Anna M. Hiszpanski

Evan R. Antoniuk, Peggy Li, Bhavya Kailkhura et al.

One of the grand challenges of utilizing machine learning for the discovery of innovative new polymers lies in the difficulty of accurately representing the complex structures of polymeric materials. Although a wide array of hand-designed polymer representations have been explored, there has yet to be an ideal solution for how to capture the periodicity of polymer structures, and how to develop polymer descriptors without the need for human feature design. In this work, we tackle these problems through the development of our periodic polymer graph representation. Our pipeline for polymer property predictions is comprised of our polymer graph representation that naturally accounts for the periodicity of polymers, followed by a message-passing neural network (MPNN) that leverages the power of graph deep learning to automatically learn chemically-relevant polymer descriptors. Across a diverse dataset of 10 polymer properties, we find that this polymer graph representation consistently outperforms hand-designed representations with a 20% average reduction in prediction error. Our results illustrate how the incorporation of chemical intuition through directly encoding periodicity into our polymer graph representation leads to a considerable improvement in the accuracy and reliability of polymer property predictions. We also demonstrate how combining polymer graph representations with message-passing neural network architectures can automatically extract meaningful polymer features that are consistent with human intuition, while outperforming human-derived features. This work highlights the advancement in predictive capability that is possible if using chemical descriptors that are specifically optimized for capturing the unique chemical structure of polymers.

Ioannis Tsaknakis, Bhavya Kailkhura, Sijia Liu et al.

Using deep learning (DL) to accelerate and/or improve scientific workflows can yield discoveries that are otherwise impossible. Unfortunately, DL models have yielded limited success in complex scientific domains due to large data requirements. In this work, we propose to overcome this issue by integrating the abundance of scientific knowledge sources (SKS) with the DL training process. Existing knowledge integration approaches are limited to using differentiable knowledge source to be compatible with first-order DL training paradigm. In contrast, our proposed approach treats knowledge source as a black-box in turn allowing to integrate virtually any knowledge source. To enable an end-to-end training of SKS-coupled-DL, we propose to use zeroth-order optimization (ZOO) based gradient-free training schemes, which is non-intrusive, i.e., does not require making any changes to the SKS. We evaluate the performance of our ZOO training scheme on two real-world material science applications. We show that proposed scheme is able to effectively integrate scientific knowledge with DL training and is able to outperform purely data-driven model for data-limited scientific applications. We also discuss some limitations of the proposed method and mention potentially worthwhile future directions.

Evan R. Antoniuk, Shehtab Zaman, Tal Ben-Nun et al.

Advances in deep learning and generative modeling have driven interest in data-driven molecule discovery pipelines, whereby machine learning (ML) models are used to filter and design novel molecules without requiring prohibitively expensive first-principles simulations. Although the discovery of novel molecules that extend the boundaries of known chemistry requires accurate out-of-distribution (OOD) predictions, ML models often struggle to generalize OOD. Furthermore, there are currently no systematic benchmarks for molecular OOD prediction tasks. We present BOOM, $\boldsymbol{b}$enchmarks for $\boldsymbol{o}$ut-$\boldsymbol{o}$f-distribution $\boldsymbol{m}$olecular property predictions -- a benchmark study of property-based out-of-distribution models for common molecular property prediction models. We evaluate more than 140 combinations of models and property prediction tasks to benchmark deep learning models on their OOD performance. Overall, we do not find any existing models that achieve strong OOD generalization across all tasks: even the top performing model exhibited an average OOD error 3x larger than in-distribution. We find that deep learning models with high inductive bias can perform well on OOD tasks with simple, specific properties. Although chemical foundation models with transfer and in-context learning offer a promising solution for limited training data scenarios, we find that current foundation models do not show strong OOD extrapolation capabilities. We perform extensive ablation experiments to highlight how OOD performance is impacted by data generation, pre-training, hyperparameter optimization, model architecture, and molecular representation. We propose that developing ML models with strong OOD generalization is a new frontier challenge in chemical ML model development. This open-source benchmark will be made available on Github.

Evan R. Antoniuk, Peggy Li, Nathan Keilbart et al.

Although generative models hold promise for discovering molecules with optimized desired properties, they often fail to suggest synthesizable molecules that improve upon the known molecules seen in training. We find that a key limitation is not in the molecule generation process itself, but in the poor generalization capabilities of molecular property predictors. We tackle this challenge by creating an active-learning, closed-loop molecule generation pipeline, whereby molecular generative models are iteratively refined on feedback from quantum chemical simulations to improve generalization to new chemical space. Compared against other generative model approaches, only our active learning approach generates molecules with properties that extrapolate beyond the training data (reaching up to 0.44 standard deviations beyond the training data range) and out-of-distribution molecule classification accuracy is improved by 79%. By conditioning molecular generation on thermodynamic stability data from the active-learning loop, the proportion of stable molecules generated is 3.5x higher than the next-best model.

