Method and system for predicting drug binding using synthetic data

Abstract

Methods for predicting drug-target binding using synthetically enhanced data involve generating a plurality of phantom ligands for a plurality of proteins in a protein structure database; generating a plurality of drug-target interaction (DTI) signatures for proteins and ligands in a (DTI) database using the plurality of phantom ligands; using the DTI features to generate a machine learning model; and predicting the likelihood of interaction of the combination of query proteins and query ligands using the machine learning model.

Description

[0001] This application is a divisional application of a Chinese patent application with application number 2020800181551, filing date January 2, 2020, and invention title "Methods and Systems for Predicting Drug Binding Using Synthetic Data".

[0002] Cross - reference to related applications

[0003] This application claims the priority of U.S. Provisional Application No. 62 / 788,682, filed on January 4, 2019, under 35 U.S.C.§119(e). The provisional application has at least one inventor in common with this application and has the title "Methods and Systems for Predicting Drug Binding Using Synthetic Data". U.S. Provisional Application No. 62 / 788,682 is incorporated herein by reference. Background of the Invention

[0004] There are computational methods to predict the interaction between ligands and proteins. Depending on the type of information used for the prediction, they are generally classified as 'ligand - based' or'structure - based'.

[0005] Protein - based predictions have the potential to learn biophysical compatibility and may thus be more general, but are highly data - constrained. Specifically, protein - based predictions use the 3D molecular structure of ligands co - crystallized with proteins to evaluate or predict interactions. These methods are computationally demanding and they tend to be trained only on 100 to 1000 different proteins. Neural networks for these tend to have a very high feature space - to - data ratio. As a result, when applied to previously unseen protein systems or drug scaffolds, the method can produce a large number of false negatives and / or false positives in docking.

[0006] Ligand - based predictions can be performed using drug - target interaction (DTI) databases with millions of records. Examples of publicly available DTI databases include ChEMBL, NCBI's Bioassay, and STITCH. However, these records tend to represent only about 2,000 out of 20,000 human proteins. Due to the high ligand - to - protein data ratio, a standard approach is to derive many different models for each of the 2,000 proteins. These are often successful and even, in many cases, outperform high - throughput experimental results, but (1) they represent only about 10% of human proteins, (2) they can be weaker when there is not much chemical diversity between examples of a single protein, and (3) the individual models do not learn the physical properties of drug - protein compatibility and miss out on benefiting from the data used to generate other models. Summary of the Invention

[0007] Generally, in one aspect, one or more embodiments relate to a method for predicting drug-target binding using synthetically enhanced data, the method comprising: generating a plurality of phantom ligands for a plurality of proteins in a protein structure database; using the plurality of phantom ligands to generate a plurality of DTI features for proteins and ligands in a drug-target interaction (DTI) database; generating a machine learning model using the plurality of DTI features; and predicting the likelihood of interaction of a combination of a query protein and a query ligand using the machine learning model.

[0008] Generally, in one aspect, one or more embodiments relate to a non-transitory computer-readable medium comprising computer-readable program code for predicting drug-target binding using synthetically enhanced data, the computer-readable program code causing a computer system to: generate a plurality of phantom ligands for a plurality of proteins in a protein structure database; use the plurality of phantom ligands to generate a plurality of DTI features for proteins and ligands in a drug-target interaction (DTI) database; generate a machine learning model using the DTI features; and predict the likelihood of interaction of a combination of a query protein and a query ligand using the machine learning model.

[0009] Generally, in one aspect, one or more embodiments relate to a system for differential drug discovery, the system comprising: a protein structure database; a phantom ligand identification engine configured to generate a plurality of phantom ligands for a plurality of proteins in the protein structure database; a phantom ligand database storing the plurality of phantom ligands; a drug-target interaction (DTI) database storing proteins and ligands; a feature generation engine configured to generate a plurality of DTI features for proteins and ligands in the DTI database using the plurality of phantom ligands in the phantom ligand database; a machine learning model training engine configured to generate a machine learning model using the DTI features; and a DTI prediction engine configured to predict the likelihood of interaction of a combination of a query protein and a query ligand using the machine learning model. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Embodiments are illustrated by way of example and are not intended to be limited by the figures of the drawings.

[0011] Figure 1A A block diagram of a system for predicting drug binding according to one or more embodiments is shown.

[0012] Figure 1B A block diagram of a protein structure database according to one or more embodiments is shown.

[0013] Figure 1C A block diagram of a phantom ligand database according to one or more embodiments is shown.

[0014] Figure 1D A block diagram of a drug - target interaction database according to one or more embodiments is shown.

[0015] Figure 1E A block diagram of a protein annotation database according to one or more embodiments is shown.

[0016] Figure 2 A flowchart depicting a method for training a machine learning model for predicting drug - target interactions according to one or more embodiments is shown.

[0017] Figure 3 A flowchart depicting a method for generating a phantom ligand database according to one or more embodiments is shown.

[0018] Figure 4 A flowchart depicting a method for generating drug - target interaction (DTI) features according to one or more embodiments is shown.

[0019] Figure 5 A flowchart depicting a method for generating a machine learning model for DTI prediction according to one or more embodiments is shown.

[0020] Figure 6 A flowchart depicting a method for predicting the interaction between a query protein and a query ligand according to one or more embodiments is shown.

[0021] Figure 7A An example of generating a phantom ligand according to one or more embodiments is shown.

[0022] Figure 7B An illustration of a concentric shell model for obtaining binding site features according to one or more embodiments is shown.

[0023] Figure 8 The generation of training data for a machine learning model according to one or more embodiments is shown.

[0024] Figure 9 A performance comparison between embodiments of the present disclosure and traditional methods is shown.

[0025] Figure 10A and Figure 10B A computing system according to one or more embodiments is shown. Detailed Description

[0026] The detailed embodiments disclosed herein will now be described in detail with reference to the accompanying drawings. For consistency, the same elements in each figure may be represented by the same reference numerals and / or the same names.

[0027] The following detailed description is merely exemplary in nature and is not intended to limit the embodiments disclosed herein or the application and use of the embodiments disclosed herein. In addition, it is not intended to be bound by any express or implied theory presented in the aforementioned technical field, background technology, summary of the invention or the following detailed description.

[0028] In the following detailed description of some embodiments disclosed herein, many specific details are set forth in order to provide a more comprehensive understanding of the various embodiments disclosed herein. However, it will be apparent to one of ordinary skill in the art that the embodiments may be practiced without these specific details. In other cases, well-known features are not described in detail to avoid unnecessarily complicating the description.

[0029] Throughout the application, ordinal numbers (e.g., first, second, third, etc.) may be used as adjectives for elements (i.e., any noun in the application). The use of ordinal numbers does not imply or create any particular order of elements, nor does it limit any element to only a single element, unless explicitly disclosed, such as by using the terms "before," "after," "single," and other such terms. Instead, the use of ordinal numbers is to distinguish between elements. By way of example, a first element is different from a second element, and a first element may include more than one element and is after (or before) a second element in the order of the elements.

