Large-scale computational drug repositioning to find treatments for rare diseases

Potential of Artificial Intelligence to Accelerate Drug Development for Rare Diseases

The growth in breadth and depth of artificial intelligence (AI) applications has been fast, running hand in hand with the increasing amount of digital data available. Here, we comment on the application of AI in the field of drug development, with a strong focus on the specific achievements and challenges posed by rare diseases. Data paucity and high costs make drug development for rare diseases especially hard. AI can enable otherwise inaccessible approaches based on the large-scale integration of heterogeneous datasets and knowledge bases, guided by expert biological understanding. Obstacles still exist for the routine use of AI in the usually conservative pharmaceutical domain, which can easily become disillusioned. It is crucial to acknowledge that AI is a powerful, supportive tool that can assist but not replace human expertise in the various phases and aspects of drug discovery and development.

In silico screening of phenylethanoid glycosides, a class of pharmacologically active compounds as natural inhibitors of SARS-CoV-2 proteases

Since the advent of Covid-19, several natural products have been investigated regarding their in silico interactions with SARS-CoV-2 proteases - 3CL^pro and PL^pro, two of the most important pharmacological targets for antiviral development. Phenylethanoid glycosides (PG) are a class of natural products present in important medicinal plants and a drug containing this group of active ingredients has been successfully used in the treatment of Covid-19 in China. Thus, a dataset with 567 derivatives of this class was built from reviews published between 1994 and 2020, and their interaction against both SARS-CoV-2 proteases was investigated. The virtual screening was performed by filtering the PGs through the evaluation of scores based on the AutoDock Vina, GOLD/ChemPLP, and GOLD/GoldScore evaluation functions. The bRO5 pharmacokinetic parameters of the PGs ranked in the previous step were analyzed and their interaction with key amino acid residues of the 3CL^pro and PL^pro enzymes was evaluated. Ninety-eight compounds were identified by computational approaches against PL^pro and 80 PGs against 3CL^pro. Of these, four interacted with key catalytic residues of PL^pro, which is an indicative of inhibitory activity, and three compounds interacted with catalytic key residues of 3CL^pro. Of these, five PGs occur in plants of the Traditional Chinese Medicine (TCM), while two are components of plants/formulations currently used in the Covid-19 protocols in China. The data presented here show the potential of PGs as selective inhibitors of SARS-CoV-2 3CL^pro and PL^pro.

Procoxacin bidirectionally inhibits osteoblastic and osteoclastic activity in bone and suppresses bone metastasis of prostate cancer

BACKGROUND

Bone is the most common site of metastasis of prostate cancer (PCa). PCa invasion leads to a disruption of osteogenic-osteolytic balance and causes abnormal bone formation. The interaction between PCa and bone stromal cells, especially osteoblasts (OB), is considered essential for the disease progression. However, drugs that effectively block the cancer-bone interaction and regulate the osteogenic-osteolytic balance remain undiscovered.

METHODS

A reporter gene system was constructed to screen compounds that could inhibit PCa-induced OB activation from 631 compounds. Then, the pharmacological effects of a candidate drug, Procoxacin (Pro), on OBs, osteoclasts (OCs) and cancer-bone interaction were studied in cellular models. Intratibial inoculation, micro-CT and histological analysis were used to explore the effect of Pro on osteogenic and osteolytic metastatic lesions. Bioinformatic analysis and experiments including qPCR, western blotting and ELISA assay were used to identify the effector molecules of Pro in the cancer-bone microenvironment. Virtual screening, molecular docking, surface plasmon resonance assay and RNA knockdown were utilized to identify the drug target of Pro. Experiments including co-IP, western blotting and immunofluorescence were performed to reveal the role of Pro binding to its target. Intracardiac inoculation metastasis model and survival analysis were used to investigate the therapeutic effect of Pro on metastatic cancer.

