Challenge In Drug Discovery Research Articles

Synthesis, Evaluation of Antiangiogenic Activity of Novel Coumarinyl 4-Thiazolidinone Derivatives.

Cancer remains in the maximum percentage probability of death in spite of the latest advancement in prior diagnosis and different approaches to treatment. The major challenge in drug discovery for medicinal chemists is to develop new cytotoxic agents to cure cancer with fewer side effects. A new series of substituted coumarinyl 4-thiazolidinone derivatives (TZL1–TZL7) were prepared by treating different 4-hydroxy-3-(1-(phenylimino0ethyl)-2H-chromen-2-one with HSCH2COOH. The Schiff bases were obtained by treating 4-hydroxy coumarin with different aniline derivatives in the presence of C2H5OH. All the title compounds were purified and subjected to TLC to check their purity. Further, it was characterized by using IR, 1 H-NMR, and mass spectral studies. Selected synthesized compounds (TZL1–TZL7) were screened for in-vivo antiangiogenic activity by using the Zebrafish model. TZL2 had the best antiangiogenic effect with an average vein count of 2.7. TZL4, TZL5, and TZL6 had an average vein count of 4.1, 3.9, and 3.4, respectively.

A Comprehensive Review of Systemic Targeted Therapies in Cancer Treatment

Abstract: Cancer is one of the significant healthcare challenges in today’s world, even after advancements in modern science, including oncology. The complex nature of the disease, which involves multiple proteins and pathways, poses a substantial challenge in drug discovery. Several therapeutic options have emerged in the last decade. Systemic cancer therapies began with the advent of chemotherapy and were revolutionized with the development of targeted therapies. The present review is a definite overview of the advances in various therapeutic options for cancer, with a particular emphasis on targeted therapy using small molecules and biologicals.

Evaluation of reinforcement learning in transformer-based molecular design

Designing compounds with a range of desirable properties is a fundamental challenge in drug discovery. In pre-clinical early drug discovery, novel compounds are often designed based on an already existing promising starting compound through structural modifications for further property optimization. Recently, transformer-based deep learning models have been explored for the task of molecular optimization by training on pairs of similar molecules. This provides a starting point for generating similar molecules to a given input molecule, but has limited flexibility regarding user-defined property profiles. Here, we evaluate the effect of reinforcement learning on transformer-based molecular generative models. The generative model can be considered as a pre-trained model with knowledge of the chemical space close to an input compound, while reinforcement learning can be viewed as a tuning phase, steering the model towards chemical space with user-specific desirable properties. The evaluation of two distinct tasks—molecular optimization and scaffold discovery—suggest that reinforcement learning could guide the transformer-based generative model towards the generation of more compounds of interest. Additionally, the impact of pre-trained models, learning steps and learning rates are investigated.Scientific contributionOur study investigates the effect of reinforcement learning on a transformer-based generative model initially trained for generating molecules similar to starting molecules. The reinforcement learning framework is applied to facilitate multiparameter optimisation of starting molecules. This approach allows for more flexibility for optimizing user-specific property profiles and helps finding more ideas of interest.

Direct Inhibition of the YAP : TEAD Interaction: An Unprecedented Drug Discovery Challenge.

The Hippo pathway, which is key in organ morphogenesis, is frequently deregulated in cancer. The TEAD (TEA domain family member) transcription factors are the most distal elements of this pathway, and their activity is regulated by proteins such as YAP (Yes-associated protein). The identification of inhibitors of the YAP : TEAD interaction is one approach to develop novel anticancer drugs: the first clinical candidate (IAG933) preventing the association between these two proteins by direct competition has just been reported. The discovery of this molecule was particularly challenging because the interface between these two proteins is large (~3500 Å2 buried in complex formation) and made up of distinct contact areas. The most critical of these involves an omega-loop (Ω-loop), a secondary structure element rarely found in protein-protein interactions. This review summarizes how the knowledge gained from structure-function studies of the interaction between the Ω-loop of YAP and TEAD was used to devise the strategy to identify potent low-molecular weight compounds that show a pronounced anti-tumor effect.

De Novo Design of Peptide Binders to Conformationally Diverse Targets with Contrastive Language Modeling.

