatory and transport functions, including 25 complete PTS sugar transport systems. Moreover, the chromosome encodes >200 extracellular proteins, many of which are predicted to be bound to the cell envelope. A large proportion of the genes encoding sugar transport and utilization, as well as genes encoding extracellular functions, appear to be clustered in a 600-kb region near the origin of replication. Many of these genes display deviation of nucleotide composition, consistent with a foreign origin. These findings suggest that these genes, which provide an important part of the interaction of L. plantarum with its environment, form a lifestyle adaptation region in the chromosome.

/pdf/complete-genome-sequence-of-lactobacillus-plantarum-wcfs1-3i6pkrsthv.pdf

Complete genome sequence of Lactobacillus plantarum WCFS1

Despite their name, synonymous mutations have significant consequences for cellular processes in all taxa. As a result, an understanding of codon bias is central to fields as diverse as molecular evolution and biotechnology. Although recent advances in sequencing and synthetic biology have helped to resolve longstanding questions about codon bias, they have also uncovered striking patterns that suggest new hypotheses about protein synthesis. Ongoing work to quantify the dynamics of initiation and elongation is as important for understanding natural synonymous variation as it is for designing transgenes in applied contexts.

/pdf/synonymous-but-not-the-same-the-causes-and-consequences-of-4xxuk0p9fz.pdf

Synonymous but not the same: the causes and consequences of codon bias.

The complete 1,751,377-bp sequence of the genome of the thermophilic archaeon Methanobacterium thermoautotrophicum deltaH has been determined by a whole-genome shotgun sequencing approach. A total of 1,855 open reading frames (ORFs) have been identified that appear to encode polypeptides, 844 (46%) of which have been assigned putative functions based on their similarities to database sequences with assigned functions. A total of 514 (28%) of the ORF-encoded polypeptides are related to sequences with unknown functions, and 496 (27%) have little or no homology to sequences in public databases. Comparisons with Eucarya-, Bacteria-, and Archaea-specific databases reveal that 1,013 of the putative gene products (54%) are most similar to polypeptide sequences described previously for other organisms in the domain Archaea. Comparisons with the Methanococcus jannaschii genome data underline the extensive divergence that has occurred between these two methanogens; only 352 (19%) of M. thermoautotrophicum ORFs encode sequences that are >50% identical to M. jannaschii polypeptides, and there is little conservation in the relative locations of orthologous genes. When the M. thermoautotrophicum ORFs are compared to sequences from only the eucaryal and bacterial domains, 786 (42%) are more similar to bacterial sequences and 241 (13%) are more similar to eucaryal sequences. The bacterial domain-like gene products include the majority of those predicted to be involved in cofactor and small molecule biosyntheses, intermediary metabolism, transport, nitrogen fixation, regulatory functions, and interactions with the environment. Most proteins predicted to be involved in DNA metabolism, transcription, and translation are more similar to eucaryal sequences. Gene structure and organization have features that are typical of the Bacteria, including genes that encode polypeptides closely related to eucaryal proteins. There are 24 polypeptides that could form two-component sensor kinase-response regulator systems and homologs of the bacterial Hsp70-response proteins DnaK and DnaJ, which are notably absent in M. jannaschii. DNA replication initiation and chromosome packaging in M. thermoautotrophicum are predicted to have eucaryal features, based on the presence of two Cdc6 homologs and three histones; however, the presence of an ftsZ gene indicates a bacterial type of cell division initiation. The DNA polymerases include an X-family repair type and an unusual archaeal B type formed by two separate polypeptides. The DNA-dependent RNA polymerase (RNAP) subunits A', A", B', B" and H are encoded in a typical archaeal RNAP operon, although a second A' subunit-encoding gene is present at a remote location. There are two rRNA operons, and 39 tRNA genes are dispersed around the genome, although most of these occur in clusters. Three of the tRNA genes have introns, including the tRNAPro (GGG) gene, which contains a second intron at an unprecedented location. There is no selenocysteinyl-tRNA gene nor evidence for classically organized IS elements, prophages, or plasmids. The genome contains one intein and two extended repeats (3.6 and 8.6 kb) that are members of a family with 18 representatives in the M. jannaschii genome.

