22_3_CFA.pptx

Target Druggability What is druggability? Importance in drug discovery Overview of different methods Precedence-based Structure-based Future-based Beyond small-molecules The druggable genome Applications in target prioritisation What is druggability? Definitions of ’druggability’ vary: The ability o...

Target Druggability What is druggability? Importance in drug discovery Overview of different methods Precedence-based Structure-based Future-based Beyond small-molecules The druggable genome Applications in target prioritisation What is druggability? Definitions of ’druggability’ vary: The ability of a protein target to bind small molecules with high affinity -sometimes (perhaps more appropriately) called ‘ligandability’ Easier to assess The ability of a protein to be modulated by a drug-like small molecule More predictive of success The likelihood of finding orally bioavailable small molecules that bind to a particular target in a disease-modfying way’ -includes consideration of likely pharmacokinetic and pharmacodynamic properties of compounds Lipinski’s Ruole of Five and drug-likeness Orally bioavailable (drug-like) small molecules tend to have properties within certain parameters Lipinski observed that poor absorption and permeability was likely if more than one of these rules is violated: Molecular Weight < 500 Da LogP (octanol/water partition coefficient) < 5 H-bond Acceptors (N + O) < 10 H-bond Donors (NH + OH) < 5 More recent quantitative methods e.g. Quantitative Estimate of Druglikeness (QED) and other descriptors e.g., rotatable bonds, polar surface area Beautiful binding sites In order to bind drug-like compounds, a protein should have a binding site with complementary properties e.g., Appropriate size to accommodate a drug-like ligand Buried, to increase interaction surface Not too polar Large exposed polar sites can be considered less druggable than smaller, more buried, hydrophobic pockets Beautiful binding sites Importance of assessing druggability Omics techniques are producing large amounts of data relating genese/proteins to disease e.g., Sequencing/GWAS studies to identify mutations Expression studies in diseased vs normal tissue Proteomics/metabolomics studies to find biomarkers Estimates suggest that around 15% of human genome may be druggable (with small molecule approach) disease-linked proteins druggable proteins Drug targets Importance of assessing druggability Important to prioritise potential targets and pursue those that are most likely to be amenable to the desired approach Choosing a target that doesn’t bind small molecules is likely to result in failure of screening experiments Choosing a target that doesn’t bind ‘drug-like’ small molecules may lead to later (even more costly) failure due to poor pharmacokinetic properties of compounds >60% projects may fall at lead identification/optimisation stages Methods for assessing druggability Protein structure contains drug-like pockets Structural analysis Protein Data Bank Predicted druggable based on sequence features Sequence analysis e.g. Machinelearning algorithms Bind endogenous drug-like ligands Metabolite/ ligand databases ChEMI, HMDB, KEGG, IUPHAR High affinity drug-like compounds available Pharmacology databases ChEMBL PubChem BindingDB Protein is an established small molecule Clinical trial Drug databases DrugBank databases ClinicalTrials TTD DailyMed.gov ChEMBL Compounds in clinical trials for the protein Increasing confidence in druggability Methods for assessing druggability – Precedence-based Protein structure contains drug-like pockets Structural analysis Protein Data Bank Predicted druggable based on sequence features Sequence analysis e.g. Machinelearning algorithms Bind endogenous drug-like ligands Metabolite/ ligand databases ChEMI, HMDB, KEGG, IUPHAR High affinity drug-like compounds available Pharmacology databases ChEMBL PubChem BindingDB Protein is an established small molecule Clinical trial Drug databases DrugBank databases ClinicalTrials TTD DailyMed.gov ChEMBL Compounds in clinical trials for the protein Increasing confidence in druggability Precedence-based assessment If a protein is the target of an approved small molecule drug this gives a very high degree of confidence in druggability Also implies that targeting this protein is (relatively) safe (depending on indication/administration route) Caveats – still not an absolute guarantee of success for a different disease/product profile e.g. agonist vs antagonist CNS-penetrant vs systemic Long vs short-acting acceptability of side effects ChEMBL is a manually curated database of bioactive molecules with drug-like properties. It brings together chemical, bioactivity and genomic data to aid the translation of genomic information into effective new drugs. Assigning approved drug targets Non-trivial in many cases Different rules for assignment Assign based on binding affinity –e.e. all targets likely to interact at physiologically relevant concentration? Assign based on compound selectivity – e.g., include all members of a family if compound non-selective? Assign based on approved indication – e.g., include