Graph-based machine learning interprets and predicts diagnostic isomer-selective ion–molecule reactions in tandem mass spectrometry

Fine, Jonathan; Liu, Judy Kuan-Yu; Beck, Armen; Alzarieni, Kawthar Z.; Ma, Xin; Boulos, Victoria; Kenttämaa, Hilkka I.; Chopra, Gaurav

doi:10.1039/d0sc02530e

Cited by 15 publications

(20 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Machine learning (ML) has many potential uses within material discovery, including to reduce the cost of property calculation compared to carrying out computational simulations (especially via quantum mechanical methods) and to remove the need for specialist modelling packages. This allows researchers to focus experimental synthesis and measurement effort on the most promising materials, reducing wasted laboratory resources, 13,14 as well as to help facilitate the exploration of larger chemical space. [15][16][17] Apart from the widely reported ML models for molecular discovery, especially drug discovery, the applications of ML to porous materials such as MOFs have gained signicant interest.…”

Section: Introductionmentioning

confidence: 99%

Explainable graph neural networks for organic cages

2022

View full text Add to dashboard Cite

show abstract

Section: Introductionmentioning

confidence: 99%

Explainable graph neural networks for organic cages

2022

View full text Add to dashboard Cite

show abstract

“…Recently, chemists have started to apply machine learning techniques to model various chemical processes for catalyst design 9,10 , prediction of organic reactions 11,12 , reactions in a mass spectrometer 13 , and for the analysis of chemical spectra. Moreover, Lai and coworkers have developed machine learning based methodologies to determine molecular features which are responsible for the viscosity behavior and the aggregation of therapeutic antibodies.…”

Section: Introductionmentioning

confidence: 99%

“…One interesting application of machine learning in chemistry is the ability to understand how underlying variables (x) relate to the output (y). The use of machine learning to understand these variables has been applied to understanding how IR and MS spectra relate to functional groups 17 and how functional groups in organic molecules relate to chemical reactivity 13 . When performing this type of analysis there are two important points to consider.…”

Section: Introductionmentioning

confidence: 99%

“…Recent works have shown that machine learning models built on small training sets can be successful for both understanding chemistry and for useful predictions. 13,[20][21][22] To our knowledge, these techniques have not been used to understand the relationships between storage conditions, chemical stability, and physical stability in drug formulation. In this work, we will outline how an initial design of experiment (DoE) was used to generate training data for both the chemical and physical stability of a peptide drug molecule in formulation matrix.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Learning relationships between chemical and physical stability for drug development

Fine

Wijewardhane

Mohideen

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Chemical and physical stabilities are two key features considered in pharmaceutical development. Chemical stability is typically reported as a combination of potency and degradation product. For peptide products, it is common to measure physical stability via aggregation or fibrillation using the fluorescent reporter Thioflavin T. Executing stability studies is a lengthy process and requires extensive resources. To reduce the resources and shorten the process for stability studies during the development of a product, we introduce a machine learning based model for predicting the chemical stability over time using both the formulation conditions as well as the aggregation curve. In this work, we explore the relationships between the formulation, stability time point, and the measurements of chemical stability and achieve a coefficient of determination on a random test set of 0.945 and a mean absolute error (MAE) of 0.421 when using a multilayer perceptron (MLP) neural network for total degradation. Similarly, we achieve a coefficient of determination of 0.908 and an MAE of 1.435 when predicting the potency using a random forest model. When measurements of physical stability are included into the model, the MAE in the prediction of TD decreases to 0.148 for the MLP model. Using a similar strategy for the prediction of potency, the MAE decreases to 0.705 for the random forest model. Therefore, we can conclude two important points: first, chemical stability can be modeled using machine learning techniques and second there is a relationship between the physical stability of a peptide and its chemical stability.

show abstract

“…Machine learning (ML) has many potential uses within material discovery, including to reduce the cost of property calculation compared to carrying out computational simulations (especially via quantum mechanical methods), to focus experimental synthesis and measurement effort on the most promising materials reducing wasted laboratory effort [13,14], as well as to help facilitate the exploration of larger chemical space [15][16][17]. Apart from the widely reported ML models for molecular discovery, especially drug discovery, the applications of ML to porous materials such as MOFs have gained significant interest [18].…”

Section: Introductionmentioning

confidence: 99%

Explainable Graph Neural Networks for Organic Cages

Yuan

Szczypiński

Jelfs

2021

Preprint

View full text Add to dashboard Cite

The development of accurate and explicable machine learning models to predict the properties of topologically complex systems is a challenge in material science. Porous organic cages, a class of polycyclic molecular materials, have potential application in molecular separations, catalysis and encapsulation. For most applications of porous organic cages, having a permanent internal cavity in the absence of solvent, a property termed “shape persistency” is critical. Here, we report the development of Graph Neural Networks (GNNs) to predict the shape persistence of organic cages. Graph neural networks are a class of neural networks where the data, in our case that of organic cages, are represented by graphs. The performance of the GNN models was measured against a previously reported computational database of organic cages formed through a range of [4+6] reactions with a variety of reaction chemistries. The reported GNNs have an improved prediction accuracy and transferability compared to random forest predictions. Apart from the improvement in predictive power, we explored the explicability of the GNNs by computing the integrated gradient of the GNN input. The contribution of monomers and molecular fragments to the shape persistence of the organic cages could be quantitatively evaluated with integrated gradient. With the added explicability of the GNNs, it is possible not only to accurately predict the property of organic materials, but also to interpret the predictions of the deep learning models and provide structural insights to the discovery of future materials.

show abstract

Graph-based machine learning interprets and predicts diagnostic isomer-selective ion–molecule reactions in tandem mass spectrometry

Abstract: We combine mass spectrometry with machine learning that is predictive and explainable using chemical reactivity flowcharts for diagnostic ion–molecule reactions.

Cited by 15 publications

References 48 publications

Explainable graph neural networks for organic cages

Explainable graph neural networks for organic cages

Learning relationships between chemical and physical stability for drug development

Explainable Graph Neural Networks for Organic Cages

Contact Info

Product

Resources

About