Designing binders to target undruggable proteins presents a formidable challenge in drug discovery, requiring innovative approaches to overcome the lack of putative binding sites. Recently, generative models have been trained to design binding proteins via three-dimensional structures of target proteins, but as a result, struggle to design binders to disordered or conformationally unstable targets. In this work, we provide a generalizable algorithmic framework to design short, target-binding linear peptides, requiring only the amino acid sequence of the target protein. To do this, we propose a process to generate naturalistic peptide candidates through Gaussian perturbation of the peptidic latent space of the ESM-2 protein language model, and subsequently screen these novel linear sequences for target-selective interaction activity via a CLIP-based contrastive learning architecture. By integrating these generative and discriminative steps, we create a Peptide Prioritization via CLIP (PepPrCLIP) pipeline and validate highly-ranked, target-specific peptides experimentally, both as inhibitory peptides and as fusions to E3 ubiquitin ligase domains, demonstrating functionally potent binding and degradation of conformationally diverse protein targets in vitro. Overall, our design strategy provides a modular toolkit for designing short binding linear peptides to any target protein without the reliance on stable and ordered tertiary structure, enabling generation of programmable modulators to undruggable and disordered proteins such as transcription factors and fusion oncoproteins.

Improved Detection of Drug-Induced Liver Injury by Integrating Predicted In Vivo and In Vitro Data.

Drug-induced liver injury (DILI) has been a significant challenge in drug discovery, often leading to clinical trial failures and necessitating drug withdrawals. Over the last decade, the existing suite of in vitro proxy-DILI assays has generally improved at identifying compounds with hepatotoxicity. However, there is considerable interest in enhancing the in silico prediction of DILI because it allows for evaluating large sets of compounds more quickly and cost-effectively, particularly in the early stages of projects. In this study, we aim to study ML models for DILI prediction that first predict nine proxy-DILI labels and then use them as features in addition to chemical structural features to predict DILI. The features include in vitro (e.g., mitochondrial toxicity, bile salt export pump inhibition) data, in vivo (e.g., preclinical rat hepatotoxicity studies) data, pharmacokinetic parameters of maximum concentration, structural fingerprints, and physicochemical parameters. We trained DILI-prediction models on 888 compounds from the DILI data set (composed of DILIst and DILIrank) and tested them on a held-out external test set of 223 compounds from the DILI data set. The best model, DILIPredictor, attained an AUC-PR of 0.79. This model enabled the detection of the top 25 toxic compounds (2.68 LR+, positive likelihood ratio) compared to models using only structural features (1.65 LR+ score). Using feature interpretation from DILIPredictor, we identified the chemical substructures causing DILI and differentiated cases of DILI caused by compounds in animals but not in humans. For example, DILIPredictor correctly recognized 2-butoxyethanol as nontoxic in humans despite its hepatotoxicity in mice models. Overall, the DILIPredictor model improves the detection of compounds causing DILI with an improved differentiation between animal and human sensitivity and the potential for mechanism evaluation. DILIPredictor required only chemical structures as input for prediction and is publicly available at https://broad.io/DILIPredictor for use via web interface and with all code available for download.

A Hierarchical Graph Neural Network Framework for Predicting Protein-Protein Interaction Modulators With Functional Group Information and Hypergraph Structure.

Accurate prediction of small molecule modulators targeting protein-protein interactions (PPIMs) remains a significant challenge in drug discovery. Existing machine learning-based models rely on manual feature engineering, which is tedious and task-specific. Recently, deep learning models based on graph neural networks have made remarkable progress in molecular representation learning. However, many graph-based approaches ignore molecular hierarchical structure modeling guided by domain knowledge. In chemistry, the functional groups of a molecule determine its interaction with specific targets. Therefore, we propose a hierarchical graph neural network framework (called HiGPPIM) for predicting PPIMs by integrating atom-level and functional group-level features of molecules. HiGPPIM constructs atom-level and functional group-level graphs based on chemical knowledge and learns graph representations using graph attention networks. Furthermore, a hypergraph attention network is designed in HiGPPIM to aggregate and transform two-level graph information. We evaluate the performance of HiGPPIM on eight PPI families and two prediction tasks, namely PPIM identification and potency prediction. Experimental results demonstrate that HiGPPIM achieves state-of-the-art performance on both tasks and that using functional group information to guide PPIM prediction is effective.

Targeting SARS-CoV-2 Spike and Main protease with phytochemicals from Herbs and spices: Molecular Docking and dynamics simulation studies.