RESULTS

Luciferase reporter gene consisted of Runx2 binding sequence, OSE2, and Alp promotor could sensitively reflect the intensity of PCa-OB interaction. Pro best matched the screening criteria among 631 compounds in drug screening. Further study demonstrated that Pro effectively inhibited the PCa-induced osteoblastic changes without killing OBs or PCa cells and directly killed OCs or suppressed osteoclastic functions at very low concentrations. Mechanism study revealed that Pro broke the feedback loop of TGF-β/C-Raf/MAPK pathway by sandwiching into 14-3-3ζ/C-Raf complex and prevented its disassociation. Pro treatment alleviated both osteogenic and osteolytic lesions in PCa-involved bones and reduced the number of metastases of PCa in vivo.

CONCLUSIONS

In summary, our study provides a drug screening strategy based on the cancer-host microenvironment and demonstrates that Pro effectively inhibits both osteoblastic and osteoclastic lesions in PCa-involved bones, which makes it a promising therapeutic agent for PCa bone metastasis.

A chronicle review of new techniques that facilitate the understanding and development of optimal individualized therapeutic strategies for chordoma

Chordoma is a rare malignant bone tumor that mainly occurs in the sacrum and the clivus/skull base. Surgical resection is the treatment of choice for chordoma, but the local recurrence rate is high with unsatisfactory prognosis. Compared with other common tumors, there is not much research and individualized treatment for chordoma, partly due to the rarity of the disease and the lack of appropriate disease models, which delay the discovery of therapeutic strategies. Recent advances in modern techniques have enabled gaining a better understanding of a number of rare diseases, including chordoma. Since the beginning of the 21st century, various chordoma cell lines and animal models have been reported, which have partially revealed the intrinsic mechanisms of tumor initiation and progression with the use of next-generation sequencing (NGS) techniques. In this study, we performed a systematic overview of the chordoma models and related sequencing studies in a chronological manner, from the first patient-derived chordoma cell line (U-CH1) to diverse preclinical models such as the patient-derived organoid-based xenograft (PDX) and patient-derived organoid (PDO) models. The use of modern sequencing techniques has discovered mutations and expression signatures that are considered potential treatment targets, such as the expression of Brachyury and overactivated receptor tyrosine kinases (RTKs). Moreover, computational and bioinformatics techniques have made drug repositioning/repurposing and individualized high-throughput drug screening available. These advantages facilitate the research and development of comprehensive and personalized treatment strategies for indicated patients and will dramatically improve their prognoses in the near feature.

GraphSite: Ligand Binding Site Classification with Deep Graph Learning

The binding of small organic molecules to protein targets is fundamental to a wide array of cellular functions. It is also routinely exploited to develop new therapeutic strategies against a variety of diseases. On that account, the ability to effectively detect and classify ligand binding sites in proteins is of paramount importance to modern structure-based drug discovery. These complex and non-trivial tasks require sophisticated algorithms from the field of artificial intelligence to achieve a high prediction accuracy. In this communication, we describe GraphSite, a deep learning-based method utilizing a graph representation of local protein structures and a state-of-the-art graph neural network to classify ligand binding sites. Using neural weighted message passing layers to effectively capture the structural, physicochemical, and evolutionary characteristics of binding pockets mitigates model overfitting and improves the classification accuracy. Indeed, comprehensive cross-validation benchmarks against a large dataset of binding pockets belonging to 14 diverse functional classes demonstrate that GraphSite yields the class-weighted F1-score of 81.7%, outperforming other approaches such as molecular docking and binding site matching. Further, it also generalizes well to unseen data with the F1-score of 70.7%, which is the expected performance in real-world applications. We also discuss new directions to improve and extend GraphSite in the future.