[0030] In one or more embodiments of the invention, elements of protein-based prediction methods and ligand-based prediction methods can be combined to obtain superior predictions of drug-target interactions. In one or more embodiments of the invention, machine learning models are used to predict drug-target interactions (DTIs).

[0031] There are a large number of data points on protein-ligand interactions in DTI databases. For example, the ChEMBL database contains approximately 15,000,000 records that describe protein and ligand pairs as bound or unbound (and often provide a measure of affinity or confidence). However, using the content of these DTI databases to predict the interactions of new protein and ligand pairs has certain limitations, such as limited coverage of the human proteome and weak predictive power when the diversity between individual protein instances is limited. Therefore, the predictive quality of machine learning models that only operate on DTI database records may be limited. Alternatively, machine learning models can operate on the 3D molecular structures of ligands co-crystallized with proteins to capture the biophysics of the interactions between proteins and ligands. There are some databases that capture protein-ligand interactions (e.g., sc-PDB), but these databases contain relatively few data points and tend to be highly redundant. In addition, these databases tend to lack protein structure diversity and chemotype diversity. Therefore, training machine learning models on these data points may be challenging due to insufficient data. In particular, machine learning models that operate on the 3D molecular structures of ligands co-crystallized with proteins have a high-dimensional feature space, combined with the limited availability of suitable training samples, making the machine learning models prone to overfitting.

[0032] In one or more embodiments of the present disclosure, a machine learning model combines the use of local 3D features and DTI records to achieve predictions that are superior to the above-mentioned conventional protein-based and ligand-based predictions. More specifically, synthetic data is generated by projecting DTI records onto 3D structural models of known protein-ligand complexes to generate local protein features that would otherwise not be available. The synthetic data generated in this way can be used to train a machine learning model. The machine learning model can then be used to make predictions for query protein and query ligand pairs.

[0033] Turning to Figure 1A , a system for predicting drug binding using synthetic data according to one or more embodiments is shown. The system (100) may include a phantom ligand recognition engine (110), a feature generation engine (120), a machine learning model training engine (130), a drug-target interaction prediction engine (150), a protein structure database (160), a phantom ligand database (170), a drug-target interaction database (180), and a protein annotation database (190). Each of these components will be described subsequently.

[0034] According to one or more embodiments, the phantom ligand recognition engine (110) includes instructions in the form of computer-readable program code to perform Figure 2 and 3At least one of the steps described in to generate a phantom ligand database (170) of phantom ligands and associated confidence scores from proteins in a protein structure database (160). The phantom ligand recognition engine (110) can obtain the phantom ligands of a protein by structurally aligning known homologs and projecting the ligands of these known homologs onto the protein at the alignment sites. Although these phantom ligands may not interact with the protein, they can serve as placeholders indicating the structural compatibility between the phantom ligands and the alignment sites of the protein. The phantom ligand recognition engine (110) is operably connected to the protein structure database (160) and the phantom ligand database (170).

[0035] According to one or more embodiments, the feature generation engine (120) includes instructions in the form of computer-readable program code to perform Figure 2 and 4 At least one of the steps described in to generate drug-target interaction (DTI) features for training a machine learning model for DTI prediction. The feature generation engine (110) can generate features of proteins and ligands using data from the phantom ligand database (170), the drug-target interaction database (180), and the protein annotation database (190). Thus, the feature generation engine (120) is operably connected to the phantom ligand database (170), the drug-target interaction database (180), and the protein annotation database (190).

[0036] Continue Figure 1A , according to one or more embodiments, the machine learning model training engine (130) includes instructions in the form of computer-readable program code to perform Figure 2 and 5 At least one of the steps described in to train a machine learning model (140) for DTI prediction. The machine learning model training engine (130) can be trained using the DTI features generated by the feature generation engine (120). Thus, the machine learning model training engine (130) is operably connected to the feature generation engine (120). The final machine learning model (140) for DTI prediction can be any type of classifier capable of predicting the interaction between a query drug and a query protein. In one or more embodiments, the machine learning model (140) for DTI prediction is a deep neural network.

[0037] According to one or more embodiments, the drug-target interaction (DTI) prediction engine (150) includes instructions in the form of computer-readable program code to perform Figure 6At least one of the steps described in to predict the drug-target interaction of a query drug and a query protein using a machine learning model (140). The DTI prediction engine (150) generates features of the query protein and the query ligand associated with the query drug, which are compatible with the machine learning model (140), and then calculates the likelihood of interaction based on the features and using the same machine learning model (140) trained by the machine learning model training engine (130). In various embodiments, one or more identical and / or different machine learning models may be used.

[0038] A protein structure database (160) according to one or more embodiments may be any type of storage unit and / or device for storing data (e.g., a file system, a database, a collection of tables, or any other storage mechanism). Reference is made below to Figure 1B Describe the protein structure database (160).

[0039] A phantom ligand database (170) according to one or more embodiments may be any type of storage unit and / or device for storing data (e.g., a file system, a database, a collection of tables, or any other storage mechanism). Reference is made below to Figure 1C Describe the phantom ligand database (170).

[0040] A drug-target interaction database (180) according to one or more embodiments may be any type of storage unit and / or device for storing data (e.g., a file system, a database, a collection of tables, or any other storage mechanism). Reference is made below to Figure 1D Describe the drug-target interaction database (180).

[0041] A protein annotation database (190) according to one or more embodiments may be any type of storage unit and / or device for storing data (e.g., a file system, a database, a collection of tables, or any other storage mechanism). Reference is made below to Figure 1D Describe the protein annotation database (190).

[0042] Go to Figure 1B , which shows a protein structure database (160) according to one or more embodiments. The protein structure database (160) may store 3D models (162A, 162B, 162N) of proteins. Each 3D model may be associated with a homolog model (164A, 164B, 164N) and / or an experimental model (166A, 166B, 166N). Examples of publicly available protein structure databases (160) include, but are not limited to, the Protein Data Bank (PDB) and SWISS-MODEL.

[0043] Go to Figure 1C, shows a phantom ligand database (170) according to one or more embodiments. For multiple 3D models of proteins (e.g., 172A, 172B, 172N), the phantom ligand database (170) can store the identified phantom ligands (e.g., 174A, 174B, 174N). In addition, for each identified phantom ligand, a confidence score based on similarity is included (e.g., 176A, 176B, 176N). The phantom ligand database can be established as described in Figure 3 as described.

[0044] Go to Figure 1D , shows a drug-target interaction database (180) according to one or more embodiments. The drug-target interaction database (180) can store interaction confidences (e.g., 186A, 186B, 186N) for multiple pairs of drugs (e.g., 182A, 182B, 182N) and targets (e.g., 184A, 184B, 184N). Examples of publicly available drug-target interaction databases (180) include but are not limited to STITCH and ChEMBL. These databases can contain many data points (~15,000,000 for ChEMBL).

