Phosphoantigens glue butyrophilin 3A1 and 2A1 to activate Vγ9Vδ2 T cells.

In both cancer and infections, diseased cells are presented to human Vγ9Vδ2 T cells through an 'inside out' signalling process whereby structurally diverse phosphoantigen (pAg) molecules are sensed by the intracellular domain of butyrophilin BTN3A11-4. Here we show how-in both humans and alpaca-multiple pAgs function as 'molecular glues' to promote heteromeric association between the intracellular domains of BTN3A1 and the structurally similar butyrophilin BTN2A1. X-ray crystallography studies visualized that engagement of BTN3A1 with pAgs forms a composite interface for direct binding to BTN2A1, with various pAg molecules each positioned at the centre of the interface and gluing the butyrophilins with distinct affinities. Our structural insights guided mutagenesis experiments that led to disruption of the intracellular BTN3A1-BTN2A1 association, abolishing pAg-mediated Vγ9Vδ2 T cell activation. Analyses using structure-based molecular-dynamics simulations, 19F-NMR investigations, chimeric receptor engineering and direct measurement of intercellular binding force revealed how pAg-mediated BTN2A1 association drives BTN3A1 intracellular fluctuations outwards in a thermodynamically favourable manner, thereby enabling BTN3A1 to push off from the BTN2A1 ectodomain to initiate T cell receptor-mediated γδ T cell activation. Practically, we harnessed the molecular-glue model for immunotherapeutics design, demonstrating chemical principles for developing both small-molecule activators and inhibitors of human γδ T cell function.

A Structural Change in Butyrophilin upon Phosphoantigen Binding Underlies Phosphoantigen-Mediated Vγ9Vδ2 T Cell Activation.

Human Vγ9Vδ2 T cells respond to microbial infections and malignancy by sensing diphosphate-containing metabolites called phosphoantigens, which bind to the intracellular domain of butyrophilin 3A1, triggering extracellular interactions with the Vγ9Vδ2 T cell receptor (TCR). Here, we examined the molecular basis of this "inside-out" triggering mechanism. Crystal structures of intracellular butyrophilin 3A proteins alone or in complex with the potent microbial phosphoantigen HMBPP or a synthetic analog revealed key features of phosphoantigens and butyrophilins required for γδ T cell activation. Analyses with chemical probes and molecular dynamic simulations demonstrated that dimerized intracellular proteins cooperate in sensing HMBPP to enhance the efficiency of γδ T cell activation. HMBPP binding to butyrophilin doubled the binding force between a γδ T cell and a target cell during "outside" signaling, as measured by single-cell force microscopy. Our findings provide insight into the "inside-out" triggering of Vγ9Vδ2 T cell activation by phosphoantigen-bound butyrophilin, facilitating immunotherapeutic drug design.

Characterization of AST-001 non-clinical pharmacokinetics: A novel selective AKR1C3-activated prodrug in mice, rats, and cynomolgus monkeys.

AST-001 is a chemically synthesized inactive nitrogen mustard prodrug that is selectively cleaved to a cytotoxic aziridine (AST-2660) via aldo-keto reductase family 1 member C3 (AKR1C3). The purpose of this study was to investigate the pharmacokinetics and tissue distribution of the prodrug, AST-001, and its active metabolite, AST-2660, in mice, rats, and monkeys. After single and once daily intravenous bolus doses of 1.5, 4.5, and 13.5 mg/kg AST-001 to Sprague-Dawley rats and once daily 1 h intravenous infusions of 0.5, 1.5, and 4.5 mg/kg AST-001 to cynomolgus monkeys, AST-001 exhibited dose-dependent pharmacokinetics and reached peak plasma levels at the end of the infusion. No significant accumulation and gender differences were observed after 7 days of repeated dosing. In rats, the half-life of AST-001 was dose independent and ranged from 4.89 to 5.75 h. In cynomolgus monkeys, the half-life of AST-001 was from 1.66 to 5.56 h and increased with dose. In tissue distribution studies conducted in Sprague-Dawley rats and in liver cancer PDX models in female athymic nude mice implanted with LI6643 or LI6280 HepG2-GFP tumor fragments, AST-001 was extensively distributed to selected tissues. Following a single intravenous dose, AST-001 was not excreted primarily as the prodrug, AST-001 or the metabolite AST-2660 in the urine, feces, and bile. A comprehensive analysis of the preclinical data and inter-species allometric scaling were used to estimate the pharmacokinetic parameters of AST-001 in humans and led to the recommendation of a starting dose of 5 mg/m2 in the first-in-human dose escalation study.