Complete genome sequence of Methanobacterium thermoautotrophicum deltaH: functional analysis and comparative genomics.

We compare and contrast genome-wide compositional biases and distributions of short oligonucleotides across 15 diverse prokaryotes that have substantial genomic sequence collections. These include seven complete genomes (Escherichia coli, Haemophilus influenzae, Mycoplasma genitalium, Mycoplasma pneumoniae, Synechocystis sp. strain PCC6803, Methanococcus jannaschii, and Pyrobaculum aerophilum). A key observation concerns the constancy of the dinucleotide relative abundance profiles over multiple 50-kb disjoint contigs within the same genome. (The profile is rhoXY* = fXY*/fX*fY* for all XY, where fX* denotes the frequency of the nucleotide X and fY* denotes the frequency of the dinucleotide XY, both computed from the sequence concatenated with its inverted complementary sequence.) On the basis of this constancy, we refer to the collection [rhoXY*] as the genome signature. We establish that the differences between [rhoXY*] vectors of 50-kb sample contigs of different genomes virtually always exceed the differences between those of the same genomes. Various di- and tetranucleotide biases are identified. In particular, we find that the dinucleotide CpG=CG is underrepresented in many thermophiles (e.g., M. jannaschii, Sulfolobus sp., and M. thermoautotrophicum) but overrepresented in halobacteria. TA is broadly underrepresented in prokaryotes and eukaryotes, but normal counts appear in Sulfolobus and P. aerophilum sequences. More than for any other bacterial genome, palindromic tetranucleotides are underrepresented in H. influenzae. The M. jannaschii sequence is unprecedented in its extreme underrepresentation of CTAG tetranucleotides and in the anomalous distribution of CTAG sites around the genome. Comparative analysis of numbers of long tetranucleotide microsatellites distinguishes H. influenzae. Dinucleotide relative abundance differences between bacterial sequences are compared. For example, in these assessments of differences, the cyanobacteria Synechocystis, Synechococcus, and Anabaena do not form a coherent group and are as far from each other as general gram-negative sequences are from general gram-positive sequences. The difference of M. jannaschii from low-G+C gram-positive proteobacteria is one-half of the difference from gram-negative proteobacteria. Interpretations and hypotheses center on the role of the genome signature in highlighting similarities and dissimilarities across different classes of prokaryotic species, possible mechanisms underlying the genome signature, the form and level of genome compositional flux, the use of the genome signature as a chronometer of molecular phylogeny, and implications with respect to the three putative eubacterial, archaeal, and eukaryote domains of life and to the origin and early evolution of eukaryotes.

Compositional biases of bacterial genomes and evolutionary implications.

What Drives Codon Choices in Human Genes

Several bacterial genomes exhibit preference for G over C on the DNA leading strand extending from the origin of replication to the ter-region in the genomes of Escherichia coli, Mycoplasma genitalium, Bacillus subtilis, and marginally in Haemophilus influenzae, Mycoplasma pneumoniae, and Helicobacter pylori. Strand compositional asymmetry is not observed in the cyanobacterium Synechocystis sp. genome nor in the archaeal genomes of Methanococcus jannaschii, Methanobacterium thermoautotrophicum, and Archaeoglobus fulgidus. A strong strand compositional asymmetry is observed in β-type but not α- or γ-type human herpesviruses featuring G > C downstream of oriL and C > G upstream of oriL. Dinucleotide relative abundances (i.e., dinucleotide representations normalized by the component nucleotide frequencies) are consonant with respect to the leading and lagging strands. Strand compositional asymmetry may reflect on differences in replication synthesis of the leading versus lagging strand, on differences between template and coding strand associated with transcription-coupled repair mechanisms, on differences in gene density between the two strands, on differences in residue and codon biases in relation to gene function, expression level, or operon organization, or on differences in single or context-dependent base mutational rates. The absence of strand asymmetry in the archaeal genomes may reflect the presence of multiple origins of replication.

