Synopsis: Despite efforts to integrate research across different subdisciplines of biology, the scale of integration remains limited. We hypothesize that future generations of Artificial Intelligence (AI) technologies specifically adapted for biological sciences will help enable the reintegration of biology. AI technologies will allow us not only to collect, connect, and analyze data at unprecedented scales, but also to build comprehensive predictive models that span various subdisciplines. They will make possible both targeted (testing specific hypotheses) and untargeted discoveries. AI for biology will be the cross-cutting technology that will enhance our ability to do biological research at every scale. We expect AI to revolutionize biology in the 21st century much like statistics transformed biology in the 20th century. The difficulties, however, are many, including data curation and assembly, development of new science in the form of theories that connect the subdisciplines, and new predictive and interpretable AI models that are more suited to biology than existing machine learning and AI techniques. Development efforts will require strong collaborations between biological and computational scientists. This white paper provides a vision for AI for Biology and highlights some challenges.
Contamination of the environment with bioactive chemicals has emerged as a potential public health risk. These substances that may cause distress or disease in humans can be found in air, water and food supplies. An open question is whether these chemicals transform into potentially more active or toxic derivatives via xenobiotic metabolizing enzymes expressed in the body. We present a new prediction tool, which we call PROXIMAL (Prediction of Xenobiotic Metabolism) for identifying possible transformation products of xenobiotic chemicals in the liver. Using reaction data from DrugBank and KEGG, PROXIMAL builds look-up tables that catalog the sites and types of structural modifications performed by Phase I and Phase II enzymes. Given a compound of interest, PROXIMAL searches for substructures that match the sites cataloged in the look-up tables, applies the corresponding modifications to generate a panel of possible transformation products, and ranks the products based on the activity and abundance of the enzymes involved. PROXIMAL generates transformations that are specific for the chemical of interest by analyzing the chemical’s substructures. We evaluate the accuracy of PROXIMAL’s predictions through case studies on two environmental chemicals with suspected endocrine disrupting activity, bisphenol A (BPA) and 4-chlorobiphenyl (PCB3). Comparisons with published reports confirm 5 out of 7 and 17 out of 26 of the predicted derivatives for BPA and PCB3, respectively. We also compare biotransformation predictions generated by PROXIMAL with those generated by METEOR and Metaprint2D-react, two other prediction tools. PROXIMAL can predict transformations of chemicals that contain substructures recognizable by human liver enzymes. It also has the ability to rank the predicted metabolites based on the activity and abundance of enzymes involved in xenobiotic transformation.
As experimental efforts are costly and time consuming, computational characterization of enzyme capabilities is an attractive alternative. We present and evaluate several machine-learning models to predict which of 983 distinct enzymes, as defined via the Enzyme Commission (EC) numbers, are likely to interact with a given query molecule. Our data consists of enzyme-substrate interactions from the BRENDA database. Some interactions are attributed to natural selection and involve the enzyme’s natural substrates. The majority of the interactions however involve non-natural substrates, thus reflecting promiscuous enzymatic activities. We frame this ‘enzyme promiscuity prediction’ problem as a multi-label classification task. We maximally utilize inhibitor and unlabeled data to train prediction models that can take advantage of known hierarchical relationships between enzyme classes. We report that a hierarchical multi-label neural network, EPP-HMCNF, is the best model for solving this problem, outperforming k-nearest neighbors similarity-based and other machine-learning models. We show that inhibitor information during training consistently improves predictive power, particularly for EPP-HMCNF. We also show that all promiscuity prediction models perform worse under a realistic data split when compared to a random data split, and when evaluating performance on non-natural substrates compared to natural substrates. We provide Python code and data for EPP-HMCNF and other models in a repository termed EPP (Enzyme Promiscuity Prediction) at https://github.com/hassounlab/EPP. Supplementary data are available at Bioinformatics online.