Evaluating the performance of drug-repurposing technologies

Drug-repurposing technologies are growing in number and maturing. However, comparisons to each other and to reality are hindered because of a lack of consensus with respect to performance evaluation. Such comparability is necessary to determine scientific merit and to ensure that only meaningful predictions from repurposing technologies carry through to further validation and eventual patient use. Here, we review and compare performance evaluation measures for these technologies using version 2 of our shotgun repurposing Computational Analysis of Novel Drug Opportunities (CANDO) platform to illustrate their benefits, drawbacks, and limitations. Understanding and using different performance evaluation metrics ensures robust cross-platform comparability, enabling us to continue to strive toward optimal repurposing by decreasing the time and cost of drug discovery and development.

The RareDis corpus: A corpus annotated with rare diseases, their signs and symptoms

Rare diseases affect a small number of people compared to the general population. However, more than 6,000 different rare diseases exist and, in total, they affect more than 300 million people worldwide. Rare diseases share as part of their main problem, the delay in diagnosis and the sparse information available for researchers, clinicians, and patients. Finding a diagnostic can be a very long and frustrating experience for patients and their families. The average diagnostic delay is between 6-8 years. Many of these diseases result in different manifestations among patients, which hampers even more their detection and the correct treatment choice. Therefore, there is an urgent need to increase the scientific and medical knowledge about rare diseases. Natural Language Processing (NLP) can help to extract relevant information about rare diseases to facilitate their diagnosis and treatments, but most NLP techniques require manually annotated corpora. Therefore, our goal is to create a gold standard corpus annotated with rare diseases and their clinical manifestations. It could be used to train and test NLP approaches and the information extracted through NLP could enrich the knowledge of rare diseases, and thereby, help to reduce the diagnostic delay and improve the treatment of rare diseases. The paper describes the selection of 1,041 texts to be included in the corpus, the annotation process and the annotation guidelines. The entities (disease, rare disease, symptom, sign and anaphor) and the relationships (produces, is a, is acron, is synon, increases risk of, anaphora) were annotated. The RareDis corpus contains more than 5,000 rare diseases and almost 6,000 clinical manifestations are annotated. Moreover, the Inter Annotator Agreement evaluation shows a relatively high agreement (F1-measure equal to 83.5% under exact match criteria for the entities and equal to 81.3% for the relations). Based on these results, this corpus is of high quality, supposing a significant step for the field since there is a scarcity of available corpus annotated with rare diseases. This could open the door to further NLP applications, which would facilitate the diagnosis and treatment of these rare diseases and, therefore, would improve dramatically the quality of life of these patients.

Suppression of LPS-Induced Inflammation and Cell Migration by Azelastine through Inhibition of JNK/NF-κB Pathway in BV2 Microglial Cells

The c-Jun N-terminal kinases (JNKs) are implicated in many neuropathological conditions, including neurodegenerative diseases. To explore potential JNK3 inhibitors from the U.S. Food and Drug Administration-approved drug library, we performed structure-based virtual screening and identified azelastine (Aze) as one of the candidates. NMR spectroscopy indicated its direct binding to the ATP-binding site of JNK3, validating our observations. Although the antihistamine effect of Aze is well documented, the involvement of the JNK pathway in its action remains to be elucidated. This study investigated the effects of Aze on lipopolysaccharide (LPS)-induced JNK phosphorylation, pro-inflammatory mediators, and cell migration in BV2 microglial cells. Aze was found to inhibit the LPS-induced phosphorylation of JNK and c-Jun. It also inhibited the LPS-induced production of pro-inflammatory mediators, including interleukin-6, tumor necrosis factor-α, and nitric oxide. Wound healing and transwell migration assays indicated that Aze attenuated LPS-induced BV2 cell migration. Furthermore, Aze inhibited LPS-induced IκB phosphorylation, thereby suppressing nuclear translocation of NF-κB. Collectively, our data demonstrate that Aze exerts anti-inflammatory and anti-migratory effects through inhibition of the JNK/NF-κB pathway in BV2 cells. Based on our findings, Aze may be a potential candidate for drug repurposing to mitigate neuroinflammation in various neurodegenerative disorders, including Alzheimer’s and Parkinson’s diseases.