/pdf/strand-compositional-asymmetry-in-bacterial-and-large-viral-5fkfbeweqt.pdf

Strand compositional asymmetry in bacterial and large viral genomes

We present a comparative proteome analysis of the five complete eukaryotic genomes (human, Drosophila melanogaster, Caenorhabditis elegans, Saccharomyces cerevisiae, Arabidopsis thaliana), focusing on individual and multiple amino acid runs, charge and hydrophobic runs. We found that human proteins with multiple long runs are often associated with diseases; these include long glutamine runs that induce neurological disorders, various cancers, categories of leukemias (mostly involving chromosomal translocations), and an abundance of Ca(2 +) and K(+) channel proteins. Many human proteins with multiple runs function in development and/or transcription regulation and are Drosophila homeotic homologs. A large number of these proteins are expressed in the nervous system. More than 80% of Drosophila proteins with multiple runs seem to function in transcription regulation. The most frequent amino acid runs in Drosophila sequences occur for glutamine, alanine, and serine, whereas human sequences highlight glutamate, proline, and leucine. The most frequent runs in yeast are of serine, glutamine, and acidic residues. Compared with the other eukaryotic proteomes, amino acid runs are significantly more abundant in the fly. This finding might be interpreted in terms of innate differences in DNA-replication processes, repair mechanisms, DNA-modification systems, and mutational biases. There are striking differences in amino acid runs for glutamine, asparagine, and leucine among the five proteomes.

https://www.pnas.org/content/pnas/99/1/333.full.pdf

Amino acid runs in eukaryotic proteomes and disease associations

Our approach in predicting gene expression levels relates to codon usage differences among gene classes. In prokaryotic genomes, genes that deviate strongly in codon usage from the average gene but are sufficiently similar in codon usage to ribosomal protein genes, to translation and transcription processing factors, and to chaperone-degradation proteins are predicted highly expressed (PHX). By these criteria, PHX genes in most prokaryotic genomes include those encoding ribosomal proteins, translation and transcription processing factors, and chaperone proteins and genes of principal energy metabolism. In particular, for the fast-growing species Escherichia coli, Vibrio cholerae, Bacillus subtilis, and Haemophilus influenzae, major glycolysis and tricarboxylic acid cycle genes are PHX. In Synechocystis, prime genes of photosynthesis are PHX, and in methanogens, PHX genes include those essential for methanogenesis. Overall, the three protein families—ribosomal proteins, protein synthesis factors, and chaperone complexes—are needed at many stages of the life cycle, and apparently bacteria have evolved codon usage to maintain appropriate growth, stability, and plasticity. New interpretations of the capacity of Deinococcus radiodurans for resistance to high doses of ionizing radiation is based on an excess of PHX chaperone-degradation genes and detoxification genes. Expression levels of selected classes of genes, including those for flagella, electron transport, detoxification, histidine kinases, and others, are analyzed. Flagellar PHX genes are conspicuous among spirochete genomes. PHX genes are positively correlated with strong Shine-Dalgarno signal sequences. Specific regulatory proteins, e.g., two-component sensor proteins, are rarely PHX. Genes involved in pathways for the synthesis of vitamins record low predicted expression levels. Several distinctive PHX genes of the available complete prokaryotic genomes are highlighted. Relationships of PHX genes with stoichiometry, multifunctionality, and operon structures are discussed. Our methodology may be used complementary to experimental expression analysis.

/pdf/predicted-highly-expressed-genes-of-diverse-prokaryotic-37fs7kukvc.pdf

Jan Mrázek

Papers

Compositional biases of bacterial genomes and evolutionary implications.

What Drives Codon Choices in Human Genes

Strand compositional asymmetry in bacterial and large viral genomes

Amino acid runs in eukaryotic proteomes and disease associations

Predicted Highly Expressed Genes of Diverse Prokaryotic Genomes