[0045] Go to Figure 1E , shows a protein annotation database (190) according to one or more embodiments. The protein annotation database (190) can store related annotations (e.g., 194A, 192B, 194N) for multiple proteins (e.g., 192A, 192B, 192N). Annotations related to proteins can include any available information about the proteins and can be added to the protein annotation database manually or computationally. For example, the UniProt database can be used.

[0046] Figure 2 , Figure 3 , Figure 4 , Figure 5 and Figure 6 illustrate a flowchart according to one or more embodiments. Figure 2 , Figure 3 , Figure 4 and Figure 5 The flowchart depicts a method for training a machine learning model to predict drug-target interactions, and Figure 6 The flowchart depicts a method for using a machine learning model to predict drug-target interactions. It can be executed by the components of the system (100) discussed above with reference to Figure 1A and Figure 2 , Figure 3 , Figure 4 , Figure 5 and Figure 6One or more steps. In one or more embodiments, steps may be omitted, repeated Figure 2 , Figure 3 , Figure 4 , Figure 5 and Figure 6 one or more of the steps shown in Figure 2 , Figure 3 , Figure 4 , Figure 5 and Figure 6 and / or performed in an order different from that shown in Figure 2 , Figure 3 , Figure 4 , Figure 5 and Figure 6 . The scope of the present invention should not be considered limited to the

[0047] Go to Figure 2 flowchart, which shows a method for generating a machine learning model to predict drug - target interactions (DTIs). Although Figure 2 is intended to introduce the main steps of generating a machine learning model, the flowchart discussed subsequently provides a more detailed description. After completing the Figure 2 method, the resulting machine learning model can be used for predictions, as Figure 6 described.

[0048] In step 200, a phantom ligand database is generated based on proteins obtained from a protein structure database. Figure 3 provides a detailed description of step 200.

[0049] In step 202, drug - target interaction (DTI) features are generated, including ligand features and protein features. Figure 4 provides a detailed description of step 202.

[0050] In step 204, a machine learning model for DTI prediction is generated. Figure 5 provides a detailed description of step 204.

[0051] Go to Figure 3 flowchart, which describes a method for generating a phantom ligand database. The steps described subsequently are used to map known ligands to the structures of their homologs (experimental and homology models) using the homology relationships between proteins, referred to as "phantoms".

[0052] In step 300, proteins are obtained from a protein structure database. For each protein, a 3D model can be retrieved.

[0053] In step 302, the obtained proteins are clustered by sequence or domain. In one or more embodiments, protein clustering is defined as any set of two or more proteins that share similarity in primary sequence or three-dimensional topology (commonly referred to as fold). Protein clusters can be obtained directly from publicly available databases such as PDB, SCOP, CATH, PFAM, or Uniprot. For example, protein sequence alignment tools and clustering tools (such as BLAST, CD-HIT, or UCLUST) can be used to manually create protein clusters based on sequence similarity. Alternatively, proteins with three-dimensional structure models can be clustered by grouping unrelated proteins that share a common topology or fold.

[0054] In step 304, one of the clusters is selected for further processing.

[0055] In step 306, pairwise structural alignments are performed on all proteins in the selected cluster. Figure 7A An example of a structural alignment of protein 3D structures is shown. Three-dimensional (3D) structural alignment attempts to establish positional equivalence between two proteins. Structural alignment can be performed by applying rotation and / or translation transformations to the coordinates of one protein to minimize the average distance between equivalent residues. For example, structural alignment can be performed on the complete protein structure or a sub-selection of residues, such as residues around a single domain or ligand-binding site. The selection of ligand-binding site residues for 3D structural alignment is the best heuristic for mapping phantom ligands.

[0056] In step 308, phantom ligands are obtained by projecting each ligand onto the cluster counterparts. A confidence score can be obtained for each projection, which consists of individual scores representing the confidence in different components of the ligand projection. An example of projecting a ligand onto the cluster counterparts is provided in Figure 7A The confidence score can be based on any measure that quantifies the uncertainty in a heuristically defined structural representation for modeling DTI interactions. At this step, the confidence score can include measures of homology model quality, such as percentage sequence identity, sequence similarity, or QMEAN. The confidence score can also include measures that describe the quality of the structural alignment, such as root mean square deviation of local or global alignments. Multiple confidence scores can also be used.

[0057] In step 310, the phantom ligands and the associated confidence scores are stored in a phantom ligand database.

[0058] In step 312, it is determined whether there are any additional clusters to process. If there are additional clusters, the execution of the method can return to step 304 to select another cluster for processing as described in steps 306 - 310. If there are no remaining additional clusters, the execution of the method can terminate. Once Figure 3 the method terminates, the phantom ligand database can contain a comprehensive set of phantom ligands and associated confidence scores for all proteins processed as described.

[0059] Go to Figure 4 The flowchart of Figure 4 describes a method for generating drug - target interaction (DTI) features. The generated features include features of ligands and proteins. These features can subsequently be used to train machine - learning models for predicting drug - target interactions. Features can be generated for many combinations of proteins and ligands to ensure the availability of sufficient training samples.

[0060] In step 400, a drug - target interaction of a combination of a ligand and a protein is selected from a DTI database. Subsequent steps are performed for this considered combination of ligand and protein. These steps can be repeated later for other combinations of ligands and proteins.

[0061] In step 402, features are generated for the selected ligand. These features can include ligand fingerprints and ligand descriptors. The fingerprint can capture the structure of the ligand in descriptor format and can be based on the SMILES representation of the underlying molecule using a fixed - length vector. For example, molecular fingerprint methods can include atom pairs, extended - connectivity fingerprints, graph - based fingerprints, torsion fingerprints, or pharmacophore fingerprints. For example, molecular weight, number of rotatable bonds, number of hydrogen - bond donors, number of hydrogen - bond acceptors, hydrophobicity, aromaticity, and functional - group composition can be used as ligand descriptors. Molecular shape descriptors such as ellipticity, geometric descriptors, branching descriptors, or chirality descriptors can also be used.

[0062] In step 404, phantom ligands are retrieved from the phantom ligand database of the selected protein.

[0063] In step 406, each phantom ligand is scored based on the difference between each phantom ligand and the drug, or more specifically, the selected ligand. Higher similarity results in a higher score. For example, distance metrics for comparing molecular fingerprints can be used to score the similarity between ligands, such as Tanimoto distance, Dice distance, or cosine distance.

[0064] Steps 404 and 406 can be performed for all protein models (e.g., homology models or experimentally - derived models) available for the selected protein.

[0065] In step 408, the phantom ligand most similar to the selected ligand is selected for further processing.

[0066] In step 410, a confidence vector is generated for the DTI feature, which consists of individual scores representing the confidence in different components of the DTI feature and their representative phantom ligands. The confidence vector may include confidence scores representing the phantom ligand projections from step 308. The confidence vector may also include confidence scores for the fingerprint similarity between the selected ligand and the most similar phantom ligand selected in step 408. Additionally, the confidence vector may include a score for the confidence in the selected DTI. The confidence in the selected DTI can be scored based on the source of the DTI data (e.g., different scores can be assigned depending on whether the DTI data was obtained using high-throughput screening or low-throughput screening, etc.).

