As many of you know I don't send Christmas cards, instead I give the monies I would have spent to MS Research. Have a great time and a successful New Year,

The majority of drug targets are proteins and knowledge of the 3D structure of the protein can be very helpful for structure based design. Whilst the PDB contains 227,933 structures there are still a number of structures that lack structural information. In 2018 Deepmind released AlphaFold an artificial Intelligence program design to predict protein 3D structure from the amino-acid sequence DOI. Since then there have a series of updates that have included the ability to handle small molecules, co-factors, nucleic acids, protein complexes etc. AlphaFold has been used in collaboration with the EBI to create AlphaFold DB which provides open access to over 200 million protein structures, covering the human proteome and the proteomes of 47 other key organisms important in research and global health. A recent addition is Foldseek a protein structural search program that allows users to search the AlphaFold Database.

David Baker, Demis Hassabis and John Jumper were awarded the 2024 Nobel Prize for Chemistry. One half of the prize has been awarded to David Baker “for computational protein design” and the other half jointly to Demis Hassabis and John M. Jumper “for protein structure prediction.”

Whilst AphaFold gets much of the publicity, it has served to spawn a number of related programs, comparison of the different options is difficult especially when looking at the various licensing options. Fortunately, Brian Naughton has posted a very useful summary. http://blog.booleanbiotech.com/alphafold3-boltz-chai1.html.

Three critical databases for drug discovery are now interlinked. The Cambridge Structural Database (CSD) a curated repository of small molecule crystal structures, ChEMBL a manually curated database of bioactive molecules with their associated biological data and PDBe a founding member of the Worldwide Protein Data Bank (wwPDB) which collects, organises and disseminates data on biological macromolecular structures, are now interlinked.

The BioChemGraph (BCG) project tackles the challenge of linking diverse data in biology by creating a resource that integrates data from the PDBe, ChEMBL, and the CSD. This has been achieved by mappings UniProt Accession ID and compound InChIKey, linkingmore than 17,000 experimentally determined protein-ligand complexes from the PDB to about 39,000 ChEMBL bioactivity records. By providing this link it is possible to not only identify binding affinity for the selected target but also much more information about the small molecule ligand, such as off-target activities, calculated physicochemical properties and also any ADME/T data that might be available. All data can be downloaded as a tsv as shown below.

ChEMBL and PDBe have collaborated to set up an automatic pipeline for generating these data. As a result, the data will be updated weekly, in sync with the PDBe release every Wednesday at 00.00 UTC.

InChis have also been used to interconnect with Cambridge Structural Database using UniChem, 235,000 CSD identifiers have been linked corresponding entries in UniChem, a “universal translator” for chemistry using InChIs to connect chemical structures and their identifiers across various databases. UniChem enables researchers to seamlessly access information about a specific molecule across a wide variety of data sources. There are currently 41 data sources (https://www.ebi.ac.uk/unichem/sources).

Protein protein interactions are always a challenge to optimise and it looks like the latest offering from Google DeepMind may be of significant help.

Protein binders that can bind tightly to a target protein are hard to design. Traditional methods are time intensive, requiring multiple rounds of extensive lab work. After the binders are created, they undergo additional experimental rounds to optimize binding affinity, so they bind tightly enough to be usefu

AlphaProteo generates novel proteins that bind to other proteins. Given the structure of a target molecule and a set of preferred binding locations on that molecule, AlphaProteo generates a candidate protein that binds to the target at those locations.

Whilst code is not available, note

If you’re a biologist, whose research could benefit from target-specific protein binding, and you’d like to register interest in being a trusted tester for AlphaProteo, please reach out to us on alphaproteo@google.com.

The term "privileged structures" was first coined by Ben Evans DOI: 10.1021/jm00120a002 who recognised the potential of certain regularly occurring structural motifs as templates for derivatization to discovery novel ligands for binding to proteins. In this seminal paper they identified a benzodiazepine and substituted indole as key structures in their work to yield CCK antagonists.

Two very popular privileged structures are N-benzyl piperidine and N-benzyl piperazine. They offer a variety of different interactions (pi-stacking, hydrophobic, electrostatic) with a relatively well defined 3D structure.

A recent publication gives a very nice summary of their use in drug discovery DOI.

