AI is a viable alternative to high throughput screening: a 318-target study

High throughput screening (HTS) is routinely used to identify bioactive small molecules. This requires physical compounds, which limits coverage of accessible chemical space. Computational approaches combined with vast on-demand chemical libraries can access far greater chemical space, provided that the predictive accuracy is sufficient to identify useful molecules. Through the largest and most diverse virtual HTS campaign reported to date, comprising 318 individual projects, we demonstrate that our AtomNet® convolutional neural network successfully finds novel hits across every major therapeutic area and protein class. We address historical limitations of computational screening by demonstrating success for target proteins without known binders, high-quality X-ray crystal structures, or manual cherry-picking of compounds. We show that the molecules selected by the AtomNet® model are novel drug-like scaffolds rather than minor modifications to known bioactive compounds. Our empirical results suggest that computational methods can substantially replace HTS as the first step of small-molecule drug discovery.

Investigation of dynamic solution interactions between NET-1 and UNC-5B by multi-wavelength analytical ultracentrifugation

NET-1 is a key chemotropic ligand that signals commissural axon migration and change in direction. NET-1 and its receptor UNC-5B switch axon growth cones from attraction to repulsion. The biophysical properties of the NET-1 + UNC-5B complex have been poorly characterized. Using multi-wavelength-AUC by adding a fluorophore to UNC-5B, we were able to separate the UNC-5B sedimentation from NET-1. Using both multi-wavelength- and single-wavelength AUC, we investigated NET-1 and UNC-5B hydrodynamic parameters and complex formation. The sedimentation velocity experiments show that NET-1 exists in a monomer-dimer equilibrium. A close study of the association shows that NET-1 forms a pH-sensitive dimer that interacts in an anti-parallel orientation. UNC-5B can form equimolar NET-1 + UNC-5B heterocomplexes with both monomeric and dimeric NET-1.

Molecular dynamics simulations of ortho-carborane nano-diamond storage within the nonpolar channel cavities of a right-handed coiled-coil tetrabrachion nanotube

Molecular dynamics simulations have been performed on a complex in which clusters of boron in the form of molecules of the nanodiamond ortho-carborane ( C 2 B 10 H 12 ) have been inserted into the four large nonpolar cavities of a nanotube of the right-handed coiled-coil ( R H C C ) t e t r a b r a c h i o n . The techniques of multi-configurational thermodynamic integration, steered molecular dynamics and umbrella sampling have been combined to investigate the energetics of storage of ortho-carborane in the cavities and to map out the free energy landscape of the RHCC - t e t r a b r a c h i o n - o r t h o - c a r b o r a n e complex along the central channel and along directions transverse to the central channel. The purpose of the study was to explore potential pathways for the diffusion of ortho-carborane between the cavities and the solvent and to assess the stability of the complex as a possible drug delivery system for boron neutron capture therapy (BNCT). The investigation reveals a complex free energy landscape with a multitude of peaks and valleys, all of which can be related to specific architectural elements of the RHCC - n a n o t u b e , and the activation barriers for ortho-carborane capture and release support the requirements for rapid cargo uptake coupled with tight binding to the cavities.

Energy flow and intersubunit signalling in GSAM: A non-equilibrium molecular dynamics study

Non-equilibrium molecular dynamics simulations of vibrational energy flow induced by the imposition of a thermal gradient have been performed on the μ2-dimeric enzyme glutamate-1-semialdehyde aminomutase (GSAM), the key enzyme in the biosynthesis of chlorophyll, in order to identify energy transport pathways and to elucidate their role as potential allosteric communication networks for coordinating functional dynamics, specifically the negative cooperativity observed in the motion of the two active site gating loops. Fully atomistic MD simulations of thermal diffusion were executed with a GROMACS simulation package on a fully solvated GSAM enzyme by heating various active site target ligands (initially, catalytic intermediates and cofactors) to 300K while holding the remainder of the protein and the solvent bath at 10K and monitoring the temperature T(t) of all the enzyme residues as a function of time over a 1ns observation window. Energy is observed to be deposited in a relatively small number of discrete chains of residues most of which contribute to specific structural or biochemical functionality. Thermal linkages between all thermally active chains were established by isolating a specific pair of chains and performing a thermal diffusion simulation on the pair, one held at 300K and the other at 10K, with the rest of the protein frozen in its initial atomic configuration and thus thermally unresponsive. Proceeding in this way, it was possible to map out multiple pathways of vibrational energy flow leading from one of the active sites through a network of contiguous residues, many of which were evolutionarily conserved and linked by hydrogen bonds, into the other active site and ultimately to the other gating loop.