Improving the Accuracy of Protein Thermostability Predictions for Single Point Mutations.

Accurately predicting the protein thermostability changes upon single point mutations in silico is a challenge that has implications for understanding diseases as well as industrial applications of protein engineering. Free energy perturbation (FEP) has been applied to predict the effect of single point mutations on protein stability for over 40 years and emerged as a potentially reliable prediction method with reasonable throughput. However, applications of FEP in protein stability calculations in industrial settings have been hindered by a number of limitations, including the inability to model mutations to and from prolines in which the bonded topology of the backbone is modified and the complexity in modeling charge-changing mutations. In this study, we have extended the FEP+ protocol to enable the accurate modeling of the effects on protein stability from proline mutations and from charge-changing mutations. We also evaluated the influence of the unfolded model in the stability calculations using increasingly longer peptides with native sequence and conformations. With the abovementioned improvements, the accuracy of FEP predictions of protein stability over a data set of 87 mutations on five different proteins has drastically improved compared with previous studies, with a mean unsigned error of 0.86 kcal/mol and root mean square error of 1.11 kcal/mol, comparable with the accuracy of previously published state-of-the-art small-molecule relative binding affinity calculations, which have been shown to be capable of driving discovery projects.

Exploring conformational search protocols for ligand-based virtual screening and 3-D QSAR modeling.

3-D ligand conformations are required for most ligand-based drug design methods, such as pharmacophore modeling, shape-based screening, and 3-D QSAR model building. Many studies of conformational search methods have focused on the reproduction of crystal structures (i.e. bioactive conformations); however, for ligand-based modeling the key question is how to generate a ligand alignment that produces the best results for a given query molecule. In this work, we study different conformation generation modes of ConfGen and the impact on virtual screening (Shape Screening and e-Pharmacophore) and QSAR predictions (atom-based and field-based). In addition, we develop a new search method, called common scaffold alignment, that automatically detects the maximum common scaffold between each screening molecule and the query to ensure identical coordinates of the common core, thereby minimizing the noise introduced by analogous parts of the molecules. In general, we find that virtual screening results are relatively insensitive to the conformational search protocol; hence, a conformational search method that generates fewer conformations could be considered "better" because it is more computationally efficient for screening. However, for 3-D QSAR modeling we find that more thorough conformational sampling tends to produce better QSAR predictions. In addition, significant improvements in QSAR predictions are obtained with the common scaffold alignment protocol developed in this work, which focuses conformational sampling on parts of the molecules that are not part of the common scaffold.

Robust Prediction of Relative Binding Energies for Protein-Protein Complex Mutations Using Free Energy Perturbation Calculations.

Computational free energy-based methods have the potential to significantly improve throughput and decrease costs of protein design efforts. Such methods must reach a high level of reliability, accuracy, and automation to be effectively deployed in practical industrial settings in a way that impacts protein design projects. Here, we present a benchmark study for the calculation of relative changes in protein-protein binding affinity for single point mutations across a variety of systems from the literature, using free energy perturbation (FEP+) calculations. We describe a method for robust treatment of alternate protonation states for titratable amino acids, which yields improved correlation with and reduced error compared to experimental binding free energies. Following careful analysis of the largest outlier cases in our dataset, we assess limitations of the default FEP+ protocols and introduce an automated script which identifies probable outlier cases that may require additional scrutiny and calculates an empirical correction for a subset of charge-related outliers. Through a series of three additional case study systems, we discuss how Protein FEP+ can be applied to real-world protein design projects, and suggest areas of further study.

Robust prediction of relative binding energies for protein-protein complex mutations using free energy perturbation calculations.

Computational free energy-based methods have the potential to significantly improve throughput and decrease costs of protein design efforts. Such methods must reach a high level of reliability, accuracy, and automation to be effectively deployed in practical industrial settings in a way that impacts protein design projects. Here, we present a benchmark study for the calculation of relative changes in protein-protein binding affinity for single point mutations across a variety of systems from the literature, using free energy perturbation (FEP+) calculations. We describe a method for robust treatment of alternate protonation states for titratable amino acids, which yields improved correlation with and reduced error compared to experimental binding free energies. Following careful analysis of the largest outlier cases in our dataset, we assess limitations of the default FEP+ protocols and introduce an automated script which identifies probable outlier cases that may require additional scrutiny and calculates an empirical correction for a subset of charge-related outliers. Through a series of three additional case study systems, we discuss how protein FEP+ can be applied to real-world protein design projects, and suggest areas of further study.