Abstract The N-benzyl piperidine (N-BP) structural motif is commonly employed in drug discovery due to its structural flexibility and three-dimensional nature. Medicinal chemists frequently utilize the N-BP motif as a versatile tool to fine-tune both efficacy and physicochemical properties in drug development. It provides crucial cation-π interactions with the target protein and also serves as a platform for optimizing stereochemical aspects of potency and toxicity. This motif is found in numerous approved drugs and clinical/preclinical candidates. This review focuses on the applications of the N-BP motif in drug discovery campaigns, emphasizing its role in imparting medicinally relevant properties. We provide an overview of approved drugs, the clinical and preclinical pipeline, and discuss its utility for specific therapeutic targets and indications, along with potential challenges.

Many compounds can cause liver injury, after oral administration the first major organ exposed is the liver. The LiverTox is a database of information on the diagnosis, cause, frequency, patterns, and management of liver injury attributable to prescription and nonprescription medications, herbals and dietary supplements.

Checking for the potential to cause liver injury is an important part of the drug discovery process and there are a number of in vitro and in vivo assays that can be used.

High dose studies in safety species are undertaken to identify potential toxicities and to determine safety margins, Clinically, the most relevant reactions include liver necrosis, hepatitis, cholestasis, vascular changes and steatosis. A drug can cause liver toxicity via multiple mechanisms, it can be the result of a direct action of the parent compound or indirectly through reactive metabolites. The drug or its metabolites may cause liver toxicity after specific receptor binding, or reactive metabolites can react with hepatic macromolecules, all leading to direct cytotoxicity. In addition, Immune-mediated idiosyncratic drug reaction has been responsible for numerous serious hepatotoxic events in humans

It would be useful to be able to predict ahead of synthesis whether a molecule was likely to cause liver injury and that is the function of DILIpredictor DOI. Using data form several thousand molecules and a variety of different assays (both in vitro and in vivo) and different species the authors have developed a predictive model. The attraction of this approach is in addition to giving an early flag of potential DILI it also highlights potential species differences and can give an insight into the mechanism.

DILIPredictor required only chemical structures as input for prediction and is publicly available at https://broad.io/DILIPredictor for use via web interface (please don't submit confidential molecules) and with all code available for download from GitHub

https://github.com/srijitseal/DILI_Predictor

I installed like this since it currently does not run with the latest version of python

conda create -n DILIpred python=3.10
conda activate DILIpred

pip install DILIpred

It can then beinstalled as follows.

conda create -n DILIpred python=3.10
conda activate DILIpred

pip install DILIpred

It can then be usef as follows

(DILIpred) chrisswain@Mac-Studio ~ % dilipred -smiles "C[C@@H]1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C"

If you use DILIPred in your work, please cite: Improved Detection of Drug-Induced Liver Injury by Integrating Predicted In Vivo and In Vitro Data Srijit Seal, Dominic Williams, Layla Hosseini-Gerami, Manas Mahale, Anne E. Carpenter, Ola Spjuth, and Andreas Bender doi: https://doi.org/10.1021/acs.chemrestox.4c00015

100%███████████████████████████████████████████████████████████████ 1/1 [00:01<00:00, 1.14s/it] 2024-07-12 08:29:44.777 | CRITICAL | dilipred.main:predict:458 - The compound is predicted DILI-Positive

The detailed output is contained in a file created.