Motivation: While traditionally utilized for identifying site-specific metabolic activity within a compound to alter its interaction with a metabolizing enzyme, predicting the site-of-metabolism (SOM) is essential in analyzing the promiscuity of enzymes on substrates. The successful prediction of SOMs and the relevant promiscuous products has a wide range of applications that include creating extended metabolic models (EMMs) that account for enzyme promiscuity and the construction of novel heterologous synthesis pathways. There is therefore a need to develop generalized methods that can predict molecular SOMs for a wide range of metabolizing enzymes.Results: This article develops a Graph Neural Network (GNN) model for the classification of an atom (or a bond) being an SOM. Our model, GNN-SOM, is trained on enzymatic interactions, available in the KEGG database, that span all enzyme commission numbers. We demonstrate that GNN-SOM consistently outperforms baseline machine learning models, when trained on all enzymes, on Cytochrome P450 (CYP) enzymes, or on non-CYP enzymes. We showcase the utility of GNN-SOM in prioritizing predicted enzymatic products due to enzyme promiscuity for two biological applications: the construction of EMMs and the construction of synthesis pathways.Availability and implementation: A python implementation of the trained SOM predictor model can be found at https://github.com/HassounLab/GNN-SOM.Supplementary information: Supplementary data are available at Bioinformatics online.
A key challenge in metabolomics is annotating measured spectra from a biological sample with chemical identities. Currently, only a small fraction of measurements can be assigned identities. Two complementary computational approaches have emerged to address the annotation problem: mapping candidate molecules to spectra, and mapping query spectra to molecular candidates. In essence, the candidate molecule with the spectrum that best explains the query spectrum is recommended as the target molecule. Despite candidate ranking being fundamental in both approaches, limited prior works incorporated rank learning tasks in determining the target molecule. We propose a novel machine learning model, Ensemble Spectral Prediction (ESP), for metabolite annotation. ESP takes advantage of prior neural network-based annotation models that utilize multilayer perceptron (MLP) networks and Graph Neural Networks (GNNs). Based on the ranking results of the MLP- and GNN-based models, ESP learns a weighting for the outputs of MLP and GNN spectral predictors to generate a spectral prediction for a query molecule. Importantly, training data is stratified by molecular formula to provide candidate sets during model training. Further, baseline MLP and GNN models are enhanced by considering peak dependencies through label mixing and multi-tasking on spectral topic distributions. When trained on the NIST 2020 dataset and evaluated on the relevant candidate sets from PubChem, ESP improves average rank by 23.7% and 37.2% over the MLP and GNN baselines, respectively, demonstrating performance gain over state-of-the-art neural network approaches. However, MLP approaches remain strong contenders when considering top five ranks. Importantly, we show that annotation performance is dependent on the training dataset, the number of molecules in the candidate set and candidate similarity to the target molecule. The ESP code, a trained model, and a Jupyter notebook that guide users on using the ESP tool is available at https://github.com/HassounLab/ESP.
Metabolic models are indispensable in guiding cellular engineering and in advancing our understanding of systems biology. As not all enzymatic activities are fully known and/or annotated, metabolic models remain incomplete, resulting in suboptimal computational analysis and leading to unexpected experimental results. We posit that one major source of unaccounted metabolism is promiscuous enzymatic activity. It is now well-accepted that most, if not all, enzymes are promiscuous—i.e., they transform substrates other than their primary substrate. However, there have been no systematic analyses of genome-scale metabolic models to predict putative reactions and/or metabolites that arise from enzyme promiscuity. Our workflow utilizes PROXIMAL—a tool that uses reactant–product transformation patterns from the KEGG database—to predict putative structural modifications due to promiscuous enzymes. Using iML1515 as a model system, we first utilized a computational workflow, referred to as Extended Metabolite Model Annotation (EMMA), to predict promiscuous reactions catalyzed, and metabolites produced, by natively encoded enzymes in Escherichia coli. We predict hundreds of new metabolites that can be used to augment iML1515. We then validated our method by comparing predicted metabolites with the Escherichia coli Metabolome Database (ECMDB). We utilized EMMA to augment the iML1515 metabolic model to more fully reflect cellular metabolic activity. This workflow uses enzyme promiscuity as basis to predict hundreds of reactions and metabolites that may exist in E. coli but may have not been documented in iML1515 or other databases. We provide detailed analysis of 23 predicted reactions and 16 associated metabolites. Interestingly, nine of these metabolites, which are in ECMDB, have not previously been documented in any other E. coli databases. Four of the predicted reactions provide putative transformations parallel to those already in iML1515. We suggest adding predicted metabolites and reactions to iML1515 to create an extended metabolic model (EMM) for E. coli.