Structure-based protein design with deep learning

Since the first revelation of proteins functioning as macromolecular machines through their three dimensional structures, researchers have been intrigued by the marvelous ways the biochemical processes are carried out by proteins. The aspiration to understand protein structures has fueled extensive efforts across different scientific disciplines. In recent years, it has been demonstrated that proteins with new functionality or shapes can be designed via structure-based modeling methods, and the design strategies have combined all available information — but largely piece-by-piece — from sequence derived statistics to the detailed atomic-level modeling of chemical interactions. Despite the significant progress, incorporating data-derived approaches through the use of deep learning methods can be a game changer. In this review, we summarize current progress, compare the arc of developing the deep learning approaches with the conventional methods, and describe the motivation and concepts behind current strategies that may lead to potential future opportunities.     

DNA Sequences from Formalin-Fixed Nematodes: Integrating Molecular and Morphological Approaches to Taxonomy

To effectively integrate DNA sequence analysis and classical nematode taxonomy, we must be able to obtain DNA sequences from formalin-fixed specimens. Microdissected sections of nematodes were removed from specimens fixed in formalin, using standard protocols and without destroying morphological features. The fixed sections provided sufficient template for multiple polymerase chain reaction-based DNA sequence analyses.     

Trichoderma reesei sequences that bind to the nuclear matrix enhance transformation frequency

Three DNA fragments, trs1, 2 and 3, were isolated from the Trichoderma reesei genome on the basis of their ability to promote autonomous replication of plasmids in Saccharomyces cerevisiae. Each trs element bound specifically to the isolated T. reesei nuclear matrix in vitro, and two of them bound in vivo, indicating that they are matrix attachment regions (MARs). A similar sequence previously isolated from Aspergillus nidulans (ans1) was also shown to bind specifically to the T. reesei nuclear matrix in vitro. The T. reesei MARs are AT-rich sequences containing 70%, 86% and 73% A+T over 2.9, 0.8 and 3.7 kb, respectively for trs1, 2 and 3. They exhibited no significant sequence homology, but were shown to contain a number of sequence motifs that occur frequently in many MARs identified in other eukaryotes. However, these motifs occurred as frequently in the trs elements as in randomly generated sequences with the same A+T content. trs1 and 3 were shown to be present as single copies in the T. reesei genome. The presence of the trs elements in transforming plasmids enhanced the frequency of integrative transformation of T. reesei up to five fold over plasmids without a trs. No evidence was obtained to suggest that the trs elements promoted efficient replication of plasmids in T. reseei. A mechanism for the enhancement of transformation frequency by the trs elements is proposed. Received: 1 March 1997 / Accepted: 13 May 1997     

Microsecond Dynamics During the Binding-induced Folding of an Intrinsically Disordered Protein

Tau is an intrinsically disordered protein implicated in many neurodegenerative diseases. The repeat domain fragment of tau, tau-K18, is known to undergo a disorder to order transition in the presence of lipid micelles and vesicles, in which helices form in each of the repeat domains. Here, the mechanism of helical structure formation, induced by a phospholipid mimetic, sodium dodecyl sulfate (SDS) at sub-micellar concentrations, has been studied using multiple biophysical probes. A study of the conformational dynamics of the disordered state, using photoinduced electron transfer coupled to fluorescence correlation spectroscopy (PET-FCS) has indicated the presence of an intermediate state, I, in equilibrium with the unfolded state, U. The cooperative binding of the ligand (L), SDS, to I has been shown to induce the formation of a compact, helical intermediate (IL₅) within the dead time (∼37 µs) of a continuous flow mixer. Quantitative analysis of the PET-FCS data and the ensemble microsecond kinetic data, suggests that the mechanism of induction of helical structure can be described by a U ↔ I ↔ IL₅ ↔ FL₅ mechanism, in which the final helical state, FL₅, forms from IL₅ with a time constant of 50–200 µs. Finally, it has been shown that the helical conformation is an aggregation-competent state that can directly form amyloid fibrils.     

An Efficient Regression Type Estimator in Survey Sampling

For the estimation of population mean in simple random sampling, an efficient regression-type estimator is proposed which is more efficient than the conventional regression estimator and hence than mean per unit estimator, ratio and product estimators and many other estimators proposed by various authors. Some numerical examples are included for illustration.     

UDP-N-acetylglucosamine 1-carboxyvinyl-transferase from Acinetobacter calcoaceticus

Abstract We have analyzed the sequence downstream of rpoN from Zcinetobacter calcoaceticus and identified an open reading frame encoding a protein with high similarity to UDP- N -acetylgucosamine 1-carboxyvinyl-transferase (MurZ). Multicopy plasmids encoding this enzyme conferred phosphomycin resistance to A. calcoaceticus . The polar effect of a rpoN mutation on the phosphomycin resistance level suggests that murZ is, in part, cotranscribed with rpoN . These observations confirm that A. calcoaceticus represents the first exceptin from a conserved genetic context of rpoN observed in several other Gram-negative bacteria.     

Nucleotide sequence of the ethidium efflux gene from Escherichia coli

The nucleotide sequence of the gene specifying the ethidium efflux system of Escherichia coli has been determined. The translated open reading frame has identified a membrane-bound polypeptide of 110 amino acids (11,960 Da) which shares 42% identity with a staphylococcal protein specifying resistance to ethidium.     

Identification of the coding region for the putidaredoxin reductase gene from the plasmid of Pseudomonas putida

The first 12 NH₂-terminal amino acids of the Pseudomonas putida putidaredoxin reductase were shown to be Met-Asn-Ala-Asn-Asp-Asn-Val-Val-Ile-Val-Gly-Thr. Comparison of these data with the DNA sequence of the BamHI-HindIII 197-base fragment derived from the PstI 2.2-kb fragment obtained from the P. putida plasmid showed that the putidaredoxin reductase gene was downstream from the cytochrome P-450 gene and the intergenic region had the 24-nucleotide sequence TAAACACATGGGAGTGCGTGCTAA. The Shine-Dalgarno sequence GGAG was detected in this region. The initiating triplet for the reductase gene was GTG, which normally codes for valine, but in the initiating codon position codes for methionine. From the amino acid sequence and X-ray data comparisons with other flavoproteins, what appears to be the AMP binding region of the FAD can be recognized in the NH₂-terminal portion of the reductase involving residues 5–35.This article was presented during the proceedings of the International Conference on Macromolecular Structure and Function, held at the National Defence Medical College, Tokorozawa, Japan, December 1985.