source,assaytype,description,value,pred,SHAP contribution to Toxicity,SHAP,smiles,smilesr DILI,DILIstFDA,This is the predicted FDA DILIst label,0.8187885151383966,1,N/A,N/A,C[C@@H]1C2=NN=C(N2CCN1C(=O)C3=CC=C(C= C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O)c2ccc(F)cc2)C3C)n1 Diverse DILI C,Heterogenous Data ,"Transient liver function abnormalities, adverse hepatic effects",0.7393781727510759,True,Positive,0.003926801602711443,C[C@@H] 1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O )c2ccc(F)cc2)C3C)n1 BESP,Mechanisms of Liver Toxicity,BESP Bile Salt Export Pump Inhibition,0.5655727513227511,True,Positive,0.0003070244326849143,C[C@@H ]1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(= O)c2ccc(F)cc2)C3C)n1 Mitotox,Mechanisms of Liver Toxicity,Mitotox ,0.10973983865879627,False,Positive,0.0006130947805277731,C[C@@H]1C2=NN= C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O)c2ccc( F)cc2)C3C)n1 Reactive Metabolite,Mechanisms of Liver Toxicity,Reactive Metabolite Formation,0.19967540492325553,False,Negative,-0.001048102741727997,C[C@@ H]1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C( =O)c2ccc(F)cc2)C3C)n1 Human hepatotoxicity,Human hepatotoxicity,"Human hepatotoxicity, hepatobiallry",0.7196576912119554,True,Positive,0.007802364907899422,C[C @@H]1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN( C(=O)c2ccc(F)cc2)C3C)n1 Animal hepatotoxicity A,Animal hepatotoxicity,"Rat, chronic oral administration, Hepatic histopathologic effects, ToxRefDB",0.5867747455286331,True,Positive,0.0030731971130978854,C[C@@H] 1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O )c2ccc(F)cc2)C3C)n1 Animal hepatotoxicity B,Animal hepatotoxicity,"Hepatocellular hypertrophy, rats, ORAD, HESS",0.6646590439473917,True,Positive,0.0013360236463188587,C[C@@H]1C2= NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O) c2ccc(F)cc2)C3C)n1 Preclinical hepatotoxicity,Animal hepatotoxicity,"Preclinical hepatotoxicity data from PharmaPendium, Leadscopre, and internal repository with 14- to 28-day rat study data",0.8576928962241468,True,Positive,0.011692057666492625,C[C@@H]1C2= NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O) c2ccc(F)cc2)C3C)n1 Diverse DILI A,Heterogenous Data ,Large-scale and diverse ddrug induced liver injury dataset,0.6324274304660036,True,Positive,0.003398315762493277,C[C@@H]1C2 =NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O) c2ccc(F)cc2)C3C)n1 source,assaytype,description,value,pred,SHAP contribution to Toxicity,SHAP,smiles,smilesr DILI,DILIstFDA,This is the predicted FDA DILIst label,0.8187885151383966,1,N/A,N/A,C[C@@H]1C2=NN=C(N2CCN1C(=O)C3=CC=C(C= C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O)c2ccc(F)cc2)C3C)n1 Diverse DILI C,Heterogenous Data ,"Transient liver function abnormalities, adverse hepatic effects",0.7393781727510759,True,Positive,0.003926801602711443,C[C@@H] 1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O )c2ccc(F)cc2)C3C)n1 BESP,Mechanisms of Liver Toxicity,BESP Bile Salt Export Pump Inhibition,0.5655727513227511,True,Positive,0.0003070244326849143,C[C@@H ]1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(= O)c2ccc(F)cc2)C3C)n1 Mitotox,Mechanisms of Liver Toxicity,Mitotox ,0.10973983865879627,False,Positive,0.0006130947805277731,C[C@@H]1C2=NN= C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O)c2ccc( F)cc2)C3C)n1 Reactive Metabolite,Mechanisms of Liver Toxicity,Reactive Metabolite Formation,0.19967540492325553,False,Negative,-0.001048102741727997,C[C@@ H]1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C( =O)c2ccc(F)cc2)C3C)n1 Human hepatotoxicity,Human hepatotoxicity,"Human hepatotoxicity, hepatobiallry",0.7196576912119554,True,Positive,0.007802364907899422,C[C @@H]1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN( C(=O)c2ccc(F)cc2)C3C)n1 Animal hepatotoxicity A,Animal hepatotoxicity,"Rat, chronic oral administration, Hepatic histopathologic effects, ToxRefDB",0.5867747455286331,True,Positive,0.0030731971130978854,C[C@@H] 1C2=NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O )c2ccc(F)cc2)C3C)n1 Animal hepatotoxicity B,Animal hepatotoxicity,"Hepatocellular hypertrophy, rats, ORAD, HESS",0.6646590439473917,True,Positive,0.0013360236463188587,C[C@@H]1C2= NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O) c2ccc(F)cc2)C3C)n1 Preclinical hepatotoxicity,Animal hepatotoxicity,"Preclinical hepatotoxicity data from PharmaPendium, Leadscopre, and internal repository with 14- to 28-day rat study data",0.8576928962241468,True,Positive,0.011692057666492625,C[C@@H]1C2= NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O) c2ccc(F)cc2)C3C)n1 Diverse DILI A,Heterogenous Data ,Large-scale and diverse ddrug induced liver injury dataset,0.6324274304660036,True,Positive,0.003398315762493277,C[C@@H]1C2 =NN=C(N2CCN1C(=O)C3=CC=C(C=C3)F)C4=NC(=NS4)C,Cc1nsc(-c2nnc3n2CCN(C(=O) c2ccc(F)cc2)C3C)n1

Overall a useful tool to have to hand.