[0067] In step 412, local features of the selected protein around the most similar phantom ligand are obtained. The local features may include binding site features present in concentric shells with increasing radii, as Figure 7B shown. For each concentric shell, multiple descriptors are provided, such as atom type descriptors specifying the atoms present within the shell. For example, 70 atom type descriptors may be provided for each shell radius. The descriptors may also include, but are not limited to, the flexibility or rigidity of the binding site within the shell region, residue contacts within the shell region, and / or any other factors representing biophysics and the geometry of the indirect binding site. However, these features may not specify the exact positions or coordinates of the atoms. The local features may also include a graphical description of the ligand binding site, which describes the distances between the amino acids around the ligand binding site in a network format. The local features can also be defined by the shape of the pocket, corresponding to the void space not occupied by protein residues. The pocket void can be determined by pocket detection methods, such as flood filling, concavity, or solvent accessibility. The local features defined by the pocket void space may include the shape of the void space, including volume, ellipticity, curvature, branching pattern, or spatial stability based on the dynamics of nearby residues. For example, the local features defined by the residues adjacent to the pocket void space may include the orientation of the residues, the geometric availability of hydrogen bond donor and acceptor groups, hydrophobicity, aromaticity, or the geometric availability of π - stacking interactions. The local features may also include a description of the ligand binding channel, which includes solvent-exposed residues near the ligand binding site that do not directly contact the ligand in the stable binding state. The ligand binding channel is expected to form transient interactions with the ligand during the dynamic processes of binding and dissociation. The ligand binding channel features may include those similar to those defining the pocket, such as the orientation of the residues, amino acid composition, and the availability of hydrogen bond donors and acceptors.

[0068] In step 414, global features of the structure and / or sequence of the selected protein are obtained by extending to an outer shell with a large radius as described in step 412. The shell radius corresponding to the local features may include, for example or thresholds. The shell radius corresponding to the domain-level description or the global protein description can have, for example, or a greater distance threshold, or may have no distance threshold. Global features can also include descriptions of domains or folds and can be derived from publicly available databases such as SCOP, CATH, or PFAM. For example, global features can also include features derived from the protein sequence and can include the presence of consensus sequence motifs. Global features can include a description of the protein folding state, such as the presence and biophysical properties of intrinsically disordered regions, hinges, loops, ordered regions, or regulatory domains. Global features can also be described based on the distance to the ligand binding site.

[0069] In step 416, functional annotations of the selected protein are obtained. The functional annotations can be obtained from a protein annotation database. The functional annotations can include, for example, Enzyme Commission (EC) numbers, Gene Ontology (GO) annotations, or Uniprot keywords. The functional annotations can also include the presence or absence of recorded protein position-specific properties, such as catalytic sites, post-translational modifications, disease associations, or genetic variations.

[0070] In step 418, features are generated for the selected protein. These features can include local features, global features, and / or functional annotations.

[0071] In step 420, it is determined whether there are additional DTIs to be processed. If there are additional DTIs, the execution of the method can return to step 400 to select another DTI for processing, as described in steps 402 - 418. If there are no remaining additional clusters, the execution of the method can terminate. Once Figure 4 the method terminates, the comprehensive feature set of the ligands and proteins listed in the DTI database can be used.

[0072] Go to Figure 5 the flowchart that describes the method for generating a machine learning model for DTI prediction. Based on the DTI features obtained as Figure 4 described, a machine learning model reflecting the compatibility between the protein environment and ligand properties is obtained.

[0073] In step 500, ligand features and protein features are obtained. The obtaining of the ligand and protein features can be performed as Figure 4 described in steps 402 and 418 of

[0074] In step 502, by Figure 4The function of the confidence vector established in step 408 filters ligands and proteins. The filtering can implement a confidence threshold, and only samples with a confidence higher than the threshold are considered for further processing. The confidence function can convert the confidence vector into a single score for filtering. For example, the confidence function can convert the confidence score into a probability and apply Bayesian statistics to evaluate the combined probability. The confidence function can apply individual cut-off thresholds to each element of the confidence vector as a means of selecting which samples are suitable for machine learning. The confidence function threshold or equation can be set by automatically testing different combinations as hyperparameters of the machine learning algorithm.

[0075] In step 504, the ligand and protein features are concatenated to generate positive training samples.

[0076] In step 506, the ligand and protein features are shuffled. The shuffled ligand and protein features are concatenated to generate negative training samples. This step may be repeated multiple times to evaluate different positive-negative training sample ratios, such as 1:1, 1:5, 1:10, 1:19, or 1:20.

[0077] In step 508, a machine learning model for DTI prediction is trained using the positive and negative training samples. For example, a learning algorithm based on backpropagation can be used. In one or more embodiments, the training samples can be weighted based on the relevant confidence vectors. In one or more embodiments, transfer learning is used to train the machine learning model more effectively. Initially, the machine learning model can be trained by applying an initial confidence threshold in step 502. In subsequent retraining phases, the confidence threshold can be increased to reduce the number of training instances and improve their quality. Additionally or alternatively, subsequent retraining phases can restrict the training instances to select classes of drugs or targets. The machine learning model can be a supervised discriminative classification or regression model, such as a random forest, support vector machine, single-layer perceptron, or multi-layer artificial neural network. Considering the number of training data points (in the 100,000s to 10,000,000s) and the dimensionality of the training data features (in the 1000s to 10,000s), artificial neural networks are particularly suitable for this task. In one embodiment, the artificial neural network is in the form of a fully connected network, having a feature input layer, two hidden layers, for example with 512 and 256 nodes respectively, and two output nodes corresponding to interacting and non-interacting pairs. In one embodiment, an artificial neural network with multiple hidden layers omits the connections between input types and is used to create separate latent spaces representing ligand fingerprints, global protein features, local protein features, and protein functional features.

[0078] Go to Figure 6The flowchart describes a method for predicting the interaction between a query protein and a query ligand. By applying a machine learning model to a set of DTI features corresponding to the query ligand and at least one known binding site of the query protein, as described in reference Figures 2 - 5 The trained machine learning model can be used to test the "compatibility" of the query protein and the query compound.

[0079] In step 600, a query protein and a query ligand are obtained. The query protein and the query ligand can be obtained from a user who wants to obtain a prediction of the interaction between the query protein and the query ligand.

[0080] In step 602, as previously described in Figure 4 Steps 404 - 412, the possible binding sites and related local features of the query protein are obtained. Thus, one or more binding sites can be obtained from one or more experiments or homology models. Alternatively, the binding sites and related local features can be obtained from the user, for example, if the user wishes to specify a particular binding site.

[0081] In step 604, the global features and protein annotations of the query protein are obtained. The global features and protein annotations can be obtained as previously described in Figure 4 Steps 414 and 416.

[0082] In step 606, features are generated for the query protein. These features can include local features, global features, and / or functional annotations.