A computational multi-targeting approach for drug repositioning for psoriasis treatment

Background

Psoriasis is an autoimmune inflammatory skin disease that affects 0.5–3% of the world’s population and current treatment options are posed with limitations. The reduced risk of failure in clinical trials for repositioned drug candidates and the time and cost-effectiveness has popularized drug reposition and computational methods in the drug research community.

Results

The current study attempts to reposition approved drugs for the treatment of psoriasis by docking about 2000 approved drug molecules against fifteen selected and validated anti-psoriatic targets. The docking results showed that a good number of the dataset interacted favorably with the targets as most of them had − 11.00 to − 10.00 kcal/mol binding free energies across the targets. The percentage of the dataset with binding affinity higher than the co-crystallized ligands ranged from 34.76% (JAK-3) to 0.73% (Rac-1). It was observed that 12 out of the 0.73% outperformed all the co-crystallized ligands across the 15 studied proteins. All the 12 drugs identified are currently indicated as either antiviral or anticancer drugs and are of purine and pyrimidine nuclei. This is not surprising given that there is similarity in the mechanism of the mentioned diseases.

Conclusion

This study, therefore, suggests that; antiviral and anticancer drugs could have anti-psoriatic effects, and molecules with purine and pyrimidine structural architecture are likely templates to consider in developing anti-psoriatic agents.

Computational Drug Repositioning Identifies Potentially Active Therapies for Chordoma

BACKGROUND

Chordomas are aggressive bone tumors that often recur despite maximal resection and adjuvant radiation. To date there are no Food and Drug Administration (FDA)-approved chemotherapies. Computational drug repositioning is an expanding approach to identify pharmacotherapies for clinical trials.

OBJECTIVE

To identify FDA-approved compounds for repurposing in chordoma.

METHODS

Previously identified highly differentially expressed genes from chordoma tissue samples at our institution were compared with pharmacogenomic interactions in the Comparative Toxicogenomics Database (CTD) using ksRepo, a drug-repositioning platform. Compounds selected by ksRepo were then validated in CH22 and UM-Chor1 human chordoma cells in Vitro.

RESULTS

A total of 13 chemical compounds were identified in silico from the CTD, and 6 were selected for preclinical validation in human chordoma cell lines based on their clinical relevance. Of these, 3 identified drugs are FDA-approved chemotherapies for other malignancies (cisplatin, cytarabine, and lucanthone). Cytarabine, a deoxyribonucleic acid polymerase inhibitor approved for the treatment of various leukemias, exhibited a significant concentration-dependent effect against CH22 and UM-Chor1 cells when compared to positive (THZ1) and negative (venetoclax) controls. Tretinoin exhibited a significant concentration-dependent cytotoxic effect in CH22, sacral chordoma-derived cell lines but to a much lesser extent in UM-Chor1, a cell line derived from skull base chordoma.

CONCLUSION

Cytarabine administration reduces the viability of human chordoma cells. The equally effective reduction in viability seen with tretinoin seems to be cell line dependent. Based on our findings, we recommend the evaluation of cytarabine and tretinoin in an expanded set of human chordoma cell lines and animal models.

A high-content drug screening strategy to identify protein level modulators for genetic diseases: A proof-of-principle in autosomal dominant leukodystrophy

In genetic diseases, the most prevalent mechanism of pathogenicity is an altered expression of dosage-sensitive genes. Drugs that restore physiological levels of these genes should be effective in treating the associated conditions. We developed a screening strategy, based on a bicistronic dual-reporter vector, for identifying compounds that modulate protein levels, and used it in a pharmacological screening approach. To provide a proof-of-principle, we chose autosomal dominant leukodystrophy (ADLD), an ultra-rare adult-onset neurodegenerative disorder caused by lamin B1 (LMNB1) overexpression. We used a stable Chinese hamster ovary (CHO) cell line that simultaneously expresses an AcGFP reporter fused to LMNB1 and a Ds-Red normalizer. Using high-content imaging analysis, we screened a library of 717 biologically active compounds and approved drugs, and identified alvespimycin, an HSP90 inhibitor, as a positive hit. We confirmed that alvespimycin can reduce LMNB1 levels by 30%-80% in five different cell lines (fibroblasts, NIH3T3, CHO, COS-7, and rat primary glial cells). In ADLD fibroblasts, alvespimycin reduced cytoplasmic LMNB1 by about 50%. We propose this approach for effectively identifying potential drugs for treating genetic diseases associated with deletions/duplications and paving the way toward Phase II clinical trials.