Abstract COVID-19, a pandemic disease has affected 480 million people and caused 6 million deaths around the world. Despite the progress made in COVID-19 drug discovery, SARS-CoV-2, the causative agent of this disease continuously mutates and rapidly evolves into new variants. This increases the challenges in drug discovery for COVID-19. As natural products serve as sources of drugs forever, this study applies computational techniques in predicting the natural compounds in herbs and spices of household origin as SARS-CoV-2 spike and protease inhibitors and also verifies the top hits against spike and protease mutants associated with SARS-CoV-2 variants of concern. This study reveals Hesperidin, Diosgenin, Fenugreekine, Epigallocatechin gallate and Quercetin as SARS-CoV-2 spike and protease inhibitors, which acts better than the drug Remdesivir. The efficiency of the top hits was also been verified against the mutants, which reveals Diosgenin and Fenugreekine as the most efficient natural compounds against SARS-CoV-2 spike mutants (N501Y, E484K, K417N and K417T), which are associated with SARS-CoV-2 variants of concern. Additionally, Hesperidin is proven to have better binding efficiency against Mpro mutants (Y54C, A191V, T190I and N142S). Overall, this study concludes that Hesperidin, Diosgenin and Fenugreekine could combat both SARS-CoV-2 and its variants effectively.

Altered desensitization and internalization patterns of rodent versus human glucose-dependent insulinotropic polypeptide (GIP) receptors. An important drug discovery challenge.

The gut hormone glucose-dependent insulinotropic polypeptide (GIP) signals via the GIP receptor (GIPR), resulting in postprandial potentiation of glucose-stimulated insulin secretion. The translation of results from rodent studies to human studies has been challenged by the unexpected effects of GIPR-targeting compounds. We, therefore, investigated the variation between species, focusing on GIPR desensitization and the role of the receptor C-terminus. The GIPR from humans, mice, rats, pigs, dogs and cats was studied in vitro for cognate ligand affinity, G protein activation (cAMP accumulation), recruitment of beta-arrestin and internalization. Variants of the mouse, rat and human GIPRs with swapped C-terminal tails were studied in parallel. The human GIPR is more prone to internalization than rodent GIPRs. Despite similar agonist affinities and potencies for Gαs activation, especially, the mouse GIPR shows reduced receptor desensitization, internalization and beta-arrestin recruitment. Using an enzyme-stabilized, long-acting GIP analogue, the species differences were even more pronounced. 'Tail-swapped' human, rat and mouse GIPRs were all fully functional in their Gαs coupling, and the mouse GIPR regained internalization and beta-arrestin 2 recruitment properties with the human tail. The human GIPR lost the ability to recruit beta-arrestin 2 when its own C-terminus was replaced by the rat or mouse tail. Desensitization of the human GIPR is dependent on the C-terminal tail. The species-dependent functionality of the C-terminal tail and the different species-dependent internalization patterns, especially between human and mouse GIPRs, are important factors influencing the preclinical evaluation of GIPR-targeting therapeutic compounds.

An Overview on Nocardiopsis Species Originating From North African Biotopes as a Promising Source of Bioactive Compounds and In Silico Genome Mining Analysis of Three Sequenced Genomes.

Actinobacteria are renowned for their prolific production of diverse bioactive secondary metabolites. In recent years, there has been an increasing focus on exploring "rare" genera within this phylum for biodiscovery purposes, notably the Nocardiopsis genus, which will be the subject of the present study. Recognizing the absence of articles describing the research process of finding bioactive molecules from the genus Nocardiopsis in North African environments. We, therefore, present a historical overview of the discoveries of bioactive molecules of the genus Nocardiopsis originating from the region, highlighting their biological activities and associated reported molecules, providing a snapshot of the current state of the field, and offering insights into future opportunities and challenges for drug discovery. Additionally, we present a genome mining analysis of three genomes deposited in public databases that have been reported to be bioactive. A total of 36 biosynthetic gene clusters(BGCs) were identified, including those known to encode bioactive molecules. Notably, a substantial portion of the BGCsshowed little to no similarity to those previously described, suggesting the possibility that the analyzed strains could be potential producers of new compounds. Further research on these genomes is essential to fully uncovering their biotechnological potential. Moving forward, we discuss the experimental designs adopted in the reported studies, as well as new avenues to guide the exploration of the Nocardiopsis genus in North Africa.

Ligandability at the Membrane Interface of GPx4 Revealed through a Reverse Micelle Fragment Screening Platform.