[0083] In step 608, as previously described in Figure 4 Step 402, the ligand fingerprint and ligand descriptors are obtained.

[0084] In step 610, features are generated for the query ligand. These features may include the ligand fingerprint and ligand descriptors.

[0085] In step 612, the machine learning model for DTI prediction is applied to the features of the query ligand and the features of the query protein to obtain a numerical score of the likelihood of interaction between the query ligand and the query protein.

[0086] The following paragraphs further illustrate embodiments of the present disclosure based on various examples. Those skilled in the art will understand that the present disclosure is not limited to these examples.

[0087] (i) Sample phantom ligand:

[0088] Go to Figure 7A, shows an example (700) for generating phantom ligands. Three hypothetical protein structures are shown (top row). Two of the three hypothetical protein structures actually interact with the ligand (top row, left and middle columns). The middle row shows various structural arrangements of the three hypothetical proteins. As a result of the structural arrangements, the ligand can project onto other proteins. Based on the similarity of the binding sites, confidence scores are assigned. The confidence score for an actual ligand-protein pairing is 1.0, but the confidence score for a phantom ligand-protein pairing is lower. The bottom row shows the generated phantom ligand-protein pairings as they may be stored in the phantom ligand database.

[0089] Figure 7B Shows an illustration of a concentric shell model (750) for obtaining binding site features according to one or more embodiments. Concentric shells with increasing radii (r) surround a central chemical structure that is considered part of the binding site. The inner shell mainly captures local features near the binding site, while the outer shell captures more and more global features. Features representing the protein can be based on the concentric shell model, thereby capturing local and global features of the protein without specifying the exact 3D geometry (e.g., at the atomic level).

[0090] (ii) Sample confidence vector:

[0091] Embodiments of the present disclosure rely on heuristic processes to increase drug-target interaction (DTI) data using hypothesized three-dimensional structure representations. These hypothesized DTI representations can provide information-rich features for machine learning, improving models aimed at predicting protein-ligand interactions. Obtaining these approximate DTI representations for any given DTI data point requires several assumptions outlined in Figure 2 Steps 200 and 202. For example, the three-dimensional protein structure used to represent DTI can come from a homology model rather than directly from experimental coordinates.

[0092] The confidence vector consists of multiple metrics that describe the measurable uncertainty in the approximate DTI representation. These metrics are accumulated in the creation of the phantom ligand database (step 200) and the projection of known DTI data onto the phantom ligand database (step 202). In one example, the confidence vector contains four elements, including: (1) the percentage of sequence identity between the homology model representing the DTI protein and its source template, (2) the RMSD from the alignment between the source structure of the phantom ligand and the homology model representing the DTI protein, (3) the Tanimoto similarity between the morgan3 fingerprint of the DTI ligand and the phantom ligand template, and (4) the confidence of the DTI data point.

[0093] In this example, the Drug-Target Interaction (DTI) database indicates that the ligand gefitinib interacts with the protein Aurora kinase A. The DTI database assigns an 85% probability to the interaction based on the accuracy of the source biophysical experiments. There is no specific interaction between gefitinib and Aurora kinase A in the source 3D structure database. When creating the phantom ligand database, a homology model of the Aurora kinase A protein was created from the closely homologous Aurora kinase B with 72.5% sequence identity. The molecule closest to gefitinib successfully mapped onto the homology model is erlotinib, which has a Morgan3 fingerprint Tanimoto similarity of 0.372. The erlotinib phantom ligand position was approximated based on a structural alignment between the Aurora kinase A homology model and the erlotinib-EGFR complex crystal structure, with a ligand binding site RMSD of Therefore, the corresponding confidence vector will be:

[0094] (iii) Sample training data and negative random permutations:

[0095] The method described focuses on augmenting Drug-Target Interaction (DTI) pairs from the DTI database with a mixture of relevant features obtained through a series of deterministic mapping relationships and heuristic modeling features (local structural features). Each row in the DTI database can be converted into a feature vector, as Figure 8 illustrated, showing the generation of training data (800) with ligand features from the corresponding drug (the column labeled "ligand features") and protein features (the columns labeled "global features", "functional features", and "local features"). The global and functional features of a protein can be retrieved from any protein using standard practices of database lookup and protein identifier mapping. The local protein features can be the result of the heuristic definition process outlined in this patent. They are modeled and thus may not be accurate. Each data row also has a corresponding confidence vector (described above but not shown in the figure), which can be used to imply hard truncation or weights for training a machine learning model.

[0096] When a neural network is trained only on true drug-target interaction positive instances extracted from a drug-target interaction dataset, the model may learn to ignore the core and obvious patterns of the interaction because they do not provide any signal to the model. Additionally, it may be necessary to control for potential significant biases towards highly representative drugs and targets in the drug-target interaction dataset. Therefore, it may be beneficial to sample negative instances using the relative proportions of each drug and target. As a result, the model can learn patterns based on both positive and negative instances. In Figure 8In this, randomized negative results permute the ligand component (white) of the eigenvector with the protein components (three gray shades) of the eigenvector. The resulting negative instances can be used to train a classification engine to balance the presence of individual ligand or protein features. Equal use of individual ligand and / or protein features in the positive and negative sets avoids network learning that any single feature is typically particularly associated with binding.

[0097] Embodiments of the present disclosure use phantom ligands to create local protein features for protein chemometrics (PCM) from a protein-ligand data set. More specifically, drug-target interaction (DTI) data is threaded onto a 3D atomic model of a protein-ligand complex to derive local protein features. The hybrid feature data set for PCM can include local (pocket), regional (domain), and global (whole protein) annotations and / or functional annotations.

[0098] Traditionally, the training data for machine learning should be high-confidence'model-quality data'. According to one or more embodiments, it may seem counterintuitive to use predictions (phantom ligands + threading) to generate training data for machine learning algorithms. Specifically, if the heuristics are not accurate enough, then, according to conventional wisdom, introducing local features derived from the combination of phantom ligands and threading has the potential to introduce additional noise, thereby degrading the performance of traditional DTIPMC. However, as Figure 9 shown by the performance comparison (900), performance improvement is achieved by introducing local features derived by the method described in this patent. Omitting the local features derived by this method is equivalent to the performance achieved by DTIPMC alone.

[0099] Specifically, Figure 9 shows a performance comparison of ranking the binding likelihood of small molecule ligands to 8717 proteins. To test the ranking, 100 molecules were randomly removed from the training data and used for testing. The figure plots the predicted ranking of the known interactions of these 100 random drugs. For example, in the absence of local features, only about 63% of the actual interactions were observed in the prediction of the top 300 proteins (top ~3.5%) in 8717 (dashed line). Including the local features estimated by this procedure, for the same threshold (solid line), the discovery rate increased to about 75%.

[0100] Various embodiments of the present disclosure have one or more of the following advantages. Embodiments of the present invention are capable of predicting drug-target interactions (DTI) using a machine learning model that reflects the compatibility between the protein environment and ligand properties. Localized 3D features are created to represent the binding site, even when no 3D information is available for the interaction under consideration.

