The Genetic Code—More Than Just a Table

Co-optimization of therapeutic antibody affinity and specificity using machine learning models that generalize to novel mutational space

Therapeutic antibody development requires selection and engineering of molecules with high affinity and other drug-like biophysical properties. Co-optimization of multiple antibody properties remains a difficult and time-consuming process that impedes drug development. Here we evaluate the use of machine learning to simplify antibody co-optimization for a clinical-stage antibody (emibetuzumab) that displays high levels of both on-target (antigen) and off-target (non-specific) binding. We mutate sites in the antibody complementarity-determining regions, sort the antibody libraries for high and low levels of affinity and non-specific binding, and deep sequence the enriched libraries. Interestingly, machine learning models trained on datasets with binary labels enable predictions of continuous metrics that are strongly correlated with antibody affinity and non-specific binding. These models illustrate strong tradeoffs between these two properties, as increases in affinity along the co-optimal (Pareto) frontier require progressive reductions in specificity. Notably, models trained with deep learning features enable prediction of novel antibody mutations that co-optimize affinity and specificity beyond what is possible for the original antibody library. These findings demonstrate the power of machine learning models to greatly expand the exploration of novel antibody sequence space and accelerate the development of highly potent, drug-like antibodies.

Full-Length Transcriptome: A Reliable Alternative for Single-Cell RNA-Seq Analysis in the Spleen of Teleost Without Reference Genome

Fish is considered as a supreme model for clarifying the evolution and regulatory mechanism of vertebrate immunity. However, the knowledge of distinct immune cell populations in fish is still limited, and further development of techniques advancing the identification of fish immune cell populations and their functions are required. Single cell RNA-seq (scRNA-seq) has provided a new approach for effective in-depth identification and characterization of cell subpopulations. Current approaches for scRNA-seq data analysis usually rely on comparison with a reference genome and hence are not suited for samples without any reference genome, which is currently very common in fish research. Here, we present an alternative, i.e. scRNA-seq data analysis with a full-length transcriptome as a reference, and evaluate this approach on samples from Epinephelus coioides-a teleost without any published genome. We show that it reconstructs well most of the present transcripts in the scRNA-seq data achieving a sensitivity equivalent to approaches relying on genome alignments of related species. Based on cell heterogeneity and known markers, we characterized four cell types: T cells, B cells, monocytes/macrophages (Mo/MΦ) and NCC (non-specific cytotoxic cells). Further analysis indicated the presence of two subsets of Mo/MΦ including M1 and M2 type, as well as four subsets in B cells, i.e. mature B cells, immature B cells, pre B cells and early-pre B cells. Our research will provide new clues for understanding biological characteristics, development and function of immune cell populations of teleost. Furthermore, our approach provides a reliable alternative for scRNA-seq data analysis in teleost for which no reference genome is currently available.

The novel Ideal Symmetry Genetic Code table - Common purine-pyrimidine symmetry net for all RNA and DNA species

Ever since Nirenberg's discovery in 1961 in which codons code individual amino acids, numerous scientists searched for symmetries within the genetic code. The standard genetic code (SGC) table is an alphabetic artificial construct based on the U-C-A-G ordering of nucleotides without natural symmetries. Up to the present, complete symmetry in the genetic code has not been found, leaving doubt as to whether the symmetrical nature as the protector of order even exists. Our novel Ideal Symmetry Genetic Code (ISyGC) table reflects a unique fundamental physicochemical purine-pyrimidine symmetry net for all more than thirty known variations of nuclear and mitochondrial genetic codes. The nuclear genetic code for RNA and DNA viruses also contains the same purine-pyrimidine symmetry net. We show that the ISyGC table leads to automatic transformation into a DNA sequence akin to the 5'3 codon and 3'5 anticodon patterns. As a result of purine-pyrimidine symmetries between codons in the ISyGC table, algorithms of the first two bases as well of the third base of codons show how tRNA cognate anticodons can recognize synonymous codons during mRNA decoding. We show that the ISyGC purine-pyrimidine net with its physicochemical properties represents an evolutionary common "frozen accident" at the onset of each genetic code creation and RNA to DNA evolution. As such, during all of evolution the unique fundamental purine-pyrimidine symmetry net of all genetic codes remains unchangeable. In this way, evolution is a road paved with symmetries.

