Artificial intelligence uncovers carcinogenic human metabolites

Mittal, Aayushi; Mohanty, Sanjay Kumar; Gautam, Vishakha; Arora, Sakshi; Saproo, Sheetanshu; Gupta, Ria; Sivakumar, Roshan; Garg, Prakriti; Aggarwal, Anmol; Raghavachary, Padmasini; Dixit, Nilesh Kumar; Singh, V.R.; Mehta, Anurag; Tayal, Juhi; Naidu, Srivatsava; Sengupta, Debarka; Ahuja, Gaurav

doi:10.1038/s41589-022-01110-7

Cited by 9 publications

(9 citation statements)

References 86 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our model can be extended to other end points with explicit task relationships that can be leveraged. For example, the carcinogenic property is related to cellular proliferation, genomic instability, oxidative stress response, antiapoptotic response, epigenetic modifications, and electrophilic property . Notably, the backbone expert network in our model can be replaced by any other GNN architecture.…”

Section: Discussionmentioning

confidence: 99%

“…For example, the carcinogenic property is related to cellular proliferation, genomic instability, oxidative stress response, antiapoptotic response, epigenetic modifications, and electrophilic property. 43 Notably, the backbone expert network in our model can be replaced by any other GNN architecture. In the future, the model capability can be further improved with the further development of molecular graph representation learning, especially the integration of 3D structural information.…”

Section: ■ Conclusionmentioning

confidence: 99%

See 1 more Smart Citation

ALipSol: An Attention-Driven Mixture-of-Experts Model for Lipophilicity and Solubility Prediction

Wang

et al. 2022

J. Chem. Inf. Model.

View full text Add to dashboard Cite

Lipophilicity (logD) and aqueous solubility (logS w) play a central role in drug development. The accurate prediction of these properties remains to be solved due to data scarcity. Current methodologies neglect the intrinsic relationships between physicochemical properties and usually ignore the ionization effects. Here, we propose an attention-driven mixture-of-experts (MoE) model named ALipSol, which explicitly reproduces the hierarchy of task relationships. We adopt the principle of divide-and-conquer by breaking down the complex end point (logD or logS w) into simpler ones (acidic pK a, basic pK a, and logP) and allocating a specific expert network for each subproblem. Subsequently, we implement transfer learning to extract knowledge from related tasks, thus alleviating the dilemma of limited data. Additionally, we substitute the gating network with an attention mechanism to better capture the dynamic task relationships on a per-example basis. We adopt local fine-tuning and consensus prediction to further boost model performance. Extensive evaluation experiments verify the success of the ALipSol model, which achieves RMSE improvement of 8.04%, 2.49%, 8.57%, 12.8%, and 8.60% on the Lipop, ESOL, AqSolDB, external logD, and external logS data sets, respectively, compared with Attentive FP and the state-of-the-art in silico tools. In particular, our model yields more significant advantages (Welch’s t-test) for small training data, implying its high robustness and generalizability. The interpretability analysis proves that the atom contributions learned by ALipSol are more reasonable compared with the vanilla Attentive FP, and the substitution effects in benzene derivatives agreed well with empirical constants, revealing the potential of our model to extract useful patterns from data and provide guidance for lead optimization.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: ■ Conclusionmentioning

confidence: 99%

ALipSol: An Attention-Driven Mixture-of-Experts Model for Lipophilicity and Solubility Prediction

Wang

et al. 2022

J. Chem. Inf. Model.

View full text Add to dashboard Cite

show abstract

“…For several of the studies, 79,80 BO approaches were used with Gaussian Process (GP) models as surrogate optimization functions. Langner et al chose to use Bayesian neural networks in order to avoid the very poor performance scaling ( N ( ) 3 ) that GPs exhibit with problem size. 84 There are methods to improve the performance of GPs for large problems, but they are not necessarily applicable in all cases.…”