only if evidence that the target is responsible for the efficacy in that disease? Assign based on expression data – e.g., identify the targets that are expressed in relevant tissue? Different treatment of protein complexes Include only binding subunit? include all individual proteins? Examples Tiotropium (chronic osbtructive pulmonary disease): Non selective muscarinic antagonist There are 5 muscarinic receptor subtypes (M1-M5) Topical agent - inhaled powder M3 receptor expressed in the lung and causes contraction of smooth muscle, leading to bronchoconstriction Benzodiazepines (e.g., insomnia): Positive allosteric modulators of GABA-A receptors GABA-A receptors are heteropentameric complexes 6 alpha subunits, 3 beta subunits, 3 gamma subunits Binding of GABA requires alpha & beta subunits (binds at interface) Benzodiazepines bind at alpha/gamma interface in alpha1,2,3 or 5-containing receptors FDA approved drugs in the ChEMBL Manually curated efficacy targets for FDA approved drugs and WHO antimalarials Targets with which drug interacts directly Targets responsible for efficacy in approved indication NOT targets responsible for adverse-effects or non-approved indications NOT targets assigned purely on basis of pharmacology data Drug type (small molecule, antibody etc), action type (e.g. agonist, antagonist), and binding site/subunit information where available Deal with non-specific drugs and targets that are protein complexes ChEMBL targets Clinical candidates If a target has compounds that have reached the clinic (e.g. phase I/II/III trials) this also provides good confidence in druggability Particularly, late-phase development candidates are likely to have shown a degreee of safety and an adequate pharmacokinetic profile But many databases providing candidate information are commercial (and expensive) Candidate target information Often difficult to identify targets for clinical candidates – may require searching of literature, complany pipeline information etc Could use bioactivity data to identify potential targets USAN stems may also give clues e.g., ‘-panel’ = AMPA receptor antagonists, ‘-kiren’ = renin inhibitors Methods for assessing druggability: Ligand-based Protein structure contains drug-like pockets Structural analysis Protein Data Bank Predicted druggable based on sequence features Sequence analysis e.g. Machinelearning algorithms Bind endogenous drug-like ligands Metabolite/ ligand databases ChEMI, HMDB, KEGG, IUPHAR High affinity drug-like compounds available Pharmacology databases ChEMBL PubChem BindingDB Protein is an established small molecule Clinical trial Drug databases DrugBank databases ClinicalTrials TTD DailyMed.gov ChEMBL Compounds in clinical trials for the protein Increasing confidence in druggability Pharmacology data A number of public resources exist that capture pharmacology data e.g., ChEMBL - https://www.ebi.ac.uk/chembl/ PubChem - https://pubmed.ncbi.nlm.nih.gov/ Bindin DB - https://www.bindingdb.org/rwd/bind/index.jsp Guide to Pharmacology - https://www.guidetopharmacology.org/ Existence of compounds that bind with high-affinity to the protein implies it is druggable But important to consider whether the compounds are ‘drug-like’, also whether there are any selectivity issues Example Endogenous ligands Even in the absence of ‘med-chem’ compounds, knowledge of the endogenous ligand/substrate of a protein can be useful in assessing druggability Databases containing ligand information can be used to identify proteins that contain small-molecule binding sites Guide to Pharmacology: https://www.guidetopharmacology.org/ PDBe (crystal structures): https://www.ebi.ac.uk/pdbe/ Proteins whose endogenous ligands are peptides/proteins (e.g., proteases), or which are involved only in protein/protein interactions are less likely to be druggable Methods for assessing druggability: Structure-based Protein structure contains drug-like pockets Structural analysis Protein Data Bank Predicted druggable based on sequence features Sequence analysis e.g. Machinelearning algorithms Bind endogenous drug-like ligands Metabolite/ ligand databases ChEMI, HMDB, KEGG, IUPHAR High affinity drug-like compounds available Pharmacology databases ChEMBL PubChem BindingDB Protein is an established small molecule Clinical trial Drug databases DrugBank databases ClinicalTrials TTD DailyMed.gov ChEMBL Compounds in clinical trials for the protein Increasing confidence in druggability Structure-based prediction methods Rely on identifying cativities in protein crystal structures and assessing the properties of these cavities to predict whether they may bind druglike molecules Rules for properties that indicate a druggable cativity are learnt from analysis of co-crystal complexes with drug-like ligands e.g., volume, surface area, polar surface area, burial etc These rules can then be applied to new (apo)protein structures to predict/score druggability Algorithms PocketFinder Trained on a set of 5616 binding sites from PDB Tested on 11510 corresponding apo-protein structures – high agreement with liganded sites Known ligand-binding site is largest site 80% time (and in top 2 