[0101] Mapping known drug-target interactions to homology models comprehensively enhances rich DTI training data, which has high-dimensional biophysical information, to train deep neural networks. Thus, the method can use the comprehensive DTI database for entries in the DTI database even when the binding location of the drug to the protein is not necessarily known.

[0102] The method according to one or more embodiments does not require detailed knowledge of the biophysics of protein-ligand interactions. Thus, precise 3D coordinates of atoms are not required, enabling mapping of drug-target interactions to protein pockets using the DTI database and homology models.

[0103] Compared to structure-based deep learning methods that rely on 3D atomic coordinates, embodiments of the present disclosure require a reduced feature space and allow for much larger training data. Further, embodiments of the present disclosure are found to generalize well. Preliminary performance evaluations indicate that the method described above is approximately 1,000,000 times faster than docking simulations. The method according to one or more embodiments does not require human intervention. Specifically, the most likely protein representation and binding site are automatically identified. As discussed in Appendices A and B, the method according to one or more embodiments can be used as an accurate in silico alternative or addition to other in silico and / or experimental methods for predicting drug-target interactions.

[0104] Embodiments of the present disclosure can have various applications. For example, embodiments can be used for proteome screening (e.g., performing toxicity prediction or phenotype deconvolution prediction), virtual screening, and general drug discovery and development.

[0105] Embodiments of the present disclosure can be implemented on a computing system. Any combination of mobile, desktop, server, router, switch, embedded device, or other types of hardware can be used. For example, as shown in Figure 10A a computing system (1000) can include one or more computer processors (1002), non-persistent storage devices (1004) (e.g., volatile memory such as random access memory (RAM), cache memory), persistent storage devices (1006) (e.g., hard disk, optical disk drive such as a compact disc (CD) drive or digital versatile disc (DVD) drive, flash memory, etc.), communication interfaces (1012) (e.g., Bluetooth interface, infrared interface, network interface, optical interface, etc.), and many other elements and functions.

[0106] The computer processor(s) (1002) may be an integrated circuit for processing instructions. For example, the computer processor(s) may be one or more cores or micro-cores of a processor. The computing system (1000) may also include one or more input devices (1010), such as a touch screen, keyboard, mouse, microphone, touchpad, electronic pen, or any other type of input device.

[0107] The communication interface (1012) may include an integrated circuit for connecting the computing system (1000) to a network (not shown) (e.g., a local area network (LAN), a wide area network (WAN) such as the Internet, a mobile network, or any other type of network) and / or another device, such as another computing device.

[0108] In addition, the computing system (1000) may include one or more output devices (1008), such as a screen (e.g., a liquid crystal display (LCD), a plasma display, a touch screen, a cathode ray tube (CRT) monitor, a projector, or other display device), a printer, an external storage device, or any other output device. One or more of the output devices may be the same as or different from the input device(s). The input and output device(s) may be connected locally or remotely to the computer processor(s) (1002), the non-persistent storage device(s) (1004), and the persistent storage device(s) (1006). There are many different types of computing systems, and the aforementioned input and output device(s) may take other forms.

[0109] Software instructions in the form of computer-readable program code for executing embodiments of the present disclosure may be stored in whole or in part, temporarily or permanently, on a non-transitory computer-readable medium (such as a CD, DVD, storage device, disk, tape, flash memory, physical memory, or any other computer-readable storage medium). Specifically, the software instructions may correspond to a computer-readable program code that, when executed by (one or more) processors, is configured to execute one or more embodiments of the present disclosure.

[0110] Figure 10A The computing system (1000) in may be connected to or part of a network. Figure 10B As shown in FIG. 1 , the network (1020) may include multiple nodes (e.g., node X (1022), node Y (1024)). Each node may correspond to a computing system, such as in Figure 10A The computing system or a group of nodes shown in the figure may correspond to Figure 10AThe computing system shown in. For example, embodiments of the present disclosure may be implemented on a node of a distributed system connected to other nodes. As another example, embodiments of the present disclosure may be implemented on a distributed computing system having multiple nodes, where each part of the present disclosure may be located on a different node within the distributed computing system. In addition, one or more elements of the foregoing computing system (1000) may be located at a remote location and connected to other elements via a network.

[0111] Although not shown in Figure 10B the nodes may correspond to blades in a server chassis connected to other nodes via a backplane. As another example, the nodes may correspond to servers in a data center. As another example, a node may correspond to a computer processor or a microkernel of a computer processor having shared memory and / or resources.

[0112] Nodes (e.g., node X (1022), node Y (1024)) in the network (1020) may be configured to provide services to client devices (1026). For example, the nodes may be part of a cloud computing system. The nodes may include functionality for receiving requests (1026) from client devices and transmitting responses (1026) to client devices. The client device (1026) may be a computing system, such as the computing system shown in Figure 10A In addition, the client device (1026) may include and / or execute all or a part of one or more embodiments of the present disclosure.

[0113] In Figure 10A and Figure 10B the computing system or group of computing systems described may include functionality for performing various operations disclosed herein. For example, the (one or more) computing systems may perform communication between processes on the same or different systems. A variety of mechanisms, using some form of active or passive communication, may facilitate data exchange between processes on the same device. Examples representative of such inter-process communication include, but are not limited to, the implementation of files, signals, sockets, message queues, pipelines, semaphores, shared memory, message passing, and memory-mapped files. Additional details related to several of these non-limiting examples are provided below.

[0114] Based on the client-server network model, a socket can be used as an interface or a communication channel endpoint to enable two-way data transfer between processes on the same device. First, following the client-server network model, a server process (e.g., a process that provides data) can create a first socket object. Next, the server process binds the first socket object, thereby associating the first socket object with a unique name and / or address. After creating and binding the first socket object, the server process then waits and listens for incoming connection requests from one or more client processes (e.g., processes that seek data). At this point, when a client process wishes to obtain data from the server process, the client process starts by creating a second socket object. Then, the client process continues to generate a connection request that includes at least the second socket object and the unique name and / or address associated with the first socket object. Then, the client process transmits the connection request to the server process. Depending on availability, the server process can accept the connection request, establish a communication channel with the client process, or a server process that is busy processing other operations can queue the connection request in a buffer until the server process is ready. The established connection notifies the client process that communication can begin. In response, the client process can generate a data request specifying the data that the client process wishes to obtain. The data request is then transmitted to the server process. Upon receiving the data request, the server process analyzes the request and collects the requested data. Finally, the server process then generates a reply that includes at least the requested data and transmits the reply to the client process. More commonly, data can be transferred as datagrams or character streams (e.g., bytes).

[0115] Shared memory refers to the allocation of virtual memory space to enable a mechanism by which data can be communicated and / or accessed by multiple processes. When implementing shared memory, an initialization process first creates a sharable segment in a persistent or non-persistent storage device. After creation, the initialization process then loads the sharable segment and subsequently maps the sharable segment into the address space associated with the initialization process. After loading, the initialization process continues to identify one or more authorized processes and grant them access rights. The one or more authorized processes can also write data to and read data from the sharable segment. Changes made to the data in the sharable segment by one process can immediately affect other processes that are also linked to the sharable segment. In addition, when one of the authorized processes accesses the sharable segment, the sharable segment is mapped into the address space of the authorized process. Typically, at any given time, in addition to the initialization process, only one authorized process can load the sharable segment.