While they account for a large portion of drug targets, membrane proteins present a unique challenge for drug discovery. Peripheral membrane proteins (PMPs), a class of water-soluble proteins that bind to membranes, are also difficult targets, particularly those that function only when bound to membranes. The protein-membrane interface in PMPs is often where functional interactions and catalysis occur, making it a logical target for inhibition. However, protein-membrane interfaces are underexplored spaces in inhibitor design, and there is a need for enhanced methods for small-molecule ligand discovery. In an effort to better initiate drug discovery efforts for PMPs, this study presents a screening methodology using membrane-mimicking reverse micelles (mmRM) and NMR-based fragment screening to assess ligandability at the protein-membrane interface. The proof-of-principle target, glutathione peroxidase 4 (GPx4), is a lipid hydroperoxidase that is essential for the oxidative protection of membranes and thereby the prevention of ferroptosis. GPx4 inhibition is promising for therapy-resistant cancer therapy, but current inhibitors are generally covalent ligands with limited clinical utility. Presented here is the discovery of noncovalent small-molecule ligands for membrane-bound GPx4 revealed through the mmRM fragment screening methodology. The fragments were tested against GPx4 under bulk aqueous conditions and displayed little to no binding to the protein without embedment into the membrane. The 9 hits had varying affinities and partitioning coefficients and revealed properties of fragments that bind within the protein-membrane interface. Additionally, a secondary screen confirmed the potential to progress the fragments by enhancing the affinity from >200 to ∼15 μM with the addition of certain hydrophobic groups. This study presents an advancement of screening capabilities for membrane-associated proteins, reveals ligandability within the GPx4 protein-membrane interface, and may serve as a starting point for developing noncovalent inhibitors of GPx4.

Improved Detection of Drug-Induced Liver Injury by Integrating Predicted in vivo and in vitro Data.

Drug-induced liver injury (DILI) has been significant challenge in drug discovery, often leading to clinical trial failures and necessitating drug withdrawals. The existing suite of in vitro proxy-DILI assays is generally effective at identifying compounds with hepatotoxicity. However, there is considerable interest in enhancing in silico prediction of DILI because it allows for the evaluation of large sets of compounds more quickly and cost-effectively, particularly in the early stages of projects. In this study, we aim to study ML models for DILI prediction that first predicts nine proxy-DILI labels and then uses them as features in addition to chemical structural features to predict DILI. The features include in vitro (e.g., mitochondrial toxicity, bile salt export pump inhibition) data, in vivo (e.g., preclinical rat hepatotoxicity studies) data, pharmacokinetic parameters of maximum concentration, structural fingerprints, and physicochemical parameters. We trained DILI-prediction models on 888 compounds from the DILIst dataset and tested on a held-out external test set of 223 compounds from DILIst dataset. The best model, DILIPredictor, attained an AUC-ROC of 0.79. This model enabled the detection of top 25 toxic compounds compared to models using only structural features (2.68 LR+ score). Using feature interpretation from DILIPredictor, we were able to identify the chemical substructures causing DILI as well as differentiate cases DILI is caused by compounds in animals but not in humans. For example, DILIPredictor correctly recognized 2-butoxyethanol as non-toxic in humans despite its hepatotoxicity in mice models. Overall, the DILIPredictor model improves the detection of compounds causing DILI with an improved differentiation between animal and human sensitivity as well as the potential for mechanism evaluation. DILIPredictor is publicly available at https://broad.io/DILIPredictor for use via web interface and with all code available for download and local implementation via https://pypi.org/project/dilipred/.

A Double-Humanized Mouse Model for Studying Host Gut Microbiome-Immune Interactions in Gulf War Illness.

Unraveling the multisymptomatic Gulf War Illness (GWI) pathology and finding an effective cure have eluded researchers for decades. The chronic symptom persistence and limitations for studying the etiologies in mouse models that differ significantly from those in humans pose challenges for drug discovery and finding effective therapeutic regimens. The GWI exposome differs significantly in the study cohorts, and the above makes it difficult to recreate a model closely resembling the GWI symptom pathology. We have used a double engraftment strategy for reconstituting a human immune system coupled with human microbiome transfer to create a humanized-mouse model for GWI. Using whole-genome shotgun sequencing and blood immune cytokine enzyme linked immunosorbent assay (ELISA), we show that our double humanized mice treated with Gulf War (GW) chemicals show significantly altered gut microbiomes, similar to those reported in a Veteran cohort of GWI. The results also showed similar cytokine profiles, such as increased levels of IL-1β, IL-6, and TNF R-1, in the double humanized model, as found previously in a human cohort. Further, a novel GWI Veteran fecal microbiota transfer was used to create a second alternative model that closely resembled the microbiome and immune-system-associated pathology of a GWI Veteran. A GWI Veteran microbiota transplant in humanized mice showed a human microbiome reconstitution and a systemic inflammatory pathology, as reflected by increases in interleukins 1β, 6, 8 (IL-1β, IL-6, IL-8), tumor necrosis factor receptor 1 (TNF R-1), and endotoxemia. In conclusion, though preliminary, we report a novel in vivo model with a human microbiome reconstitution and an engrafted human immune phenotype that may help to better understand gut-immune interactions in GWI.