Cestode strobilation: prediction of developmental genes and pathways

BACKGROUND

Cestoda is a class of endoparasitic worms in the flatworm phylum (Platyhelminthes). During the course of their evolution cestodes have evolved some interesting aspects, such as their increased reproductive capacity. In this sense, they have serial repetition of their reproductive organs in the adult stage, which is often associated with external segmentation in a developmental process called strobilation. However, the molecular basis of strobilation is poorly understood. To assess this issue, an evolutionary comparative study among strobilated and non-strobilated flatworm species was conducted to identify genes and proteins related to the strobilation process.

RESULTS

We compared the genomic content of 10 parasitic platyhelminth species; five from cestode species, representing strobilated parasitic platyhelminths, and five from trematode species, representing non-strobilated parasitic platyhelminths. This dataset was used to identify 1813 genes with orthologues that are present in all cestode (strobilated) species, but absent from at least one trematode (non-strobilated) species. Development-related genes, along with genes of unknown function (UF), were then selected based on their transcriptional profiles, resulting in a total of 34 genes that were differentially expressed between the larval (pre-strobilation) and adult (strobilated) stages in at least one cestode species. These 34 genes were then assumed to be strobilation related; they included 12 encoding proteins of known function, with 6 related to the Wnt, TGF-β/BMP, or G-protein coupled receptor signaling pathways; and 22 encoding UF proteins. In order to assign function to at least some of the UF genes/proteins, a global gene co-expression analysis was performed for the cestode species Echinococcus multilocularis. This resulted in eight UF genes/proteins being predicted as related to developmental, reproductive, vesicle transport, or signaling processes.

CONCLUSIONS

Overall, the described in silico data provided evidence of the involvement of 34 genes/proteins and at least 3 developmental pathways in the cestode strobilation process. These results highlight on the molecular mechanisms and evolution of the cestode strobilation process, and point to several interesting proteins as potential developmental markers and/or targets for the development of novel antihelminthic drugs.

Drug Repositioning for Noonan and LEOPARD Syndromes by Integrating Transcriptomics With a Structure-Based Approach

Noonan and LEOPARD syndromes (NS and LS) belong to a group of related disorders called RASopathies characterized by abnormalities of multiple organs and systems including hypertrophic cardiomyopathy and dysmorphic facial features. There are no approved drugs for these two rare diseases, but it is known that a missense mutation in PTPN11 genes is associated with approximately 50% and 70% of NS and LS cases, respectively. In this study, we implemented a hybrid computational drug repositioning framework by integrating transcriptomic and structure-based approaches to explore potential treatment options for NS and LS. Specifically, disease signatures were derived from the transcriptomic profiles of human induced pluripotent stem cells (iPSCs) from NS and LS patients and reverse correlated to drug transcriptomic signatures from CMap and L1000 projects on the basis that if disease and drug transcriptomic signatures are reversely correlated, the drug has the potential to treat that disease. The compounds that were ranked top based on their transcriptomic profiles were docked to mutated and wild-type 3D structures of PTPN11 by an adjusted Induced Fit Docking (IFD) protocol. In addition, we prioritized repositioned candidates for NS and LS by a consensus ranking strategy. Network analysis and phenotypic anchoring of the transcriptomic data could discriminate the two diseases at the molecular level. Furthermore, the adjusted IFD protocol was able to recapitulate the binding specificity of potential drug candidates to mutated 3D structures, revealing the relevant amino acids. Importantly, a list of potential drug candidates for repositioning was identified including 61 for NS and 43 for LS and was further verified from literature reports and on-going clinical trials. Altogether, this hybrid computational drug repositioning approach has highlighted a number of drug candidates for NS and LS and could be applied to identifying drug candidates for other diseases as well.