[0116] Without departing from the scope of the present disclosure, other techniques may be used to share data between processes, such as the various data described in this application. These processes may be part of the same or different applications and may be executed on the same or different computing systems.

[0117] Instead of or in addition to sharing data between processes, a computing system implementing one or more embodiments of the present disclosure may include functionality to receive data from a user. For example, in one or more embodiments, a user may submit data via a graphical user interface (GUI) on a user device. Data may be submitted via the graphical user interface by the user selecting one or more graphical user interface widgets or inserting text and other data into the graphical user interface widgets using a touchpad, keyboard, mouse, or any other input device. In response to selecting a particular item, information about the particular item may be obtained by a computer processor from a persistent or non-persistent storage device. When the user selects an item, the content of the obtained data about the particular item may be displayed on the user device in response to the user's selection.

[0118] As another example, a request to obtain data about a particular item may be sent to a server operatively connected to the user device via a network. For example, a user may select a uniform resource locator (URL) link within a web client of the user device, thereby initiating a hypertext transfer protocol (HTTP) or other protocol request to be sent to the web host associated with the URL. In response to the request, the server may extract data about the particular selected item and send the data to the device that initiated the request. Once the user device has received data about a particular item, the content of the received data about the particular item may be displayed on the user device in response to the user's selection. Continuing with the above example, the data received from the server after selection of the URL link may provide a web page in hypertext markup language (HTML) that can be rendered by the web client and displayed on the user device.

[0119] Once data is obtained, such as by using the techniques described above or from a storage device, a computing system, when implementing one or more embodiments of the present disclosure, may extract one or more data items from the obtained data. For example, it may be by Figure 10AThe computing system in [[ID=]] performs the extraction as described below. First, the organization pattern of the data (e.g., syntax, pattern, layout) is determined, which can be based on one or more of the following: location (e.g., bit or column location, the Nth token in a data stream, etc.); property (where a property is associated with one or more values); or hierarchical / tree structure (including node levels of different levels of detail, such as in nested packet headers or nested document sections). Then, in the context of the organization pattern, the original, unprocessed data symbol stream is parsed into a token stream (or hierarchical structure) (where each token can have an associated token "type").

[0120] Next, extraction criteria are used to extract one or more data items from the token stream or structure, where the extraction criteria are processed according to the organization pattern to extract one or more tokens (or nodes from the hierarchical structure). For location-based data, the (one or more) tokens at the (one or more) locations identified by the extraction criteria are extracted. For property / value-based data, the (one or more) tokens and / or (one or more) nodes associated with the (one or more) properties that satisfy the extraction criteria are extracted. For hierarchical / hierarchical data, the (one or more) tokens associated with the (one or more) nodes that match the extraction criteria are extracted. The extraction criteria can be as simple as an identifier string or can be a query provided to a structured data repository (where the data repository can be organized according to a database schema or data format, such as XML).

[0121] The extracted data can be used for further processing by the computing system. For example, Figure 10AWhen the computing system executes one or more embodiments of the present disclosure, it may perform data comparison. Data comparison can be used to compare two or more data values (e.g., A, B). For example, one or more embodiments can determine whether A > B, A = B, A != B, A < B, etc. The comparison can be performed by submitting operation codes A, B that specify operations related to the comparison to an arithmetic logic unit (ALU) (i.e., a circuit that performs arithmetic and / or bitwise logical operations on two data values). The ALU outputs the numerical result of the operation and / or one or more status flags related to the numerical result. For example, the status flag can indicate whether the numerical result is positive, negative, zero, etc. By selecting the appropriate operation codes and then reading the numerical result and / or status flags, the comparison can be performed. For example, to determine whether A > B, B can be subtracted from A (i.e., A - B), and the status flag can be read to determine whether the result is positive (i.e., if A > B, then A - B > 0). In one or more embodiments, as determined using the ALU, if A = B or if A > B, then B can be regarded as a threshold, and A is regarded as meeting the threshold. In one or more embodiments of the present disclosure, A and B can be vectors, and comparing A with B requires comparing the first element of vector A with the first element of vector B, comparing the second element of vector A with the second element of vector B, etc. In one or more embodiments, if A and B are strings, the binary values of the strings can be compared.

[0122] Figure 10A The computing system in can implement and / or be connected to a data repository. For example, one type of data repository is a database. A database is a collection of information configured to facilitate data retrieval, modification, reorganization, and deletion. A database management system (DBMS) is a software application that provides an interface for a user to define, create, query, update, or manage a database.

[0123] A user or software application can submit a statement or query to the DBMS. Then the DBMS interprets the statement. The statement can be a select statement, update statement, create statement, delete statement, etc. that requests information. Moreover, the statement can include parameters that specify data or data containers (database, table, record, column, view, etc.), (one or more) identifiers, conditions (comparison operators), functions (e.g., join, full join, count, average, etc.), sorting (such as ascending, descending), or others. The DBMS can execute the statement. For example, the DBMS can access memory buffers, reference, or index files for reading, writing, deleting, or any combination thereof in response to the statement. The DBMS can load data from persistent or non-persistent storage devices and perform calculations in response to a query. The DBMS can return the (one or more) results to the user or software application.

[0124] Figure 10A A computing system can include functions that provide raw and / or processed data, such as the results of comparisons and other processing. For example, providing data can be accomplished through various presentation methods. Specifically, data can be provided through a user interface provided by a computing device. The user interface can include a GUI that displays information on a display device (a computer monitor or a touch screen on a handheld computing device). The GUI can include various GUI widgets that organize what data is shown and how the data is provided to the user. In addition, the GUI can provide data directly to the user, for example, data provided as an actual data value through text, or a visual representation of the data presented by the computing device, such as through a visualization data model.

[0125] For example, the GUI can first obtain a notification from a software application requesting the provision of a specific data object within the GUI. Next, the GUI can identify the data object type associated with the specific data object, for example, by obtaining data from a data attribute within the data object that identifies the data object type. Then, the GUI can determine any rules specified for displaying the data object type, for example, rules specified by a software framework for a data object class or rules specified according to any local parameters defined by the GUI for presenting the data object type. Finally, the GUI can obtain data values from the specific data object and present a visual representation of the data values within the display device according to the specified rules for the data object type.

[0126] Data can also be provided through various audio methods. In particular, the data can be presented in an audio format and provided as sound through one or more speakers operatively connected to the computing device.

[0127] Data can also be provided to the user through tactile methods. For example, tactile methods can include vibrations or other physical signals generated by the computing system. For example, vibrations generated by a handheld computing device with a predetermined duration and vibration intensity can be used to provide data to the user to convey the data.