Section: ■ New Progress Autonomous Experimentationmentioning

confidence: 99%

“…The latest AI image generation models can transform text strings into images that can be nearly indistinguishable from high-quality human generated art and photography . In medicine, machine learning (ML) models are being used to identify carcinogens and diagnose diseases such as Parkinson’s that, previously, could not be identified from biomarkers. , Decades long scientific problems, such as the classic “protein folding” problem, are being tackled by AI that produce results which approach the resolution of our best measurements . With each major advance, it is clear that we have yet to realize the full impact of AI and ML.…”

Section: Introductionmentioning

confidence: 99%

Emerging Trends in Machine Learning: A Polymer Perspective

Martin

Audus

2023

ACS Polym. Au

View full text Add to dashboard Cite

In the last five years, there has been tremendous growth in machine learning and artificial intelligence as applied to polymer science. Here, we highlight the unique challenges presented by polymers and how the field is addressing them. We focus on emerging trends with an emphasis on topics that have received less attention in the review literature. Finally, we provide an outlook for the field, outline important growth areas in machine learning and artificial intelligence for polymer science and discuss important advances from the greater material science community.

show abstract

“…Post-metabolite loading, the cells were either treated or untreated with the α-factor at a nal concentration of 30 µM for 2 hours at desired growth conditions. Post these steps, the cells were harvested, and RNA was isolated using the protocol described in Mittal et al 94 .…”

Section: Rna-sequencingmentioning

confidence: 99%

Artificial Intelligence Uncovers Evolutionarily Conserved Intracellular Allosteric Modulators of GPCR-Gα Interface

Mohanty

Mittal

Gaur

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Traditionally, the exogenous allosteric modulators of G protein-coupled receptors (GPCRs) have been extensively investigated due to their pharmacological significance. However, to date, only a handful of endogenous intracellular allosteric modulators are known, that too with inconclusive binding information and their associated phenotypes. This limitation primarily stems from the non-availability of robust computational techniques that entails unbiased cavity identification across GPCR protein topology, cavity-specific ligand design, their synthesis, and cross-validation. Here, we introduce Gcoupler, which leverages an integrative approach combining de novo ligand design, statistical methods, and Graph Neural Networks for rationally predicting high-affinity ligands. Gcoupler offers an efficient and comparatively faster route to explore endogenous allosteric sites of GPCRs, including the GPCR-Gα interface. We tested and validated the applicability of Gcoupler in decrypting the cellular metabolites that could intracellularly but directly modulate the Ste2 (GPCR)-mediated pheromone-induced cell death in yeast. Our rigorous interrogation using Gcoupler and experimental approaches, including yeast genetic screening, RNA Sequencing, high-resolution metabolomics, and functional assays, identified endogenous hydrophobic metabolites as intracellular allosteric inhibitors of Ste2p signaling. Elevated intracellular levels of these metabolites, either naturally, through genetic alterations, or exogenous treatment, rescue the pheromone-induced programmed cell death. Mechanistic insights revealed that metabolites harbor high-binding affinity to the conserved GPCR-Gα interface and trigger a cohesive response that potentially obstructs downstream signaling. Finally, by utilizing isoproterenol-induced, GPCR-mediated human and neonatal rat cardiac hypertrophy models, we observed that elevated metabolite levels attenuate hypertrophic response, reinforcing the functional and evolutionary relevance of this mechanism. In summary, our study reports a robust computational method that uncovered a novel, evolutionary conserved, and metabolite-driven regulatory mechanism of GPCR signaling.

show abstract

Artificial intelligence uncovers carcinogenic human metabolites

Cited by 9 publications

References 86 publications

ALipSol: An Attention-Driven Mixture-of-Experts Model for Lipophilicity and Solubility Prediction

ALipSol: An Attention-Driven Mixture-of-Experts Model for Lipophilicity and Solubility Prediction

Emerging Trends in Machine Learning: A Polymer Perspective

Artificial Intelligence Uncovers Evolutionarily Conserved Intracellular Allosteric Modulators of GPCR-Gα Interface

Contact Info

Product

Resources

About