Energetics of Storage and Diffusion of Water and Cyclo-Octasulfur for a Nonpolar Cavity of RHCC Tetrabrachion by Molecular Dynamics Simulations

Tetrabrachion forms the key component of the S-layer of Staphylothermus marinus. Molecular dynamics simulations have been used to study the energetics of occupancy of cavity 3 of the right-handed coiled-coil stalk of tetrabrachion by both water molecules and cyclooctasulfur S₈ crowns, as well as to determine possible pathways and free energy barriers for the diffusion of both water and cyclooctasulfur through the peptide walls of RHCC tetrabrachion between cavity 3 and bulk solvent. Calculations of the transfer free energy from solvent to cavity show that clusters of six, seven and eight water molecules are marginally stable in cavity 3, but that occupancy of the cavity by a cyclooctasulfur ring is favoured significantly over water clusters of all sizes. Thermal activation simulations at T = 400K revealed that water molecules diffusing through the wall pass through a sequence of metastable configurations where they are temporarily immobilized by forming networks of hydrogen bonds with specific wall residues. Calculations of the free energy of these metastable configurations using multi-configurational thermodynamic integration yielded a free energy profile with a principal free energy maximum ∆G~50 kJ/mol and a slight activation asymmetry in favour of the direction from cavity to solvent. Potential exit pathways for cyclooctasulfur were investigated with the methods of steered molecular dynamics and umbrella sampling. The cyclooctasulfur was steered through a gap in the tetrabrachion wall along a linear path from cavity 3 into the solvent and the resulting trajectory was subdivided into 22 sampling windows. The free energy profile created for the trajectory by umbrella sampling showed a sharp principal maximum as a function of the reaction coordinate with asymmetric free energy barriers ∆G_exit~220 kJ/mol and ∆G_entrance~100 kJ/mol for cavity exit and entrance, respectively.

Solution Structure of C. elegans UNC-6: A Nematode Paralogue of the Axon Guidance Protein Netrin-1

UNCoordinated-6 (UNC-6) was the first member of the netrin family to be discovered in Caenorhabditis elegans. With homology to human netrin-1, it is a key signaling molecule involved in directing axon migration in nematodes. Similar to netrin-1, UNC-6 interacts with multiple receptors (UNC-5 and UNC-40, specifically) to guide axon migration in development. As a result of the distinct evolutionary path of UNC-6 compared to vertebrate netrins, we decided to employ an integrated approach to study its solution behavior and compare it to the high-resolution structure we previously published on vertebrate netrins. Dynamic light scattering and analytical ultracentrifugation on UNC-6 (with and without its C-domain) solubilized in a low-ionic strength buffer suggested that UNC-6 forms high-order oligomers. An increase in the buffer ionic strength resulted in a more homogeneous preparation of UNC-6, that was used for subsequent solution x-ray scattering experiments. Our biophysical analysis of UNC-6 ΔC solubilized in a high-ionic strength buffer suggested that it maintains a similar head-to-stalk arrangement as netrins -1 and -4. This phenomenon is thought to play a role in the signaling behavior of UNC-6 and its ability to move throughout the extracellular matrix.

Association of a Novel ACTA1 Mutation With a Dominant Progressive Scapuloperoneal Myopathy in an Extended Family

IMPORTANCE

New genomic strategies can now be applied to identify a diagnosis in patients and families with previously undiagnosed rare genetic conditions. The large family evaluated in the present study was described in 1966 and now expands the phenotype of a known neuromuscular gene.

OBJECTIVE

To determine the genetic cause of a slowly progressive, autosomal dominant, scapuloperoneal neuromuscular disorder by using linkage and exome sequencing.

DESIGN, SETTING, AND PARTICIPANTS

Fourteen affected individuals in a 6-generation family with a progressive scapuloperoneal disorder were evaluated. Participants were examined at pediatric, neuromuscular, and research clinics from March 1, 2005, to May 31, 2014. Exome and linkage were performed in genetics laboratories of research institutions.

MAIN OUTCOMES AND MEASURES

Examination and evaluation by magnetic resonance imaging, ultrasonography, electrodiagnostic studies, and muscle biopsies (n = 3). Genetic analysis included linkage analysis (n = 17) with exome sequencing (n = 7).