A model of k-mer surprisal to quantify local sequence information content surrounding splice regions

Molecular sequences carry information. Analysis of sequence conservation between homologous loci is a proven approach with which to explore the information content of molecular sequences. This is often done using multiple sequence alignments to support comparisons between homologous loci. These methods therefore rely on sufficient underlying sequence similarity with which to construct a representative alignment. Here we describe a method using a formal metric of information, surprisal, to analyse biological sub-sequences without alignment constraints. We applied our model to the genomes of five different species to reveal similar patterns across a panel of eukaryotes. As the surprisal of a sub-sequence is inversely proportional to its occurrence within the genome, the optimal size of the sub-sequences was selected for each species under consideration. With the model optimized, we found a strong correlation between surprisal and CG dinucleotide usage. The utility of our model was tested by examining the sequences of genes known to undergo splicing. We demonstrate that our model can identify biological features of interest such as known donor and acceptor sites. Analysis across all annotated coding exon junctions in Homo sapiens reveals the information content of coding exons to be greater than the surrounding intron regions, a consequence of increased suppression of the CG dinucleotide in intronic space. Sequences within coding regions proximal to exon junctions exhibited novel patterns within DNA and coding mRNA that are not a function of the encoded amino acid sequence. Our findings are consistent with the presence of secondary information encoding features such as DNA and RNA binding sites, multiplexed through the coding sequence and independent of the information required to define the corresponding amino-acid sequence. We conclude that surprisal provides a complementary methodology with which to locate regions of interest in the genome, particularly in situations that lack an appropriate multiple sequence alignment.

Global importance of RNA secondary structures in protein-coding sequences

Motivation

The protein-coding sequences of messenger RNAs are the linear template for translation of the gene sequence into protein. Nevertheless, the RNA can also form secondary structures by intramolecular base-pairing.

Results

We show that the nucleotide distribution within codons is biased in all taxa of life on a global scale. Thereby, RNA secondary structures that require base-pairing between the position 1 of a codon with the position 1 of an opposing codon (here named RNA secondary structure class c₁) are under-represented. We conclude that this bias may result from the co-evolution of codon sequence and mRNA secondary structure, suggesting that RNA secondary structures are generally important in protein-coding regions of mRNAs. The above result also implies that codon position 2 has a smaller influence on the amino acid choice than codon position 1.

Supplementary information

Supplementary data are available at Bioinformatics online.

A binary representation of the genetic code

This article introduces a novel binary representation of the canonical genetic code based on both the structural similarities of the nucleotides, as well as the physicochemical properties of the encoded amino acids. Each of the four mRNA bases is assigned a unique 2-bit identifier, so that the 64 triplet codons are each indexed by a 6-bit label. The ordering of the bits reflects the hierarchical organization manifested by the DNA replication/repair and tRNA translation systems. In this system, transition and transversion mutations are naturally expressed as binary operations, and the severities of the different point mutations can be analyzed. Using a principal component analysis, it is shown that the physicochemical properties of amino acids related to protein folding also correlate with certain bit positions of their respective labels. Thus, the likelihood for a point mutation to be conservative, and less likely to cause a change in protein functionality, can be estimated.

TRNA mutations that affect decoding fidelity deregulate development and the proteostasis network in zebrafish

Mutations in genes that encode tRNAs, aminoacyl-tRNA syntheases, tRNA modifying enzymes and other tRNA interacting partners are associated with neuropathies, cancer, type-II diabetes and hearing loss, but how these mutations cause disease is unclear. We have hypothesized that levels of tRNA decoding error (mistranslation) that do not fully impair embryonic development can accelerate cell degeneration through proteome instability and saturation of the proteostasis network. To test this hypothesis we have induced mistranslation in zebrafish embryos using mutant tRNAs that misincorporate Serine (Ser) at various non-cognate codon sites. Embryo viability was affected and malformations were observed, but a significant proportion of embryos survived by activating the unfolded protein response (UPR), the ubiquitin proteasome pathway (UPP) and downregulating protein biosynthesis. Accumulation of reactive oxygen species (ROS), mitochondrial and nuclear DNA damage and disruption of the mitochondrial network, were also observed, suggesting that mistranslation had a strong negative impact on protein synthesis rate, ER and mitochondrial homeostasis. We postulate that mistranslation promotes gradual cellular degeneration and disease through protein aggregation, mitochondrial dysfunction and genome instability.

Neuropathy- and myopathy-associated mutations in human small heat shock proteins: Characteristics and evolutionary history of the mutation sites

