Investigation of the structure-odor relationship using a Transformer model

Zheng, Xuejun; Tomiura, Yoichi; Hayashi, Kenshi

doi:10.1186/s13321-022-00671-y

Cited by 4 publications

(9 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To obtain non-transfer features, we use state-of-the-art by zheng 53 . We train the SMILEStransformer 53 from scratch using our perceptual training data. To evaluate whether the model's performance is limited by the amount of data, we conduct training on progressively larger volumes of data from 20% to 80% of the whole dataset.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Optimizing Learning Across Multimodal Transfer Features for Modelling Olfactory Perception

Shin,

Pei,

Kumari

et al. 2024

Preprint

View full text Add to dashboard Cite

For humans and other animals, the sense of smell provides crucial information in many situations of everyday life. Still, the study of olfactory perception has received only limited attention outside of the biological sciences. From an AI perspective, the complexity of the interactions between olfactory receptors and volatile molecules and the scarcity of comprehensive olfactory datasets present unique challenges in this sensory domain. Previous works have explored the relationship between molecular structure and odor descriptors using fully supervised training approaches. However, these methods are data-intensive and poorly generalized due to labeled data scarcity, particularly for rare-class samples. Our study partially tackles the challenges of data scarcity and label skewness through multimodal transfer learning. We investigate the potential of large molecular foundation models trained on extensive unlabeled molecular data to effectively model olfactory perception. Additionally, we explore the integration of different molecular representations, including molecular graphs and text-based SMILES encodings, to achieve data efficiency and generalization of the learned model, particularly on sparsely represented classes. By leveraging complementary representations, we aim to learn robust perceptual features of odorants. However, we observe that traditional methods of combining modalities do not yield substantial gains in high-dimensional skewed label spaces. To address this challenge, we introduce a novel \emph{label-balancer} technique specifically designed for high-dimensional multi-label and multi-modal training. The label-balancer technique distributes learning objectives across modalities to optimize collaboratively for distinct subsets of labels. Our results suggest that multi-modal transfer features learned using the label-balancer technique are more effective and robust, surpassing the capabilities of traditional uni- or multi-modal approaches, particularly on rare-class samples.

show abstract

Section: Resultsmentioning

confidence: 99%

“…Figure 5a shows the performance comparison between perceptual features learned with and without using a pre-trained SMILES-transformer model. To obtain non-transfer features, we use state-of-the-art by zheng 53 . We train the SMILEStransformer 53 from scratch using our perceptual training data.…”

Section: Resultsmentioning

confidence: 99%

Optimizing Learning Across Multimodal Transfer Features for Modelling Olfactory Perception

Shin,

Pei,

Kumari

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…‘Woody’ odorants are associated with rigid bulky hydrocarbon skeletons [ 23 ]. Whereas a subgroup discovery algorithm revealed the rule that ‘woody’ molecules are hydrophobic and rather not cyclic nor aromatic [ 24 ], investigations using a Transformer model suggested that woody molecules are often ring structures [ 26 ]. This is in accordance with our results, where cyclohexane structures were assigned the second highest influence values whereas aromatic structures scored low.…”

Section: Resultsmentioning

confidence: 99%

“…Though many advances in odor prediction have been achieved in recent years [ 10 – 20 ], we unfortunately still know little about the relationship between a molecule’s structure and its odor [ 21 – 23 ] to an extent where we can provide chemists with a toolbox for designing molecular structures with a specific odor in mind. However, sophisticated computational methods have led to new insights into these relationships [ 24 – 26 ] and allow prediction whether a molecule is odorous at all [ 27 ]. Adding to the hurdles in the field, there is dispute over the dimensionality of the odor space [ 7 , 28 ].…”

Section: Introductionmentioning

confidence: 99%

OWSum: algorithmic odor prediction and insight into structure-odor relationships

et al. 2023

View full text Add to dashboard Cite

We derived and implemented a linear classification algorithm for the prediction of a molecule’s odor, called Olfactory Weighted Sum (OWSum). Our approach relies solely on structural patterns of the molecules as features for algorithmic treatment and uses conditional probabilities combined with tf-idf values. In addition to the prediction of molecular odor, OWSum provides insights into properties of the dataset and allows to understand how algorithmic classifications are reached by quantitatively assigning structural patterns to odors. This provides chemists with an intuitive understanding of underlying interactions. To deal with ambiguities of the natural language used to describe odor, we introduced descriptor overlap as a metric for the quantification of semantic overlap between descriptors. Thus, grouping of descriptors and derivation of higher-level descriptors becomes possible. Our approach poses a large leap forward in our capabilities to understand and predict molecular features.

show abstract

“…With the results of machine learning, the study will further explore the chemical structure and mechanism of action of green odor molecules, revealing the potential laws of their functioning. Additionally, by using an internet service that is freely available to everyone, the binary classifiers developed through the application of machine learning can be employed, saving time and money [ 12 , 13 , 14 , 15 , 16 , 17 ]. This study presents a new breakthrough in the field of green odor research by applying machine learning for the first time to green odor molecular prediction.…”

Section: Introductionmentioning

confidence: 99%

Mlp4green: A Binary Classification Approach Specifically for Green Odor

Yang,

Qian,

et al. 2024

IJMS

View full text Add to dashboard Cite

Fresh green leaves give off a smell known as “green odor.” It has antibacterial qualities and can be used to attract or repel insects. However, a common method for evaluating green odor molecules has never existed. Machine learning techniques are widely used in research to forecast molecular attributes for binary classification. In this work, the green odor molecules were first trained and learned using machine learning methods, and then clustering analysis and molecular docking were performed to further explore their molecular characteristics and mechanisms of action. For comparison, four algorithmic models were employed, MLP performed the best in all metrics, including Accuracy, Precision, Average Precision, Matthews coefficient, and Area under curve. We determined by difference analysis that, in comparison to non-green odor molecules, green odor molecules have a lower molecular mass and fewer electrons. Based on the MLP algorithm, we constructed a binary classification prediction website for green odors. The first application of deep learning techniques to the study of green odor molecules can be seen as a signal of a new era in which green odor research has advanced into intelligence and standardization.

show abstract

Investigation of the structure-odor relationship using a Transformer model

Cited by 4 publications

References 31 publications

Optimizing Learning Across Multimodal Transfer Features for Modelling Olfactory Perception

Optimizing Learning Across Multimodal Transfer Features for Modelling Olfactory Perception

OWSum: algorithmic odor prediction and insight into structure-odor relationships

Mlp4green: A Binary Classification Approach Specifically for Green Odor

Contact Info

Product

Resources

About