Phytophenols as Promoieties for Prodrug Design.

One lucrative method for overcoming challenges in drug discovery or for enhancing undesirable properties of already-approved medications is prodrug design. The goal of this review is to present researchers with a profile of naturally occurring Phytophenols as carriers that would be used for prodrug synthesis as well as their advantages. Phytophenols offer several advantages when used as promoieties as they also possess antioxidant and analgesic properties, they are obtained naturally and their safety profile is well established. Several phytophenols like menthol, thymol, eugenol, guaiacol, sesamol, vanillin, and umbelliferone are some of the phytophenols that have several beneficial properties and are extensively employed in the field of food processing and medicine. In the current review, we have listed all types of promoieties that are used for prodrug synthesis and phytophenols are reviewed in detail, which may help researchers to select phytophenols based on their need and suitability for drug candidates.

Site Identification and Next Choice Protocol for Hit-to-Lead Optimization.

Time efficiency and cost savings are major challenges in drug discovery and development. In this process, the hit-to-lead stage is expected to improve efficiency because it primarily exploits the trial-and-error approach of medicinal chemists. This study proposes a site identification and next choice (SINCHO) protocol to improve the hit-to-lead efficiency. This protocol selects an anchor atom and growth site pair, which is desirable for a hit-to-lead strategy starting from a 3D complex structure. We developed and fine-tuned the protocol using a training data set and assessed it using a test data set of the preceding hit-to-lead strategy. The protocol was tested for experimentally determined structures and molecular dynamics (MD) ensembles. The protocol had a high prediction accuracy for applying MD ensembles, owing to the consideration of protein flexibility. The SINCHO protocol enables medicinal chemists to visualize and modify functional groups in a hit-to-lead manner.

Synthesis and interest in medicinal chemistry of β-phenylalanine derivatives (β-PAD): an update (2010-2022).

β-Phenylalanine derivatives (β-PAD) represent a structural family of therapeutic interest, either as components of drugs or as starting materials for access to key compounds. As scaffolds for medicinal chemistry work, β-PAD offer the advantage of great diversity and modularity, a chiral pseudopeptidic character that opens up the capacity to be recognized by natural systems, and greater stability than natural α-amino acids. Nevertheless, their synthesis remains a challenge in drug discovery and numerous methods have been devoted to their preparation. This review is an update of the access routes to β-PAD and their various therapeutic applications.

Paradigms and Success Stories of Natural Products in Drug Discovery Against Neurodegenerative Disorders (NDDs).

Neurodegenerative disorders (NDDs) are multifaceted complex disorders that have put a great health and economic burden around the globe nowadays. The multi-factorial nature of NDDs has presented a great challenge in drug discovery and continuous efforts are in progress in search of suitable therapeutic candidates. Nature has a great wealth of active principles in its lap that has cured the human population since ancient times. Natural products have revealed several benefits over conventional synthetic medications and scientists have shifted their vision towards exploring the therapeutic potentials of natural products in the past few years. The structural mimicking of natural compounds to endogenous ligands has presented them as a potential therapeutic candidate to prevent the development of NDDs. In the presented review, authors have summarized demographical facts about various NDDs including Alzheimer's disease (AD), Parkinson's disease (PD), Huntington's disease (HD) and various types of sclerosis in the brain. The significant findings of new active principles of natural origin along with their therapeutic potentials on NDDs have been included. Also, a description of clinical trials and patents on natural products has been enlisted in this compilation. Although natural products have shown promising success in drug discovery against NDDs, still their use is associated with several ethical issues which need to be solved in the upcoming time.

SGCLDGA: unveiling drug-gene associations through simple graph contrastive learning.