Erratum: A novel selective AKR1C3-activated prodrug AST-3424/OBI-3424 exhibits broad anti-tumor activity.

[This corrects the article on p. 3645 in vol. 11, PMID: 34354865.].

The In-Vitro Antitumor Effects of AST-3424 Monotherapy and Combination Therapy With Oxaliplatin or 5-Fluorouracil in Primary Liver Cancer.

Background: Primary liver cancer (PLC) is a common and highly lethal malignancy in the world. Approximately 85% of PLC is hepatocellular carcinoma (HCC), and this study mainly focuses on HCC. The onset of liver cancer is insidious and often complicated with basic liver disease. Meanwhile, its clinical symptoms are atypical, and the degree of malignancy is high. What is worse is that its treatment is difficult, and the prognosis is poor. All these factors make its mortality close to its incidence. AST-3424 is a prodrug of a potent nitrogen mustard, which targets the tumor by its specific and selective mode of activation and results in the concentration of the drug in the tumor and plays a higher intensity of antitumor effect with reduced side effects. The purpose of this study was to explore the in-vitro antitumor activity and mechanism of AST-3424 monotherapy and combination therapy with oxaliplatin (OXA) or 5-fluorouracil (5-Fu). Moreover, it can provide an experimental basis for further studies. Methods: Tumor growth of HCC cells was examined by using the Cell Counting Kit-8 (CCK-8), flow cytometry, and clone formation assays. Tumor migration of HCC cells was examined by using the Transwell assay. The in-vitro antitumor activity of AST-3424 monotherapy and combination therapy with OXA and 5-Fu was quantified by growth and metastasis inhibition rate. The underlying molecular mechanism was investigated by using Western blotting. Results: The inhibiting effects of AST-3424 were significant in both HepG2 cells and PLC/PRF/5 cells. Moreover, HepG2 cells showed higher sensitivity to AST-3424. With increasing AST-3424 concentration, AKR1C3 protein expression level was downregulated significantly. The inhibition of AST-3424 was significantly higher than OXA, 5-Fu, Sor (sorafenib), and Apa (apatinib) in both HCC cells. AST-3424 monotherapy and combination therapy with OXA or 5-Fu all strongly inhibited the proliferation of HCC cells, blocked HCC cells in the S phase, promoted apoptosis induction, and suppressed the migration of HCC cells. Among them, the antitumor effect of AST-3424 in combination with OXA was obviously enhanced. Western blotting analysis demonstrated the regulation of P21, Bax, Caspase3, PARP, MMP-2, MMP-9, and p-Smad proteins in the presence of AST-3424 monotherapy and combination therapy with OXA or 5-Fu, indicating that its antitumor mechanisms may be associated with the regulation of the TGF-ß signaling cascade. Conclusion: The in-vitro studies revealed that AST-3424 in combination with both OXA and 5-Fu showed an increased antitumor effect, and the combination with OXA resulted in a synergistic effect. Together with the in-vitro results, additional in-vitro and in-vivo studies are warranted to further certify its antitumor effects and explore more potential antitumor mechanisms.

A novel selective AKR1C3-activated prodrug AST-3424/OBI-3424 exhibits broad anti-tumor activity.

AST-3424/OBI-3424 (denoted by 3424) is a novel prodrug bis-alkylating agent activated by AKR1C3. AKR1C3 is overexpressed in many types of cancer, particularly in liver, non-small cell lung, gastric, renal and CRPC cancer. Currently 3424 is being studied in phase 1/2 clinical trials for the treatment of solid and hematologic cancers, and it represents potentially a novel, selective anti-cancer agent for multiple indications. In this study, AKR1C3-dependent activation of 3424 was investigated in vitro using recombinant human AKR1C3. AKR1C3-dependent cytotoxicity of 3424 was determined in a wide range of human cancer cell lines with different AKR1C3 expression levels. In addition, anti-tumor activity of 3424 was also investigated in a broad panel of CDX and PDX models. AKR1C3-dependent activation of prodrug 3424 was evident by monitoring the decrease of 3424 and generation of the active form, 2660. Kinetic analysis indicated that AKR1C3 exhibited higher catalytic efficiency towards 3424 compared to the physiological substrates. There was a strong correlation between 3424 cytotoxic potency and AKR1C3 expression. The racemic mixture induced DNA cross-linking in a concentration dependent manner. Tumor growth inhibition of 3424 was shown to be better than or comparable to the standard of care chemotherapy at clinically achievable doses as a single agent in various CDX models with high expression of AKR1C3, including liver HepG2, lung H460, castration-resistant prostate VCaP, gastric SNU-16, and kidney A498 cancer cell lines. The excellent anti-tumor efficacy of 3424 was further demonstrated in PDX models which have high level of AKR1C3 expression, but not in a model with low level of AKR1C3 expression. In the combination therapy, we showed that 3424 could enhance the efficacy of the standard care of chemotherapy in the CDX models. The results described here highlight that 3424 exhibits AKR1C3-dependent cytotoxicity in vitro and anti-tumor activity in vivo in a wide range of human cancer types, which support further development of 3424 as an anti-cancer agent for treating different types of cancers and the use of AKR1C3 as a biomarker to profile cancer patients and further guide patient selection for therapy with 3424.