Drug repositioning by prediction of drug's anatomical therapeutic chemical code via network-based inference approaches

Drug discovery and development is a time-consuming and costly process. Therefore, drug repositioning has become an effective approach to address the issues by identifying new therapeutic or pharmacological actions for existing drugs. The drug's anatomical therapeutic chemical (ATC) code is a hierarchical classification system categorized as five levels according to the organs or systems that drugs act and the pharmacology, therapeutic and chemical properties of drugs. The 2nd-, 3rd- and 4th-level ATC codes reserved the therapeutic and pharmacological information of drugs. With the hypothesis that drugs with similar structures or targets would possess similar ATC codes, we exploited a network-based approach to predict the 2nd-, 3rd- and 4th-level ATC codes by constructing substructure drug-ATC (SD-ATC), target drug-ATC (TD-ATC) and Substructure&Target drug-ATC (STD-ATC) networks. After 10-fold cross validation and two external validations, the STD-ATC models outperformed the SD-ATC and TD-ATC ones. Furthermore, with KR as fingerprint, the STD-ATC model was identified as the optimal model with AUC values at 0.899 ± 0.015, 0.916 and 0.893 for 10-fold cross validation, external validation set 1 and external validation set 2, respectively. To illustrate the predictive capability of the STD-ATC model with KR fingerprint, as a case study, we predicted 25 FDA-approved drugs (22 drugs were actually purchased) to have potential activities on heart failure using that model. Experiments in vitro confirmed that 8 of the 22 old drugs have shown mild to potent cardioprotective activities on both hypoxia model and oxygen-glucose deprivation model, which demonstrated that our STD-ATC prediction model would be an effective tool for drug repositioning.

Binding site matching in rational drug design: algorithms and applications

Interactions between proteins and small molecules are critical for biological functions. These interactions often occur in small cavities within protein structures, known as ligand-binding pockets. Understanding the physicochemical qualities of binding pockets is essential to improve not only our basic knowledge of biological systems, but also drug development procedures. In order to quantify similarities among pockets in terms of their geometries and chemical properties, either bound ligands can be compared to one another or binding sites can be matched directly. Both perspectives routinely take advantage of computational methods including various techniques to represent and compare small molecules as well as local protein structures. In this review, we survey 12 tools widely used to match pockets. These methods are divided into five categories based on the algorithm implemented to construct binding-site alignments. In addition to the comprehensive analysis of their algorithms, test sets and the performance of each method are described. We also discuss general pharmacological applications of computational pocket matching in drug repurposing, polypharmacology and side effects. Reflecting on the importance of these techniques in drug discovery, in the end, we elaborate on the development of more accurate meta-predictors, the incorporation of protein flexibility and the integration of powerful artificial intelligence technologies such as deep learning.

A network-based approach to identify deregulated pathways and drug effects in metabolic syndrome

Metabolic syndrome is a pathological condition characterized by obesity, hyperglycemia, hypertension, elevated levels of triglycerides and low levels of high-density lipoprotein cholesterol that increase cardiovascular disease risk and type 2 diabetes. Although numerous predisposing genetic risk factors have been identified, the biological mechanisms underlying this complex phenotype are not fully elucidated. Here we introduce a systems biology approach based on network analysis to investigate deregulated biological processes and subsequently identify drug repurposing candidates. A proximity score describing the interaction between drugs and pathways is defined by combining topological and functional similarities. The results of this computational framework highlight a prominent role of the immune system in metabolic syndrome and suggest a potential use of the BTK inhibitor ibrutinib as a novel pharmacological treatment. An experimental validation using a high fat diet-induced obesity model in zebrafish larvae shows the effectiveness of ibrutinib in lowering the inflammatory load due to macrophage accumulation.