Drug repurposing offers a viable strategy for discovering new drugs and therapeutic targets through the analysis of drug-gene interactions. However, traditional experimental methods are plagued by their costliness and inefficiency. Despite graph convolutional network (GCN)-based models' state-of-the-art performance in prediction, their reliance on supervised learning makes them vulnerable to data sparsity, a common challenge in drug discovery, further complicating model development. In this study, we propose SGCLDGA, a novel computational model leveraging graph neural networks and contrastive learning to predict unknown drug-gene associations. SGCLDGA employs GCNs to extract vector representations of drugs and genes from the original bipartite graph. Subsequently, singular value decomposition (SVD) is employed to enhance the graph and generate multiple views. The model performs contrastive learning across these views, optimizing vector representations through a contrastive loss function to better distinguish positive and negative samples. The final step involves utilizing inner product calculations to determine association scores between drugs and genes. Experimental results on the DGIdb4.0 dataset demonstrate SGCLDGA's superior performance compared with six state-of-the-art methods. Ablation studies and case analyses validate the significance of contrastive learning and SVD, highlighting SGCLDGA's potential in discovering new drug-gene associations. The code and dataset for SGCLDGA are freely available at https://github.com/one-melon/SGCLDGA.

Antimicrobial Lasso Peptide Cloacaenodin Utilizes a Unique TonB-Dependent Transporter to Access Susceptible Bacteria.

The development of new antimicrobial agents effective against Gram-negative bacteria remains a major challenge in drug discovery. The lasso peptide cloacaenodin has potent antimicrobial activity against multiple strains in the Enterobacter genus, one of the ESKAPE pathogens. Here, we show that cloacaenodin uses a previously uncharacterized TonB-dependent transporter, which we name CloU, to cross the outer membrane (OM) of susceptible bacteria. Inner membrane transport is mediated by the protein SbmA. CloU is distinct from the known OM transporters (FhuA and PupB) utilized by other antimicrobial lasso peptides and thus offers important insight into the spectrum of activity of cloacaenodin. Using knowledge of the transport pathway to predict other cloacaenodin-susceptible strains, we demonstrate the activity of cloacaenodin against clinical isolates of Enterobacter and of a Kluyvera strain. Further, we use molecular dynamics simulations and mutagenesis of CloU to explain the variation in cloacaenodin susceptibility observed across different strains of Enterobacter. This work expands the currently limited understanding of lasso peptide uptake and advances the potential of cloacaenodin as an antibiotic.

Abstract 902: Explainable AI: Graph machine learning for response prediction and biomarker discovery

Abstract Accurately predicting drug sensitivity and understanding what is driving it are major challenges in drug discovery. Graphs are a natural framework for capturing diverse pharmacological data for efficacy predictions, thanks to their ability to integrate multimodal data and represent relationships such as gene-gene or drug-target interactions as edges. They have also had proven success across a range of other drug discovery tasks including repositioning and target identification. In this study, we sought to address the explainability challenges of drug response predictions. Recent developments in the field of Graph AI have led to improvements in interpretability mechanisms that highlight parts of a graph which are driving predictions. We have conducted a comprehensive review of multiple major approaches for tackling drug efficacy prediction using graph methods, benchmarking the performance and interpretability of these algorithms across indications. Methods: We assembled a combined dataset of GDSC1 and GDSC2 drug response data in cell lines, with multiomic cell line data and drug target and chemical structure data. We then applied graph-based approaches for the prediction of binarized IC50 on an indication-by-indication basis. Approach 1 involved the creation of a ‘GDSC knowledge graph’, where drug response and cell line ‘omic information is represented in an unweighted knowledge graph: cell lines are connected to genes expressed in them, drugs are connected to genes they target, and so on. We then used state-of-the-art graph embedding techniques to predict IC50 using paired drug and cell line embeddings. In Approach 2 we used a weighted knowledge graph instead, and generated embeddings using heterogeneous graph neural networks (HGNNs). In Approach 3, we modelled response prediction as a graph classification task, where one single graph captures one drug-cell line interaction. The graph classifier and HGNN models both have in-built interpretability mechanisms, including graph attention, that can signify the genes in the cell line which were most important for the eventual prediction. We can also integrate biomedical prior knowledge with all these models by capturing gene-pathway and gene-gene data in the graphs. Results: Our models outperformed benchmark models including DNNs and GBMs, and identified both established and novel response biomarkers in NSCLC cell lines (AUC = 0.94, Accuracy = 89%). We have also applied our models to Breast Cancer, Pancreatic Cancer, Colorectal Cancer and Haematological malignancies with similar predictive performance and explainability. Conclusions:Our graph analytical framework for response predictions showed better performance than benchmarking models and provided insights from explainability. This framework is easily extendable to response and ‘omic data from any disease model and patient studies. Citation Format: Jake Cohen-Setton, Krishna Bulusu, Jonathan Dry, Ben Sidders. Explainable AI: Graph machine learning for response prediction and biomarker discovery [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 902.