Understanding the molecular activity of alkaline sphingomyelinase (NPP7) by computer modeling.

The enzymes in the nucleotide pyrophosphatase/phosphodiesterase (NPP) family have various substrates such as nucleotides, phospholipids, and sphingolipids. The substrate specificity in relation to their structures is largely unknown because no mammalian NPP complex has been crystallized. NPP7, also called alkaline sphingomyelinase (alk-SMase), is a NPP family member that may have important implications in carcinogenesis and cholesterol absorption. The sequence of NPP7 is 36% similar to that of the closest NPP member, but NPP7 has no activity against nucleotides. In this work, we predict the three-dimensional structure of NPP7 by homology modeling using a recently crystallized NPP from bacteria. Using the model, we studied the substrate specificity of the enzyme by docking. The model generated explains the functional changes in previous mutagenesis studies and rationalizes the structural basis for the lack of activity toward nucleotides. An effort to shift the substrate specificity from sphingomyelin (SM) to nucleotide was not successful but revealed a site-directed mutation that increased activity toward SM. In conclusion, this is the first study to predict the structure of a mammalian NPP and its substrate specificity by molecular modeling. The information may be helpful in understanding the functional differences of NPP members.

Virtual screening, selection and development of a benzindolone structural scaffold for inhibition of lumazine synthase.

Virtual screening of a library of commercially available compounds versus the structure of Mycobacterium tuberculosis lumazine synthase identified 2-(2-oxo-1,2-dihydrobenzo[cd]indole-6-sulfonamido)acetic acid (9) as a possible lead compound. Compound 9 proved to be an effective inhibitor of M. tuberculosis lumazine synthase with a K(i) of 70microM. Lead optimization through replacement of the carboxymethylsulfonamide sidechain with sulfonamides substituted with alkyl phosphates led to a four-carbon phosphate 38 that displayed a moderate increase in enzyme inhibitory activity (K(i) 38microM). Molecular modeling based on known lumazine synthase/inhibitor crystal structures suggests that the main forces stabilizing the present benzindolone/enzyme complexes involve pi-pi stacking interactions with Trp27 and hydrogen bonding of the phosphates with Arg128, the backbone nitrogens of Gly85 and Gln86, and the side chain hydroxyl of Thr87.

OBI-3424, a Novel AKR1C3-Activated Prodrug, Exhibits Potent Efficacy against Preclinical Models of T-ALL.

PURPOSE: OBI-3424 is a highly selective prodrug that is converted by aldo-keto reductase family 1 member C3 (AKR1C3) to a potent DNA-alkylating agent. OBI-3424 has entered clinical testing for hepatocellular carcinoma and castrate-resistant prostate cancer, and it represents a potentially novel treatment for acute lymphoblastic leukemia (ALL). EXPERIMENTAL DESIGN: We assessed AKR1C3 expression by RNA-Seq and immunoblotting, and evaluated the in vitro cytotoxicity of OBI-3424. We investigated the pharmacokinetics of OBI-3424 in mice and nonhuman primates, and assessed the in vivo efficacy of OBI-3424 against a large panel of patient-derived xenografts (PDX). RESULTS: AKR1C3 mRNA expression was significantly higher in primary T-lineage ALL (T-ALL; n = 264) than B-lineage ALL (B-ALL; n = 1,740; P < 0.0001), and OBI-3424 exerted potent cytotoxicity against T-ALL cell lines and PDXs. In vivo, OBI-3424 significantly prolonged the event-free survival (EFS) of nine of nine ALL PDXs by 17.1-77.8 days (treated/control values 2.5-14.0), and disease regression was observed in eight of nine PDXs. A significant reduction (P < 0.0001) in bone marrow infiltration at day 28 was observed in four of six evaluable T-ALL PDXs. The importance of AKR1C3 in the in vivo response to OBI-3424 was verified using a B-ALL PDX that had been lentivirally transduced to stably overexpress AKR1C3. OBI-3424 combined with nelarabine resulted in prolongation of mouse EFS compared with each single agent alone in two T-ALL PDXs. CONCLUSIONS: OBI-3424 exerted profound in vivo efficacy against T-ALL PDXs derived predominantly from aggressive and fatal disease, and therefore may represent a novel treatment for aggressive and chemoresistant T-ALL in an AKR1C3 biomarker-driven clinical trial.

