Quantifying model generalization to out-of-distribution data has been a longstanding challenge in machine learning. Addressing this issue is crucial for leveraging machine learning in scientific discovery, where models must generalize to new molecules or materials. Current methods typically split data into train and test sets using various criteria — temporal, sequence identity, scaffold, or random cross-validation — before evaluating model performance. However, with so many splitting criteria available, existing approaches offer limited guidance on selecting the most appropriate one, and they do not provide mechanisms for incorporating prior knowledge about the target deployment distribution(s). To tackle this problem, we have developed a novel metric, AU-GOOD, which quantifies expected model performance under conditions of increasing dissimilarity between train and test sets, while also accounting for prior knowledge about the target deployment distribution(s), when available. This metric is broadly applicable to biochemical entities, including proteins, small molecules, nucleic acids, or cells; as long as a relevant similarity function is defined for them. Recognizing the wide range of similarity functions used in biochemistry, we propose criteria to guide the selection of the most appropriate metric for partitioning. We also introduce a new partitioning algorithm that generates more challenging test sets, and we propose statistical methods for comparing models based on AU-GOOD. Finally, we demonstrate the insights that can be gained from this framework by applying it to two different use cases: developing predictors for pharmaceutical properties of small molecules, and using protein language models as embeddings to build biophysical property predictors.

Foundation Models (FMs) hold transformative potential to accelerate scientific discovery, yet reaching their full capacity in complex, highly multimodal domains such as genomics, drug discovery, and materials science requires a deeper consideration of the contextual nature of the scientific knowledge . We revisit the synergy between FMs and Multimodal Knowledge Graph (MKG) representation and learning, exploring their potential to enhance predictive and generative tasks in biomedical contexts like drug discovery. We seek to exploit MKGs to improve generative AI models’ ability to capture intricate domain-specific relations and facilitate multimodal fusion. This integration promises to accelerate discovery workflows by providing more meaningful multimodal knowledge-enhanced representations and contextual evidence. Despite this potential, challenges and opportunities remain, including fusing multiple sequential, structural and knowledge modalities and models leveraging the strengths of each; developing scalable architectures for multi-task multi-dataset learning; creating end-to-end workflows to enhance the trustworthiness of biomedical FMs using knowledge from heterogeneous datasets and scientific literature; the domain data bottleneck and the lack of a unified representation between natural language and chemical representations; and benchmarking, specifically the transfer learning to tasks with limited data (e.g., unseen molecules and proteins, rear diseases). Finally, fostering openness and collaboration is key to accelerate scientific breakthroughs.

Recent research on predicting the binding affinity between drug molecules and proteins use representations learned, through unsupervised learning techniques, from large databases of molecule SMILES and protein sequences. While these representations have significantly enhanced the predictions, they are usually based on a limited set of modalities, and they do not exploit available knowledge about existing relations among molecules and proteins. Our study reveals that enhanced representations, derived from multimodal knowledge graphs describing relations among molecules and proteins, lead to state-of-the-art results in well-established benchmarks (first place in the leaderboard for Therapeutics Data Commons benchmark ``Drug-Target Interaction Domain Generalization Benchmark", with an improvement of 8 points with respect to previous best result). Moreover, our results significantly surpass those achieved in standard benchmarks by using conventional pre-trained representations that rely only on sequence or SMILES data. We release our multimodal knowledge graphs, integrating data from seven public data sources, and which contain over 30 million triples. Pretrained models from our proposed graphs and benchmark task source code are also released.

MOTIVATION: Automated machine learning (AutoML) solutions can bridge the gap between new computational advances and their real-world applications by enabling experimental scientists to build their own custom models. We examine different steps in the development life-cycle of peptide bioactivity binary predictors and identify key steps where automation cannot only result in a more accessible method, but also more robust and interpretable evaluation leading to more trustworthy models. RESULTS: We present a new automated method for drawing negative peptides that achieves better balance between specificity and generalization than current alternatives. We study the effect of homology-based partitioning for generating the training and testing data subsets and demonstrate that model performance is overestimated when no such homology correction is used, which indicates that prior studies may have overestimated their performance when applied to new peptide sequences. We also conduct a systematic analysis of different protein language models as peptide representation methods and find that they can serve as better descriptors than a naive alternative, but that there is no significant difference across models with different sizes or algorithms. Finally, we demonstrate that an ensemble of optimized traditional machine learning algorithms can compete with more complex neural network models, while being more computationally efficient. We integrate these findings into AutoPeptideML, an easy-to-use AutoML tool to allow researchers without a computational background to build new predictive models for peptide bioactivity in a matter of minutes. AVAILABILITY AND IMPLEMENTATION: Source code, documentation, and data are available at https://github.com/IBM/AutoPeptideML and a dedicated web-server at http://peptide.ucd.ie/AutoPeptideML. A static version of the software to ensure the reproduction of the results is available at https://zenodo.org/records/13363975.