Gyeong Hoon Yi, Jiwoo Choi, Hyeongyun Song et al.

Efficiently extracting data from tables in the scientific literature is pivotal for building large-scale databases. However, the tables reported in materials science papers exist in highly diverse forms; thus, rule-based extractions are an ineffective approach. To overcome this challenge, we present MaTableGPT, which is a GPT-based table data extractor from the materials science literature. MaTableGPT features key strategies of table data representation and table splitting for better GPT comprehension and filtering hallucinated information through follow-up questions. When applied to a vast volume of water splitting catalysis literature, MaTableGPT achieved an extraction accuracy (total F1 score) of up to 96.8%. Through comprehensive evaluations of the GPT usage cost, labeling cost, and extraction accuracy for the learning methods of zero-shot, few-shot and fine-tuning, we present a Pareto-front mapping where the few-shot learning method was found to be the most balanced solution owing to both its high extraction accuracy (total F1 score>95%) and low cost (GPT usage cost of 5.97 US dollars and labeling cost of 10 I/O paired examples). The statistical analyses conducted on the database generated by MaTableGPT revealed valuable insights into the distribution of the overpotential and elemental utilization across the reported catalysts in the water splitting literature.

Donald Loveland, Shusen Liu, Bhavya Kailkhura et al.

Graph neural network (GNN) explanations have largely been facilitated through post-hoc introspection. While this has been deemed successful, many post-hoc explanation methods have been shown to fail in capturing a model's learned representation. Due to this problem, it is worthwhile to consider how one might train a model so that it is more amenable to post-hoc analysis. Given the success of adversarial training in the computer vision domain to train models with more reliable representations, we propose a similar training paradigm for GNNs and analyze the respective impact on a model's explanations. In instances without ground truth labels, we also determine how well an explanation method is utilizing a model's learned representation through a new metric and demonstrate adversarial training can help better extract domain-relevant insights in chemistry.

Shusen Liu, Bhavya Kailkhura, Jize Zhang et al.

The scientific community has been increasingly interested in harnessing the power of deep learning to solve various domain challenges. However, despite the effectiveness in building predictive models, fundamental challenges exist in extracting actionable knowledge from deep neural networks due to their opaque nature. In this work, we propose techniques for exploring the behavior of deep learning models by injecting domain-specific actionable attributes as tunable "knobs" in the analysis pipeline. By incorporating the domain knowledge in a generative modeling framework, we are not only able to better understand the behavior of these black-box models, but also provide scientists with actionable insights that can potentially lead to fundamental discoveries.

Shusen Liu, Bhavya Kailkhura, Jize Zhang et al.

The scientific community has been increasingly interested in harnessing the power of deep learning to solve various domain challenges. However, despite the effectiveness in building predictive models, fundamental challenges exist in extracting actionable knowledge from the deep neural network due to their opaque nature. In this work, we propose techniques for exploring the behavior of deep learning models by injecting domain-specific actionable concepts as tunable ``knobs'' in the analysis pipeline. By incorporating the domain knowledge with generative modeling, we are not only able to better understand the behavior of these black-box models, but also provide scientists with actionable insights that can potentially lead to fundamental discoveries.

Bhavya Kailkhura, Brian Gallagher, Sookyung Kim et al.

Material scientists are increasingly adopting the use of machine learning (ML) for making potentially important decisions, such as, discovery, development, optimization, synthesis and characterization of materials. However, despite ML's impressive performance in commercial applications, several unique challenges exist when applying ML in materials science applications. In such a context, the contributions of this work are twofold. First, we identify common pitfalls of existing ML techniques when learning from underrepresented/imbalanced material data. Specifically, we show that with imbalanced data, standard methods for assessing quality of ML models break down and lead to misleading conclusions. Furthermore, we found that the model's own confidence score cannot be trusted and model introspection methods (using simpler models) do not help as they result in loss of predictive performance (reliability-explainability trade-off). Second, to overcome these challenges, we propose a general-purpose explainable and reliable machine-learning framework. Specifically, we propose a novel pipeline that employs an ensemble of simpler models to reliably predict material properties. We also propose a transfer learning technique and show that the performance loss due to models' simplicity can be overcome by exploiting correlations among different material properties. A new evaluation metric and a trust score to better quantify the confidence in the predictions are also proposed. To improve the interpretability, we add a rationale generator component to our framework which provides both model-level and decision-level explanations. Finally, we demonstrate the versatility of our technique on two applications: 1) predicting properties of crystalline compounds, and 2) identifying novel potentially stable solar cell materials.