Potent and highly selective hypoxia-activated achiral phosphoramidate mustards as anticancer drugs.

A series of achiral hypoxia-activated prodrugs were synthesized on the basis of the DNA cross-linking toxin of the prodrug, ifosfamide. The hypoxia-selective cytotoxicity of several of the compounds was improved over previously reported racemic mixtures of chiral bioreductive phosphoramidate prodrugs. Prodrugs activated by 2-nitroimidazole reduction demonstrated up to 400-fold enhanced cytotoxicity toward H460 cells in culture under hypoxia versus their potency under aerobic conditions. Compounds were further assessed for their stability to cytochrome P450 metabolism using a liver microsome assay. The 2-nitroimidazole containing lead compound 3b (TH-302) was selectively potent under hypoxia and stable to liver microsomes. It was active in an in vivo MIA PaCa-2 pancreatic cancer orthotopic xenograft model as a monotherapy and demonstrated dramatic efficacy when used in combination with gemcitabine, extending survival with one of eight animals tumor free at day-44. Compound 3b has emerged as a promising antitumor agent that shows excellent in vivo efficacy and is currently being evaluated in the clinic.

A novel class of tubulin inhibitors that exhibit potent antiproliferation and in vitro vessel-disrupting activity.

PURPOSE: Since anticancer agents that interfere with microtubule function are in widespread use and have a broad spectrum of activity against both hematological malignancies and solid tumors, there is an urgent need to develop novel tubulin inhibitors with broader activities and avoiding drug resistance. METHODS AND RESULTS: In this study, we describe the characterization of select lead compounds from a novel class of indazole-based tubulin inhibitors. Three lead compounds, TH-337, TH-482 and TH-494, exhibit potent antiproliferative activity against cell lines derived from human pancreatic carcinoma, human breast adenocarcinoma and human colorectal adenocarcinoma cells. The three compounds were also tested for cytotoxicity against a panel of clinically relevant drug resistant cancer cell lines that either overexpress the drug resistance pumps MDR-1, MRP-1 and BCRP-1 or have altered Topoisomerase II activity. TH-482 and -494 retained cytotoxic activities against all of the resistant cell lines tested; however, TH-337 exhibited decreased cytotoxicity in the cell line overexpressing BCRP-1, indicating that TH-337 is a substrate of that pump. We show that TH-482's antiproliferative activity is due to cell cycle arrest at the G(2)/M phase. We demonstrate that TH-482 binds specifically to the colchicine site of tubulin and that it inhibits tubulin polymerization in vitro in a concentration-dependent manner. The in vitro anti-vascular activities of TH-482 were assessed using the HUVEC-C cell line. TH-482 inhibits in vitro neovessel formation and disrupts pre-established vessels using HUVEC-C cells. TH-482 also increases permeability of vascular endothelial cells in a concentration- and time-dependent manner. CONCLUSIONS: TH-482 demonstrates potent in vitro efficacy as a novel tubulin-targeted anti-proliferative and anti-vascular agent and notably is more potent in antiproliferative assays than the benchmark compound combretastatin A-4. These results identify TH-482 as a potent tubulin inhibitor, and support the investigation of its in vivo efficacy and pharmacokinetic properties as the prototype of a new class of anti-tubulin agents.

Potent antitubulin tumor cell cytotoxins based on 3-aroyl indazoles.

A series of 3-aroyl indazoles was synthesized. Modification of the C-7 position resulted in a significant structure-activity relationship (SAR) with acetylene modifications conferring unusual potency in a tumor cell cytotoxicity assay. The most potent compounds exceeded the activity of combretastatin A4 (CA-4), showing single digit nM IC50 values against all cell lines tested including those with known efflux resistance pumps. The inhibition of in vitro tubulin polymerization was comparable to CA-4, consistent with tubulin being the target for these compounds. Competition binding experiments employing [3H]colchicine and purified tubulins demonstrated that the compound specifically binds to the colchicine site.