[0128] The above description of the functions only presents several examples of the functions performed by Figure 10A the computing system and / or Figure 10B the client device in

[0129] Although the present disclosure has been described with respect to a limited number of embodiments, those skilled in the art who benefit from the present disclosure will understand that other embodiments can be designed without departing from the scope of the present disclosure as disclosed herein. Therefore, the scope of the present disclosure should be defined only by the appended claims.

[0130] The embodiments and examples set forth herein are for the purpose of best explaining the present invention and its specific applications, so that those skilled in the art can make and use the present invention. However, those skilled in the art will recognize that the foregoing description and examples are presented for purposes of illustration and example only. The description set forth is not intended to be exhaustive or to limit the invention to the precise form disclosed.

[0131] Although the present invention has been described with respect to a limited number of embodiments, those skilled in the art, having the benefit of this disclosure, will understand that other embodiments can be designed that do not depart from the scope of the invention disclosed herein. Accordingly, the scope of the invention should be limited only by the appended claims.

Claims

1. A method for predicting drug-target binding using synthetic enhanced data, the method comprising: Generate multiple phantom ligands for multiple proteins in a protein structure database; Use the multiple phantom ligands to generate multiple drug-target interaction (DTI) features for proteins and ligands in a DTI database; Generate a machine learning model using the multiple DTI features; and Use the machine learning model to predict the likelihood of interaction for a combination of a query protein and a query ligand.

2. The method according to claim 1, wherein generating the plurality of phantom ligands comprises: For a protein cluster selected from the multiple proteins: Perform a structural alignment of the proteins in the protein cluster; After the structural alignment, obtain the multiple phantom ligands by projecting the ligand of one of the proteins in the cluster onto all other proteins in the cluster; For each of the multiple phantom ligands, obtain a confidence score.

3. The method according to claim 2, wherein the protein clustering is obtained based on one selected from the group consisting of: similarity of sequences, similarity of three-dimensional topological structures, and existing clustering in the database.

4. The method according to claim 2, wherein the confidence score quantifies the uncertainty of the associated phantom ligand.

5. The method according to claim 1, wherein generating the plurality of DTI features comprises: For each of multiple combinations of ligands and proteins in the DTI database: Select the phantom ligand most similar to the ligand considered for the combination from the multiple phantom ligands; and Generate features for the protein considered for the combination, where the generated features characterize the protein considered for the combination.

6. The method according to claim 5, wherein the generated features include one selected from the group consisting of: at least one local feature, which includes binding site features in concentric shells with increasing radii, at least one global feature other than the binding site features, and at least one functional annotation.

7. The method according to claim 5, wherein the selection of the most similar phantom ligand is performed based on a distance metric.

8. The method according to claim 5, wherein generating the plurality of DTI features further comprises: Obtain a confidence vector representing the confidence in multiple components of the DTI feature associated with the most similar phantom ligand.

9. The method according to claim 8, wherein the confidence of the plurality of components includes at least one selected from the group consisting of: a first confidence score quantifying the uncertainty associated with the most similar phantom ligand, A second confidence score that quantifies the fingerprint similarity between the ligand used for the combination and the most similar phantom ligand, and A third confidence score that depends on the source from which the DTI features are obtained.

10. The method according to claim 1, wherein generating the machine learning model comprises: Obtain positive training samples based on the multiple DTI features of proteins and ligands; Obtain negative training samples based on the multiple DTI features by performing at least one random permutation of the multiple DTI features of proteins and ligands; Use the positive training samples and the negative training samples to train the machine learning model for DTI prediction.

11. The method according to claim 10, wherein generating the machine learning model comprises, before obtaining the positive training samples and the negative training samples: Using a confidence threshold applied to a confidence vector associated with the plurality of DTI features to filter the plurality of DTI features of the proteins and ligands.

12. The method according to claim 1, wherein the machine learning model is one selected from a classifier model and a regression model.

13. The method according to claim 1, wherein predicting the likelihood of interaction of the combination of the query protein and the query ligand comprises: Obtain possible binding sites and related local features of a query protein based on the multiple phantom ligands; Generate features of the query protein, where the features of the query protein include the local features; Generate features of the query ligand, where the features of the query ligand include a ligand fingerprint and ligand descriptors; and Apply the machine learning model to the features of the query protein and the features of the query ligand to obtain the likelihood of interaction between the query ligand and the query protein.

14. The method according to claim 13, wherein the features of the query protein further comprise at least one selected from the group consisting of global features and functional annotations.

15. A non-transitory computer-readable medium comprising computer-readable program code for predicting drug-target binding using integrated enhanced data, the computer-readable program code causing a computer system to: Generate a plurality of phantom ligands for a plurality of proteins in a protein structure database; Use the plurality of phantom ligands to generate a plurality of DTI features for proteins and ligands in a drug-target interaction (DTI) database; Generate a machine learning model using the DTI features; and Use the machine learning model to predict the likelihood of interaction for a combination of a query protein and a query ligand.

16. A system for differential drug discovery, the system comprising: A protein structure database; A phantom ligand identification engine configured to generate a plurality of phantom ligands for a plurality of proteins in the protein structure database; Store a phantom ligand database of the multiple phantom ligands; Store a drug-target interaction (DTI) database of proteins and ligands; A feature generation engine configured to generate multiple DTI features for proteins and ligands in the DTI database using the multiple phantom ligands in the phantom ligand database; A machine learning model training engine configured to generate a machine learning model using the DTI features; and A DTI prediction engine configured to use the machine learning model to predict the likelihood of interaction for a combination of a query protein and a query ligand.

17. The system according to claim 16, wherein generating the plurality of DTI features comprises: For each of multiple combinations of ligands and proteins in the DTI database: Select the phantom ligand most similar to the ligand considered for the combination from the multiple phantom ligands; and To consider protein generation features for the combination, where the generated features characterize the protein considered for the combination.

18. The method according to claim 17, wherein the generated features comprise one selected from the group consisting of: at least one local feature, which comprises binding site features in concentric shells with increasing radii, at least one global feature in addition to the binding site features, and at least one functional annotation.

19. The system according to claim 17, wherein generating the plurality of DTI features further comprises: Obtain a confidence vector representing the confidence in multiple components of the DTI features associated with the most similar phantom ligand, where the confidence in the multiple components includes at least one selected from the group consisting of: a first confidence score quantifying the uncertainty associated with the most similar phantom ligand, a second confidence score quantifying the fingerprint similarity between the ligand considered for the combination and the most similar phantom ligand, and a third confidence score depending on the source from which the DTI features are obtained.

20. The system according to claim 16, wherein predicting the likelihood of interaction of the combination of the query protein and the query ligand comprises: Obtain possible binding sites and associated local features of a query protein based on the multiple phantom ligands; Generate features of the query protein, the features of the query protein including the local features; Generate features of the query ligand, the features of the query ligand including a ligand fingerprint and a ligand descriptor; and Apply the machine learning model to the features of the query protein and the features of the query ligand to obtain the likelihood of interaction between the query ligand and the query protein.

Images