RESULTS

Clinical findings included progressive muscle weakness in an initially scapuloperoneal and distal distribution, including wrist extensor weakness, finger and foot drop, scapular winging, mild facial weakness, Achilles tendon contractures, and diminished or absent deep tendon reflexes. Both age at onset and progression of the disease showed clinical variability within the family. Muscle biopsy specimens demonstrated type I fiber atrophy and trabeculated fibers without nemaline rods. Analysis of exome sequences within the linkage region (4.8 megabases) revealed missense mutation c.591C>A p.Glu197Asp in a highly conserved residue in exon 4 of ACTA1. The mutation cosegregated with disease in all tested individuals and was not present in unaffected individuals.

CONCLUSIONS AND RELEVANCE

This family defines a new scapuloperoneal phenotype associated with an ACTA1 mutation. A highly conserved protein, ACTA1 is implicated in multiple muscle diseases, including nemaline myopathy, actin aggregate myopathy, fiber-type disproportion, and rod-core myopathy. To our knowledge, mutations in Glu197 have not been reported previously. This residue is highly conserved and located in an exposed position in the protein; the mutation affects the intermolecular and intramolecular electrostatic interactions as shown by structural modeling. The mutation in this residue does not appear to lead to rod formation or actin accumulation in vitro or in vivo, suggesting a different molecular mechanism from that of other ACTA1 diseases.

Human-gyrovirus-Apoptin triggers mitochondrial death pathway--Nur77 is required for apoptosis triggering

The human gyrovirus derived protein Apoptin (HGV-Apoptin) a homologue of the chicken anemia virus Apoptin (CAV-Apoptin), a protein with high cancer cells selective toxicity, triggers apoptosis selectively in cancer cells. In this paper, we show that HGV-Apoptin acts independently from the death receptor pathway as it induces apoptosis in similar rates in Jurkat cells deficient in either FADD (fas-associated death domain) function or caspase-8 (key players of the extrinsic pathway) and their parental clones. HGV-Apoptin induces apoptosis via the activation of the mitochondrial intrinsic pathway. It induces both mitochondrial inner and outer membrane permebilization, characterized by the loss of the mitochondrial potential and the release into cytoplasm of the pro-apoptotic molecules including apoptosis inducing factor and cytochrome c. HGV-Apoptin acts via the apoptosome, as lack of expression of apoptotic protease-activating factor 1 in murine embryonic fibroblast strongly protected the cells from HGV-Apoptin–induced apoptosis. Moreover, QVD-oph a broad-spectrum caspase inhibitor delayed HGV-Apoptin–induced death. On the other hand, overexpression of the anti-apoptotic BCL-XL confers resistance to HGV-Apoptin–induced cell death. In contrast, cells that lack the expression of the pro-apoptotic BAX and BAK are protected from HGV-Apoptin induced apoptosis. Furthermore, HGV-Apoptin acts independently from p53 signal but triggers the cytoplasmic translocation of Nur77. Taking together these data indicate that HGV-Apoptin acts through the mitochondrial pathway, in a caspase-dependent manner but independently from the death receptor pathway.

Apoptins: selective anticancer agents

Therapies that selectively target cancer cells for death have been the center of intense research recently. One potential therapy may involve apoptin proteins, which are able to induce apoptosis in cancer cells leaving normal cells unharmed. Apoptin was originally discovered in the Chicken anemia virus (CAV); however, human gyroviruses (HGyV) have recently been found that also harbor apoptin-like proteins. Although the cancer cell specific activity of these apoptins appears to be well conserved, the precise functions and mechanisms of action are yet to be fully elucidated. Strategies for both delivering apoptin to treat tumors and disseminating the protein inside the tumor body are now being developed, and have shown promise in preclinical animal studies.

The RNA helicase RHAU (DHX36) unwinds a G4-quadruplex in human telomerase RNA and promotes the formation of the P1 helix template boundary

Human telomerase RNA (hTR) contains several guanine tracts at its 5′-end that can form a G4-quadruplex structure. Previous evidence suggests that a G4-quadruplex within this region disrupts the formation of an important structure within hTR known as the P1 helix, a critical element in defining the template boundary for reverse transcription. RNA associated with AU-rich element (RHAU) is an RNA helicase that has specificity for DNA and RNA G4-quadruplexes. Two recent studies identify a specific interaction between hTR and RHAU. Herein, we confirm this interaction and identify the minimally interacting RNA fragments. We demonstrate the existence of multiple quadruplex structures within the 5′ region of hTR and find that these regions parallel the minimal sequences capable of RHAU interaction. We confirm the importance of the RHAU-specific motif in the interaction with hTR and demonstrate that the helicase activity of RHAU is sufficient to unwind the quadruplex and promote an interaction with 25 internal nucleotides to form a stable P1 helix. Furthermore, we have found that a 5′-terminal quadruplex persists following P1 helix formation that retains affinity for RHAU. Finally, we have investigated the functional implications of this interaction and demonstrated a reduction in average telomere length following RHAU knockdown by small interfering RNA (siRNA).