Predicting the Effect of Amino Acid Single-Point Mutations on Protein Stability-Large-Scale Validation of MD-Based Relative Free Energy Calculations.

The stability of folded proteins is critical to their biological function and for the efficacy of protein therapeutics. Predicting the energetic effects of protein mutations can improve our fundamental understanding of structural biology, the molecular basis of diseases, and possible routes to addressing those diseases with biological drugs. Identifying the effect of single amino acid point mutations on the thermodynamic equilibrium between the folded and unfolded states of a protein can pinpoint residues of critical importance that should be avoided in the process of improving other properties (affinity, solubility, viscosity, etc.) and suggest changes at other positions for increasing stability in protein engineering. Multiple computational tools have been developed for in silico predictions of protein stability in recent years, ranging from sequence-based empirical approaches to rigorous physics-based free energy methods. In this work, we show that FEP+, which is a free energy perturbation method based on all-atom molecular dynamics simulations, can provide accurate thermal stability predictions for a wide range of biologically relevant systems. Significantly, the FEP+ approach, while originally developed for relative binding free energies of small molecules to proteins and not specifically fitted for protein stability calculations, performs well compared to other methods that were fitted specifically to predict protein stability. Here, we present the broadest validation of a rigorous free energy-based approach applied to protein stability reported to date: 700+ single-point mutations spanning 10 different protein targets. Across the entire data set, we correctly classify the mutations as stabilizing or destabilizing in 84% of the cases, and obtain statistically significant predictions as compared with experiment [average error of ~1.6kcal/mol and coefficient of determination (R2) of 0.40]. This study demonstrates, for the first time in a large-scale validation, that rigorous free energy calculations can be used to predict changes in protein stability from point mutations without parameterization or system-specific customization, although further improvements should be possible with additional sampling and a better representation of the unfolded state of the protein. Here, we describe the FEP+ method as applied to protein stability calculations, summarize the large-scale retrospective validation results, and discuss limitations of the method, along with future directions for further improvements.

Generation of specific inhibitors of SUMO-1- and SUMO-2/3-mediated protein-protein interactions using Affimer (Adhiron) technology.

Because protein-protein interactions underpin most biological processes, developing tools that target them to understand their function or to inform the development of therapeutics is an important task. SUMOylation is the posttranslational covalent attachment of proteins in the SUMO family (SUMO-1, SUMO-2, or SUMO-3), and it regulates numerous cellular pathways. SUMOylated proteins are recognized by proteins with SUMO-interaction motifs (SIMs) that facilitate noncovalent interactions with SUMO. We describe the use of the Affimer system of peptide display for the rapid isolation of synthetic binding proteins that inhibit SUMO-dependent protein-protein interactions mediated by SIMs both in vitro and in cells. Crucially, these synthetic proteins did not prevent SUMO conjugation either in vitro or in cell-based systems, enabling the specific analysis of SUMO-mediated protein-protein interactions. Furthermore, through structural analysis and molecular modeling, we explored the molecular mechanisms that may underlie their specificity in interfering with either SUMO-1-mediated interactions or interactions mediated by either SUMO-2 or SUMO-3. Not only will these reagents enable investigation of the biological roles of SUMOylation, but the Affimer technology used to generate these synthetic binding proteins could also be exploited to design or validate reagents or therapeutics that target other protein-protein interactions.

AutoQSAR: an automated machine learning tool for best-practice quantitative structure-activity relationship modeling.

AIM: We introduce AutoQSAR, an automated machine-learning application to build, validate and deploy quantitative structure-activity relationship (QSAR) models. METHODOLOGY/RESULTS: The process of descriptor generation, feature selection and the creation of a large number of QSAR models has been automated into a single workflow within AutoQSAR. The models are built using a variety of machine-learning methods, and each model is scored using a novel approach. Effectiveness of the method is demonstrated through comparison with literature QSAR models using identical datasets for six end points: protein-ligand binding affinity, solubility, blood-brain barrier permeability, carcinogenicity, mutagenicity and bioaccumulation in fish. CONCLUSION: AutoQSAR demonstrates similar or better predictive performance as compared with published results for four of the six endpoints while requiring minimal human time and expertise.