sites 92.7% of time) Average volume of binding sites 610.8A DogSiteScorer Undruggable had more short-range hydrophilic-hydrophilic interaction and less short-range lipophilic-lipophilic interactions Issues with ligand/strucutre-based methods Ligand/structure-based druggability assessment gives highest degree of confidence, but not useful for novel targets or those that don’t have crystal structures available Methods are needed to help prioritise the remainder of the proteins in the human genome (or indeed other genomes), some of which may be ‘druggable’ but have not yet been investigated Homology may help in some cases (e.g., member of known drug target family), but this is only based on ‘past success’ and not very useful in assessing novel families Methods for assessing druggability: sequence-based methods Protein structure contains drug-like pockets Structural analysis Protein Data Bank Predicted druggable based on sequence features Sequence analysis e.g. Machinelearning algorithms Bind endogenous drug-like ligands Metabolite/ ligand databases ChEMI, HMDB, KEGG, IUPHAR High affinity drug-like compounds available Pharmacology databases ChEMBL PubChem BindingDB Protein is an established small molecule Clinical trial Drug databases DrugBank databases ClinicalTrials TTD DailyMed.gov ChEMBL Compounds in clinical trials for the protein Increasing confidence in druggability Sequence and feature-based druggability Earlier definitions of druggable genome relied on identifying proteins with known drug-binding domains More recently curated list of known compound-binding domain based on ChEMBL database Sequence similarity/domain analysisi may help in some cases (e.g., member of known drug target family), but this is only based on ‘past success’ and not very useful in assessing novel families Various machine-learning/modelling methods can be used to identify more general ‘features’ of what makes a good drug target Since these are based only on aminoacid sequence, they can be applied to whole genomes Large numbers of different descriptrs can be calulated e.g. amino acid composition, lenght, hydrophobicity transmembrane domains, signal peptide, glycosylation sites secondary structure, domain composition subcellular localisation Beyond small molecule druggability Based on current estimates, the overlap between small-molecule druggable and disease-modifying targets may be relatively small Therefore, other approaches may be necessary to target proteins that do not have ‘beautiful’ small molecule binding sites: Inhibition of protein-protein interactions protein therapeutics SiRNA Protein therapeutic druggability The ability to target a protein with monoclonal antibodies/protein drugs depends largely on extra-cellular location e.g., protein should be segreted or membrane bound This can be determined in a variety of ways e.g. Precedence (e.g., known protein therapeutic drugs) Annotation/experimental evidence (e.g., known membrane protein, isolated from plasma) Predictive methods (e.g., transmembrane domains, signal peptide, subcellular localisation prediction algorithms) Protein-protein interactions Proteins generally participate in protein-protein interactions, even if they don’t have a small molecule binding site, so modulating this interaction may be a highly effective means of modulating a target However, protein-protein interaction interfaces are often large, flat surfaces (i.e., not beautiful binding sites) Targeting these is difficult, particularly with an oral small molecule drug, peptidomimetics more likely Surfaces may have ‘hot-spots’ that contribute most of the binding affinity, allowing smaller inhibitors to target these Many algorithms now being developed to predict such hot-spots and assess druggability The druggable genome Applications of druggable genome Understanding wheter a given target is druggable (and with what evidence) critical in target prioritisation Targets of FDA approved drugs ➤ drug repurposing opportunities Targets of cllinical candidates ➤ target validation, drug trial support and stratification Targets binding small molecules ➤ prioritisation/initiation/de-risking, discovery programs Extracellular proteins ➤ address important future role of mAbs Members of druggable families ➤ identification of novel tractable targets Esempio – drug repurposing 101 risk loci (42 novel) form 100000 subjects: 98 candidate genes Use of drug target information Authors looked at whether any of candidate proteins were targets for existing drugs (approved and experimental) 27 proteins were targets of approved drugs for RA (confirmed associations) Additional targets identified that tha d drugs for other indications – potential drug repurposing opportunities: e.g. CDK4 and CDK6 identified as candidate genes Palbociclib is an approved CDK4/6 inhibitor for breast cancer Understanding of mechanism of action and targets of existing drugs is therefore crucial in exploiting genetic association data Even if approved drugs/clinical candidates not already available for a target, knowledge of druggability important in prioritizing proteins for follow-up studies, identifying tool compounds etc