Pesticides benefit agriculture by increasing crop yield, quality, and security. However, pesticides may inadvertently harm bees, which are valuable as pollinators. Thus, candidate pesticides in development pipelines must be assessed for toxicity to bees. Leveraging a dataset of 382 molecules with toxicity labels from honey bee exposure experiments, we train a support vector machine (SVM) to predict the toxicity of pesticides to honey bees. We compare two representations of the pesticide molecules: (i) a random walk feature vector listing counts of length- L walks on the molecular graph with each vertex- and edge-label sequence and (ii) the Molecular ACCess System (MACCS) structural key fingerprint (FP), a bit vector indicating the presence/absence of a list of pre-defined subgraph patterns in the molecular graph. We explicitly construct the MACCS FPs but rely on the fixed-length- L random walk graph kernel (RWGK) in place of the dot product for the random walk representation. The L-RWGK-SVM achieves an accuracy, precision, recall, and F1 score (mean over 2000 runs) of 0.81, 0.68, 0.71, and 0.69, respectively, on the test data set—with L = 4 being the mode optimal walk length. The MACCS-FP-SVM performs on par/marginally better than the L-RWGK-SVM, lends more interpretability, but varies more in performance. We interpret the MACCS-FP-SVM by illuminating which subgraph patterns in the molecules tend to strongly push them toward the toxic/non-toxic side of the separating hyperplane.

Pesticides benefit agriculture by increasing crop yield, quality, and security. However, pesticides may inadvertently harm bees, which are agriculturally and ecologically vital as pollinators. The development of new pesticides---driven by pest resistance to and demands to reduce negative environmental impacts of incumbent pesticides---necessitates assessments of pesticide toxicity to bees. We leverage a data set of 382 molecules labeled from honey bee toxicity experiments to train a classifier that predicts the toxicity of a new pesticide molecule to honey bees. Traditionally, the first step of a molecular machine learning task is to explicitly convert molecules into feature vector representations for input to the classifier. Instead, we (i) adopt the fixed-length random walk graph kernel to express the similarity between any two molecular graphs and (ii) use the kernel trick to train a support vector machine (SVM) to classify the bee toxicity of pesticides represented as molecular graphs. We assess the performance of the graph-kernel-SVM classifier under different walk lengths used to describe the molecular graphs. The optimal classifier, with walk length 5, achieves an (mean over 100 runs) accuracy, precision, and recall of 0.83, 0.71, and 0.72 on a test data set.

Pesticides benefit agriculture by increasing crop yield, quality, and security. However, pesticides may inadvertently harm bees, which are agriculturally and ecologically vital as pollinators. The development of new pesticides---driven by pest resistance to and demands to reduce negative environmental impacts of incumbent pesticides---necessitates assessments of pesticide toxicity to bees. We leverage a data set of 382 molecules labeled from honey bee toxicity experiments to train a classifier that predicts the toxicity of a new pesticide molecule to honey bees. Traditionally, the first step of a molecular machine learning task is to explicitly convert molecules into feature vector representations for input to the classifier. Instead, we (i) adopt the fixed-length random walk graph kernel to express the similarity between any two molecular graphs and (ii) use the kernel trick to train a support vector machine (SVM) to classify the bee toxicity of pesticides represented as molecular graphs. We assess the performance of the graph-kernel-SVM classifier under different walk lengths used to describe the molecular graphs. The optimal classifier, with walk length 4, achieves a (mean over 100 runs) accuracy, precision, recall, and F1 score of 0.82, 0.69, 0.74, and 0.71 on the test data set.

Pesticides benefit agriculture by increasing crop yield, quality, and security. However, pesticides may inadvertently harm bees, which are valuable as pollinators. Thus, candidate pesticides in development pipelines must be assessed for toxicity to bees. Leveraging a data set of 382 molecules with toxicity labels from honey bee exposure experiments, we train a support vector machine (SVM) to predict the toxicity of pesticides to honey bees. We compare two representations of the pesticide molecules: (i) a random walk feature vector listing counts of length-L walks on the molecular graph with each vertex- and edge-label sequence and (ii) the MACCS structural key fingerprint (FP), a bit vector indicating the presence/absence of a list of pre-defined subgraph patterns in the molecular graph. We explicitly construct the MACCS FPs, but rely on the fixed-length-L random walk graph kernel (RWGK) in place of the dot product for the random walk representation. The L-RWGK-SVM achieves an accuracy, precision, recall, and F1 score (mean over 2000 runs) of 0.81, 0.68, 0.71, and 0.69 on the test data set---with L=4 the mode optimal walk length. The MACCS-FP-SVM performs on par/marginally better than the L-RWGK-SVM, lends more interpretability, but varies more in performance. We interpret the MACCS-FP-SVM by illuminating which subgraph patterns in the molecules tend to strongly push them towards the toxic/non-toxic side of the separating hyperplane.

Pesticides benefit agriculture by increasing crop yield, quality, and security. However, pesticides may inadvertently harm bees, which are agriculturally and ecologically vital as pollinators. The development of new pesticides---driven by pest resistance to and demands to reduce negative environmental impacts of incumbent pesticides---necessitates assessments of pesticide toxicity to bees. We leverage a data set of 382 molecules labeled from honey bee toxicity experiments to train a classifier that predicts the toxicity of a new pesticide molecule to honey bees. Traditionally, the first step of a molecular machine learning task is to explicitly convert molecules into feature vector representations for input to the classifier. Instead, we (i) adopt the fixed-length random walk graph kernel to express the similarity between any two molecular graphs and (ii) use the kernel trick to train a support vector machine (SVM) to classify the bee toxicity of pesticides represented as molecular graphs. We assess the performance of the graph-kernel-SVM classifier under different walk lengths used to describe the molecular graphs. The optimal classifier, with walk length 5, achieves an (mean over 100 runs) accuracy, precision, and recall of 0.83, 0.71, and 0.72 on a test data set.