In many machine learning tasks, models are trained to predict structure data such as graphs. For example, in natural language processing, it is very common to parse texts into dependency trees or abstract meaning representation (AMR) graphs. On the other hand, ensemble methods combine predictions from multiple models to create a new one that is more robust and accurate than individual predictions. In the literature, there are many ensembling techniques proposed for classification or regression problems, however, ensemble graph prediction has not been studied thoroughly. In this work, we formalize this problem as mining the largest graph that is the most supported by a collection of graph predictions. As the problem is NP-Hard, we propose an efficient heuristic algorithm to approximate the optimal solution. To validate our approach, we carried out experiments in AMR parsing problems. The experimental results demonstrate that the proposed approach can combine the strength of state-of-the-art AMR parsers to create new predictions that are more accurate than any individual models in five standard benchmark datasets.

Nowadays, most users carry high computing power mobile devices where speech recognition is certainly one of the main technologies available in every modern smartphone, although battery draining and application performance (resource shortage) have a big impact on the experienced quality. Shifting applications and services to the cloud may help to improve mobile user satisfaction as demonstrated by several ongoing efforts in the mobile cloud area. However, the quality of speech recognition is still not sufficient in many complex cases to replace the common hand written text, especially when prompt reaction to short-term provisioning requests is required. To address the new scenario, this paper proposes a mobile cloud infrastructure to support the extraction of semantics information from speech recognition in the Social Care domain, where carers have to speak about their patients conditions in order to have reliable notes used afterward to plan the best support. We present not only an architecture proposal, but also a real prototype that we have deployed and thoroughly assessed with different queries, accents, and in presence of load peaks, in our experimental mobile cloud Platform as a Service (PaaS) testbed based on Cloud Foundry.

Efficiently detecting conversation threads from a pool of messages, such as social network chats, emails, comments to posts, news etc., is relevant for various applications, including Web Marketing, Information Retrieval and Digital Forensics. Existing approaches focus on text similarity using keywords as features that are strongly dependent on the dataset. Therefore, dealing with new corpora requires further costly analyses conducted by experts to find out new relevant features. This paper introduces a novel method to detect threads from any type of conversational texts overcoming the issue of previously determining specific features for each dataset. To automatically determine the relevant features of messages we map each message into a three dimensional representation based on its semantic content, the social interactions in terms of sender/recipients and its timestamp; then clustering is used to detect conversation threads. In addition, we propose a supervised approach to detect conversation threads that builds a classification model which combines the above extracted features for predicting whether a pair of messages belongs to the same thread or not. Our model harnesses the distance measure of a message to a cluster representing a thread to capture the probability that a message is part of that same thread. We present our experimental results on seven datasets, pertaining to different types of messages, and demonstrate the effectiveness of our method in the detection of conversation threads, clearly outperforming the state of the art and yielding an improvement of up to a 19%.

We present an approach to access and consolidate complex information spanning multiple specialist domains and make it available to non-experts. We are using a combination of business rules and contextual exploration to reduce interface complexity and improve consumability. We present a use case and a prototype on top of a real-world enterprise solution for coordinating Social care and Health care. We evaluate our system through a user study. Our results indicate that our approach reduces the time required to obtain business results compared to a baseline graph exploration approach.

The dynamics of social events happening in large metropolitan areas are extremely complex. Location-based user generated data could be an exceptionally rich source of information about events, however the vastness and the heterogeneity of such information makes it almost impossible for city managers to have a comprehensive view. Some events are planned and advertised only among restricted communities, for many it is problematic to estimate the number of attendees, and in general it is very difficult to have information about citizen mobility in areas nearby or where events happen. In this paper, we describe EXSED, an interactive intelligent tool to support the visual exploration of social events dynamics along the spatial, temporal and organizational dimensions. Major functionalities include: (i) studying events that are either officially/unofficially planned and with global/local scope, (ii) extract discussion topics for an event and investigate their temporal and spatial profile, (iii) estimate event attendees and correlate their mobility patterns with the event evolution. We present preliminary results of a field study with officials from Dublin City Council (Ireland), who evaluated our tool on real world user-generated tweeter data and events in the city.

Today plenty of data is emerging from various city systems. Beyond the classical Web resources, large amounts of data are retrieved from sensors, devices, social networks, governmental applications, or service networks. In such a diversity of information, answering specific information needs of city inhabitants requires holistic IR techniques, capable of harnessing different types of city data and turned it into actionable insights to answer different queries. This tutorial will present deep insights, challenges, opportunities and techniques to make heterogeneous city data searchable and show how emerging IR techniques models can be employed to retrieve relevant information for the citizens.

This work investigates the process of selecting, extracting and reorganizing content from Semantic Web information sources, to produce an ontology meeting the specifications of a particular domain and/or task. The process is combined with traditional text-based ontology learning methods to achieve tolerance to knowledge incompleteness. The paper describes the approach and presents experiments in which an ontology was built for a diet evaluation task. Although the example presented concerns the specific case of building a nutritional ontology, the methods employed are domain independent and transferrable to other use cases.

PowerAqua is a Question Answering system, which takes as input a natural language query and is able to return answers drawn from relevant semantic resources found anywhere on the Semantic Web. In this paper we provide two novel contributions: First, we detail a new component of the system, the Triple Similarity Service, which is able to match queries effectively to triples found in different ontologies on the Semantic Web. Second, we provide a first evaluation of the system, which in addition to providing data about PowerAqua's competence, also gives us important insights into the issues related to using the Semantic Web as the target answer set in Question Answering. In particular, we show that, despite the problems related to the noisy and incomplete conceptualizations, which can be found on the Semantic Web, good results can already be obtained.