Repurposing of drugs as STAT3 inhibitors for cancer therapy

Drug repurposing is a valuable approach in delivering new cancer therapeutics rapidly into the clinic. Existing safety and patient tolerability data for drugs already in clinical use represent an untapped resource in terms of identifying therapeutic agents for off-label protein targets. The multicellular effects of STAT3 mediated by a range of various upstream signaling pathways make it an attractive therapeutic target with utility in a range of diseases including cancer, and has led to the development of a variety of STAT3 inhibitors. Moreover, heightened STAT3 transcriptional activation in tumor cells and within the cells of the tumor microenvironment contribute to disease progression. Consequently, there are many STAT3 inhibitors in preclinical development or under evaluation in clinical trials for their therapeutic efficacy predominantly in inflammatory diseases and cancer. Despite these advances, many challenges remain in ultimately providing STAT3 inhibitors to patients as cancer treatments, highlighting the need not only for a better understanding of the mechanisms associated with STAT3 activation, but also how various pharmaceutical agents suppress STAT3 activity in various cancers. In this review we discuss the importance of STAT3-dependent functions in cancer, review the status of compounds designed as direct-acting STAT3 inhibitors, and describe some of the strategies for repurposing of drugs as STAT3 inhibitors for cancer therapy.

Comparing AutoDock and Vina in Ligand/Decoy Discrimination for Virtual Screening

AutoDock and Vina are two of the most widely used protein–ligand docking programs. The fact that these programs are free and available under an open source license, also makes them a very popular first choice for many users and a common starting point for many virtual screening campaigns, particularly in academia. Here, we evaluated the performance of AutoDock and Vina against an unbiased dataset containing 102 protein targets, 22,432 active compounds and 1,380,513 decoy molecules. In general, the results showed that the overall performance of Vina and AutoDock was comparable in discriminating between actives and decoys. However, the results varied significantly with the type of target. AutoDock was better in discriminating ligands and decoys in more hydrophobic, poorly polar and poorly charged pockets, while Vina tended to give better results for polar and charged binding pockets. For the type of ligand, the tendency was the same for both Vina and AutoDock. Bigger and more flexible ligands still presented a bigger challenge for these docking programs. A set of guidelines was formulated, based on the strengths and weaknesses of both docking program and their limits of validation.

DeepDrug3D: Classification of ligand-binding pockets in proteins with a convolutional neural network

Comprehensive characterization of ligand-binding sites is invaluable to infer molecular functions of hypothetical proteins, trace evolutionary relationships between proteins, engineer enzymes to achieve a desired substrate specificity, and develop drugs with improved selectivity profiles. These research efforts pose significant challenges owing to the fact that similar pockets are commonly observed across different folds, leading to the high degree of promiscuity of ligand-protein interactions at the system-level. On that account, novel algorithms to accurately classify binding sites are needed. Deep learning is attracting a significant attention due to its successful applications in a wide range of disciplines. In this communication, we present DeepDrug3D, a new approach to characterize and classify binding pockets in proteins with deep learning. It employs a state-of-the-art convolutional neural network in which biomolecular structures are represented as voxels assigned interaction energy-based attributes. The current implementation of DeepDrug3D, trained to detect and classify nucleotide- and heme-binding sites, not only achieves a high accuracy of 95%, but also has the ability to generalize to unseen data as demonstrated for steroid-binding proteins and peptidase enzymes. Interestingly, the analysis of strongly discriminative regions of binding pockets reveals that this high classification accuracy arises from learning the patterns of specific molecular interactions, such as hydrogen bonds, aromatic and hydrophobic contacts. DeepDrug3D is available as an open-source program at https://github.com/pulimeng/DeepDrug3D with the accompanying TOUGH-C1 benchmarking dataset accessible from https://osf.io/enz69/.