Intersubunit signaling in glutamate-1-semialdehyde-aminomutase

Enzymes are highly dynamic and tightly controlled systems. However, allosteric communication linked to catalytic turnover is poorly understood. We have performed an integrated approach to trap several catalytic intermediates in the alpha2-dimeric key enzyme of chlorophyll biosynthesis, glutamate-1-semialdehyde aminomutase. Our data reveal an active-site "gating loop," which undergoes a dramatic conformational change during catalysis, that is simultaneously open in one subunit and closed in the other. This loop movement requires a beta-sheet-to-alpha-helix transition to assume the closed conformation, thus facilitating transport of substrate toward, and concomitantly forming, an integral part of the active site. The accompanying intersubunit cross-talk, which controls negative cooperativity between the allosteric pair, was explored at the atomic level. The central elements of the communication triad are the cofactor bound to different catalytic intermediates, the interface helix, and the gating loop. Together, they form a molecular switch in which the cofactor acts as a central signal transmitter linking the subunit interface with the gating loop.

The laminin-binding domain of agrin is structurally related to N-TIMP-1

Agrin is the key organizer of postsynaptic differentiation at the neuromuscular junction. This organization activity requires the binding of agrin to the synaptic basal lamina. Binding is conferred by the N-terminal agrin (NtA) domain, which mediates a high-affinity interaction with the coiled coil domain of laminins. Here, we report the crystal structure of chicken NtA at 1.6 A resolution. The structure reveals that NtA harbors an oligosaccharide/oligonucleotide-binding fold with several possible sites for the interaction with different ligands. A high structural similarity of NtA with the protease inhibition domain in tissue inhibitor of metalloproteinases-1 (TIMP-1) supports the idea of additional functions of agrin besides synaptogenic activity.

Coiled coils: a highly versatile protein folding motif

The alpha-helical coiled coil is one of the principal subunit oligomerization motifs in proteins. Its most characteristic feature is a heptad repeat pattern of primarily apolar residues that constitute the oligomer interface. Despite its simplicity, it is a highly versatile folding motif: coiled-coil-containing proteins exhibit a broad range of different functions related to the specific 'design' of their coiled-coil domains. The architecture of a particular coiled-coil domain determines its oligomerization state, rigidity and ability to function as a molecular recognition system. Much progress has been made towards understanding the factors that determine coiled-coil formation and stability. Here we discuss this highly versatile protein folding and oligomerization motif with regard to its structural architecture and how this is related to its biological functions.

Crystal structure of a naturally occurring parallel right-handed coiled coil tetramer

The crystal structure of a polypeptide chain fragment from the surface layer protein tetrabrachion from Staphylothermus marinus has been determined at 1.8 A resolution. As proposed on the basis of the presence of 11-residue repeats, the polypeptide chain fragment forms a parallel right-handed coiled coil structure. Complementary hydrophobic interactions and complex networks of surface salt bridges result in an extremely thermostable tetrameric structure with remarkable properties. In marked contrast to left-handed coiled coil tetramers, the right-handed coiled coil reveals large hydrophobic cavities that are filled with water molecules. As a consequence, the packing of the hydrophobic core differs markedly from that of a right-handed parallel coiled coil tetramer that was designed on the basis of left-handed coiled coil structures.

Toward a high-resolution structure of phospholamban: design of soluble transmembrane domain mutants