Mutations in four of the ten human small heat shock proteins (sHSP) are associated with various forms of motor neuropathies and myopathies. In HspB1, HspB3, and HspB8 all known mutations cause motor neuropathies, whereas in HspB5 they cause myopathies. Several features are common to the majority of these mutations: (i) they are missense mutations, (ii) most associated disease phenotypes exhibit a dominant inheritance pattern and late disease onset, (iii) in the primary protein sequences, the sites of most mutations are located in the conserved α-crystallin domain and the variable C-terminal extensions, and (iv) most human mutation sites are highly conserved among the vertebrate orthologs and have been historically exposed to significant purifying selection. In contrast, a minor fraction of these mutations deviate from these rules: they are (i) frame shifting, nonsense, or elongation mutations, (ii) associated with recessive or early onset disease phenotypes, (iii) positioned in the N-terminal domain of the proteins, and (iv) less conserved among the vertebrates and were historically not subject to a strong selective pressure. In several vertebrate sHSPs (including primate sHSPs), homologous sites differ from the human sequence and occasionally even encode the same amino acid residues that cause the disease in humans. Apparently, a number of these mutations sites are not crucial for the protein function in single species or entire taxa, and single species even seem to have adopted mechanisms that compensate for potentially adverse effects of 'mutant-like' sHSPs. The disease-associated dominant sHSP missense mutations have a number of cellular consequences that are consistent with gain-of-function mechanisms of genetic dominance: dominant-negative effects, the formation of cytotoxic amyloid protein oligomers and precipitates, disruption of cytoskeletal networks, and increased downstream enzymatic activities. Future therapeutic concepts should aim for reducing these adverse effects of mutant sHSPs in patients. Indeed, initial experimental results are encouraging.

Genetic variants of the NOTCH3 gene in the elderly and magnetic resonance imaging correlates of age-related cerebral small vessel disease

Cerebral small vessel disease-related brain lesions such as white matter lesions and lacunes are common findings of magnetic resonance imaging in the elderly. These lesions are thought to be major contributors to disability in old age, and risk factors that include age and hypertension have been established. The radiological, histopathologic and clinical phenotypes of age-related cerebral small vessel disease remarkably resemble autosomal dominant arteriopathy with subcortical infarcts and leucoencephalopathy, which is caused by mutations in NOTCH3. We hypothesized that genetic variations in NOTCH3 also play a role in age-related cerebral small vessel disease. We directly sequenced all 33 exons, the promoter and 3'-untranslated region of NOTCH3 in 195 participants with either coalescent white matter lesions or lacunes and compared the results to 82 randomly selected participants with no focal changes on magnetic resonance images in the Austrian Stroke Prevention Study. We detected nine common and 33 rare single nucleotide polymorphisms, of which 20 were novel. All common single nucleotide polymorphisms were genotyped in the entire cohort (n = 888), and four of them, rs1043994, rs10404382, rs10423702 and rs1043997, were associated significantly with both the presence and progression of white matter lesions. The association was confined to hypertensives, a result which we replicated in the Cohorts for Heart and Ageing Research in Genomic Epidemiology Consortium on an independent sample of 4773 stroke-free hypertensive elderly individuals of European descent (P = 0.04). The 33 rare single nucleotide polymorphisms were scattered over the NOTCH3 gene with three being located in the promoter region, 24 in exons (18 non-synonymous), three in introns and three in the 3'-untranslated region. None of the single nucleotide polymorphisms affected a cysteine residue. Sorting Intolerant From Tolerant, PolyPhen2 analyses and protein structure simulation consistently predicted six of the non-synonymous single nucleotide polymorphisms (H170R, P496L, V1183M, L1518M, D1823N and V1952M) to be functional, with four being exclusively or mainly detected in subjects with severe white matter lesions. In four individuals with rare non-synonymous single nucleotide polymorphisms, we noted anterior temporal lobe hyperintensity, hyperintensity in the external capsule, lacunar infarcts or subcortical lacunar lesions. None of the observed abnormalities were specific to cerebral autosomal dominant arteriopathy with subcortical infarcts and leucoencephalopathy. This is the first comprehensive study investigating (i) the frequency of NOTCH3 variations in community-dwelling elderly and (ii) their effect on cerebral small vessel disease related magnetic resonance imaging phenotypes. We show that the NOTCH3 gene is highly variable with both common and rare single nucleotide polymorphisms spreading across the gene, and that common variants at the NOTCH3 gene increase the risk of age-related white matter lesions in hypertensives. Additional investigations are required to explore the biological mechanisms underlying the observed association.

The rules of variation: amino acid exchange according to the rotating circular genetic code

General guidelines for the molecular basis of functional variation are presented while focused on the rotating circular genetic code and allowable exchanges that make it resistant to genetic diseases under normal conditions. The rules of variation, bioinformatics aids for preventative medicine, are: (1) same position in the four quadrants for hydrophobic codons, (2) same or contiguous position in two quadrants for synonymous or related codons, and (3) same quadrant for equivalent codons. To preserve protein function, amino acid exchange according to the first rule takes into account the positional homology of essential hydrophobic amino acids with every codon with a central uracil in the four quadrants, the second rule includes codons for identical, acidic, or their amidic amino acids present in two quadrants, and the third rule, the smaller, aromatic, stop codons, and basic amino acids, each in proximity within a 90 degree angle. I also define codifying genes and palindromati, CTCGTGCCGAATTCGGCACGAG.