Taxonomic Landscape of the Dark Proteomes: Whole-Proteome Scale Interplay Between Structural Darkness, Intrinsic Disorder, and Crystallization Propensity

Growth rate of the protein sequence universe dramatically exceeds the speed of expansion for the protein structure universe, generating an immense dark proteome that includes proteins with unknown structure. A whole-proteome scale analysis of 5.4 million proteins from 987 proteomes in the three domains of life and viruses to systematically dissect an interplay between structural coverage, degree of putative intrinsic disorder, and predicted propensity for structure determination is performed. It has been found that Archaean and Bacterial proteomes have relatively high structural coverage and low amounts of disorder, whereas Eukaryotic and Viral proteomes are characterized by a broad spread of structural coverage and higher disorder levels. The analysis reveals that dark proteomes (i.e., proteomes containing high fractions of proteins with unknown structure) have significantly elevated amounts of intrinsic disorder and are predicted to be difficult to solve structurally. Although the majority of dark proteomes are of viral origin, many dark viral proteomes have at least modest crystallization propensity and only a handful of them are enriched in the intrinsic disorder. The disorder, structural coverage, and propensity are mapped for structural determination onto a novel proteome-level sequence similarity network to analyze the interplay of these characteristics in the taxonomic landscape.

eModel-BDB: a database of comparative structure models of drug-target interactions from the Binding Database

Background

The structural information on proteins in their ligand-bound conformational state is invaluable for protein function studies and rational drug design. Compared to the number of available sequences, not only is the repertoire of the experimentally determined structures of holo-proteins limited, these structures do not always include pharmacologically relevant compounds at their binding sites. In addition, binding affinity databases provide vast quantities of information on interactions between drug-like molecules and their targets, however, often lacking structural data. On that account, there is a need for computational methods to complement existing repositories by constructing the atomic-level models of drug-protein assemblies that will not be determined experimentally in the near future.

Results

We created eModel-BDB, a database of  200,005 comparative models of drug-bound proteins based on   1,391,403 interaction data obtained from the Binding Database and the PDB library of 31 January 2017. Complex models in eModel-BDB were generated with a collection of the state-of-the-art techniques, including protein meta-threading, template-based structure modeling, refinement and binding site detection, and ligand similarity-based docking. In addition to a rigorous quality control maintained during dataset generation, a subset of weakly homologous models was selected for the retrospective validation against experimental structural data recently deposited to the Protein Data Bank. Validation results indicate that eModel-BDB contains models that are accurate not only at the global protein structure level but also with respect to the atomic details of bound ligands.

Conclusions

Freely available eModel-BDB can be used to support structure-based drug discovery and repositioning, drug target identification, and protein structure determination.

Comparative assessment of strategies to identify similar ligand-binding pockets in proteins

Background

Detecting similar ligand-binding sites in globally unrelated proteins has a wide range of applications in modern drug discovery, including drug repurposing, the prediction of side effects, and drug-target interactions. Although a number of techniques to compare binding pockets have been developed, this problem still poses significant challenges.

Results

We evaluate the performance of three algorithms to calculate similarities between ligand-binding sites, APoc, SiteEngine, and G-LoSA. Our assessment considers not only the capabilities to identify similar pockets and to construct accurate local alignments, but also the dependence of these alignments on the sequence order. We point out certain drawbacks of previously compiled datasets, such as the inclusion of structurally similar proteins, leading to an overestimated performance. To address these issues, a rigorous procedure to prepare unbiased, high-quality benchmarking sets is proposed. Further, we conduct a comparative assessment of techniques directly aligning binding pockets to indirect strategies employing structure-based virtual screening with AutoDock Vina and rDock.

Conclusions

Thorough benchmarks reveal that G-LoSA offers a fairly robust overall performance, whereas the accuracy of APoc and SiteEngine is satisfactory only against easy datasets. Moreover, combining various algorithms into a meta-predictor improves the performance of existing methods to detect similar binding sites in unrelated proteins by 5–10%. All data reported in this paper are freely available at https://osf.io/6ngbs/.