Determination of a high-resolution structure of the phospholamban (PLB) transmembrane domain by X-ray crystallography or NMR is handicapped by the hydrophobic nature of the peptide. Interestingly, the crystal structure of the five-stranded parallel coiled-coil oligomerization domain from cartilage oligomeric matrix protein (COMPcc) shows marked similarities to a model proposed for the pentameric transmembrane domain of PLB. Contrary to the putative coiled-coil domain of PLB, COMPcc contains mostly hydrophilic amino acids on the surface, resulting in a soluble molecule. Here, we report the design of soluble PLB transmembrane domain variants by combining the surface residues of COMPcc and the hydrophobic interior of the transmembrane domain of PLB. The soluble PLB variants formed pentameric structures as revealed by analytical ultracentrifugation. After redox shuffling, they showed unspecific disulfide bridge patterns similar to that of the chemically synthesized wild-type PLB transmembrane domain. These results suggest a structural homology between the soluble PLB mutants and the wild-type PLB transmembrane domain. Together with the data reported in the literature, they furthermore indicate that residues Leu37, Ile40, Leu44, and Ile47 of the PLB sequence specify pentamer formation. In contrast, a designed recombinant COMPcc mutant, COMP-ARCC, which was engineered to contain the two PLB cysteines that potentially could form an interchain disulfide bridge, formed a specific disulfide bond pattern. This finding indicates structural differences between the transmembrane domain of PLB and COMPcc. The soluble PLB variants may be used to determine a high-resolution structure of the PLB pentamer by X-ray crystallography.

Site-directed mutagenesis and structural interpretation of the nidogen binding site of the laminin gamma1 chain

A precise molecular map of the nidogen binding site of laminins was obtained by site-directed mutagenesis and structural analysis of the 56 residue LE module gamma1III4 of their gamma1 chain. This demonstrated the crucial importance of the sequence DPNAV (position 800-804) in the disulfide-bonded loop a, with major contributions made by all residues except P801. Different substitutions of these residues emphasized the essential role of the negative charge (D800) and carboxamide group (N802) as well as their spacings and hydrophobic contacts (V804) for interaction, and predict direct contacts of these three residues with a complementary binding region of nidogen. An inactivating A803-V substitution, however, may lead to a distorted loop structure. A lower but still significant contribution originates from the non-contiguous link/loop c sequence LKCIY (positions 815-819) which is spatially close to the loop a sequence. The link residues (L815 and K816) provide main chain hydrogen bonds to N806 and a side chain hydrogen bond to the V804 carbonyl and thus stabilize the conformation of loop a. The side chains of I818 and Y819 together with P842 from loop d form hydrophobic contacts that provide further stability but could possibly also participate in direct ligation. The nidogen binding epitope is therefore localized on a narrow ridge and has a length of approximately 17 angstroms. The data also indicate a strong conservation of the epitope in the laminin gamma1 chains of several invertebrates.

Crystal structure of three consecutive laminin-type epidermal growth factor-like (LE) modules of laminin gamma1 chain harboring the nidogen binding site

The structure of three consecutive laminin-type EGF-like (LE) modules of mouse laminin gammma1 chain, gamma1III3-5 (positions 738 to 899), has been determined by multiple isomorphous replacement in a crystal of space group p6(4)22 (a=b=74.57 angstroms, c = 185.11 angstroms and gamma = 120 degrees). The crystal structure was refined using restrained crystallographic refinement to an R-factor of 19.72% for 14,983 independent reflections with intensities F(obs)> 0 at 2.1 angstroms resolution, with root mean square deviation of 0.012 angstroms and 1.690 degrees from ideal bond lengths and bond angles, respectively. The final model consisted of 1179 (non-hydrogen) protein atoms within 162 residues and 119 water molecules. The molecule showed a rod-like structure of about 76 angstroms length with individual modules twisted relative to each other by about 70 degrees. Each module has the same disulfide bond connections Cys1-Cys3 (loop a), Cys2-Cys4 (loop b), Cys5-Cys6 (loop c) and Cys7-Cys8 (loop d), the first three being identical to epidermal growth factor (EGF). All three LE modules showed little secondary structure which was mainly restricted to loop d, but they differed in several other details of their structure. The interface contacts between the LE modules are based on hydrogen bonds and hydrophobic interactions between the hydrophobic core of loop d of the preceding module and the first cysteine and an exposed residue in loop b of the following module. Module 4 was previously shown to contribute the major nidogen binding site of laminis and site-directed mutagenesis demonstrated a specific binding role for Asp800, Asn802, Val804 and Tyr819 in loops a and c. The side-chain of these four residues are all located on the surface in a linear array and separated by a distance of 17 angstroms between Tyr819 and Val804. The entire nidogen binding site is stabilized via main-chain hydrogen bonds which are in part derived from the link between loops b and c (residues Leu815 and Lys816). The data demonstrate the unique nature of the LE modules and only a remote similarity to EGF. They also indicate that the crucial residues in the binding loops provide direct contacts with nidogen and explain the synergism between loops a and c which is essential for binding.