Fantastic news from the Cambridge Crystallographic Data Centre (CCDC), a curated data set of protein structures from the Protein Data Bank (PDB) with predicted hydrogen positions is now available for download. The dataset is taken from the Protein Data Bank (PDB) and has the positions of hydrogens accurately computed, this provides a comprehensive snapshot of protein cavities in the PDB, identifying potential binding sites for small molecules with accurately predicted hydrogen positions for all components.

The news article is here https://www.ccdc.cam.ac.uk/discover/blog/accelerating-drug-discovery-with-the-ccdc-aws-and-intel/.

This large subset of the protein data bank which has be processed using the CCDC's protonation workflow so that reasonable proton positions have been modelled can be downloaded here.

https://www.ccdc.cam.ac.uk/support-and-resources/downloads/.

Whilst general large language models have hit the headlines in recent years, there is a school of thought that smaller domain specific models may actually more useful, in particular in areas like chemistry https://pubs.rsc.org/en/content/articlelanding/2023/dd/d2dd00087c and https://arxiv.org/abs/2402.09391.

A recent preprint describes Tx-LLM a large language model (LLM) for drug discovery https://arxiv.org/pdf/2406.06316. This work from Google Research and Google DeepMind details Tx-LLM, a LLM specifically designed to enhance drug discovery.

Developing therapeutics is a lengthy and expensive process that requires the satisfaction of many different criteria, and AI models capable of expediting the process would be invaluable. However, the majority of current AI approaches address only a narrowly defined set of tasks, often circumscribed within a particular domain. To bridge this gap, we introduce Tx-LLM, a generalist large language model (LLM) fine-tuned from PaLM-2 which encodes knowledge about diverse therapeutic modalities. Tx-LLM is trained using a collection of 709 datasets that target 66 tasks spanning various stages of the drug discovery pipeline. Using a single set of weights, Tx-LLM simultaneously processes a wide variety of chemical or biological entities (small molecules, proteins, nucleic acids, cell lines, diseases) interleaved with free-text, allowing it to predict a broad range of associated properties, achieving competitive with state-of-the-art (SOTA) performance on 43 out of 66 tasks and exceeding SOTA on 22. Among these, Tx-LLM is particularly powerful and exceeds best-in-class performance on average for tasks combining molecular SMILES representations with text such as cell line names or disease names, likely due to context learned during pretraining. We observe evidence of positive transfer between tasks with diverse drug types (e.g., tasks involving small molecules and tasks involving proteins), and we study the impact of model size, domain finetuning, and prompting strategies on performance. We believe Tx-LLM represents an important step towards LLMs encoding biochemical knowledge and could have a future role as an end-to-end tool across the drug discovery development pipeline.

The model was trained using 709 drug discovery datasets comprising 66 tasks formatted for instruction tuning from Therapeutics instruction Tuning (TxT) https://tdcommons.ai collection for tasks across the drug discovery spectrum. These tasks include:

• Evaluating drug efficacy and safety.
• Predicting molecular targets.
• Assessing the ease of manufacturing drugs.

Biomedical research often requires collaborations between scientists with different skills so I'm always delighted to highlight grants that help foster collaboration.

This opportunity is intended to support new partnerships between diverse groups of researchers within the remit of Medical Research Council (MRC).

The grant will allow you to:

Establish new, high-value collaborative activities or capabilities
Add value to high-quality scientific programmes that are already supported by grants from MRC and other funders

Funding is available for between one and five years

More details here https://www.ukri.org/opportunity/responsive-mode-partnership/?utmmedium=email&utmsource=govdelivery.

I suspect many of the readers of this site have used some of EMBL-EBI resources and tools are some time.

EMBL’s European Bioinformatics Institute maintains the world’s most comprehensive range of freely available and up-to-date molecular data resources.

https://www.ebi.ac.uk/services/data-resources-and-tools.

These of course include the Alphafold database. ChEMBL, PDB in Europe and much. Much more.

In order to support continued funding they do regular user surveys, and the latest is now online.

https://www.surveymonkey.com/r/HJKYKTT

It takes around 15 mins to complete so please take a tea break and fill it in.