Integrating knowledge from scientific literature is essential in biomedical research. However, the rapid growth of scientific literature makes staying up to date increasingly challenging. Retrieval-Augmented Generation (RAG) offers a promising framework, but its effectiveness in specialized biomedical domains remains unclear. In this work, we propose a two-stage retrieval pipeline for RAG, with a focus on Bordetella pertussis as a case study. Our method first applies hard filtering with synonym expansion to eliminate irrelevant passages, and then performs hybrid search, followed by reranking. We evaluate our approach using a dataset of 58 pertussis-related queries with automatic relevance judgments from multiple large language models (LLMs). Experimental results show that our pipeline improves MAP@10 by 13.4-20.4 points compared with existing methods and achieves the highest MRR@10. Furthermore, consistent improvements across different LLMs highlight the effectiveness of our approach.

For non-native English speakers, learning programming presents unique challenges, particularly in languages like Python, which heavily rely on English syntax and documentation. This study explores the challenges faced by non-native English-speaking Japanese university students enrolled in introductory Python programming courses. A survey was held to examined various aspects of their learning experience, including their programming proficiency, English fluency, use of translation tools, and the impact of bilingual resources on their learning. Clustering analysis was employed to group the participants into three distinct clusters: beginners, intermediate learners, and advanced learners. Cluster 0 (beginners) had low programming experience and English fluency, relying on non-English materials and translation tools. Cluster 1 (intermediate) showed moderate skills, using bilingual resources and translation aids. Cluster 2 (advanced) had higher proficiency but faced challenges with debugging and code optimization. Our findings highlight the need for tailored resources to support students navigating English-based programming environments.

Stroke is a serious cerebrovascular condition, in which brain cells die due to an abrupt blockage of arteries supplying blood and oxygen or due to bleeding in the brain tissue when a blood vessel bursts or ruptures. Because stroke occurs suddenly in most people, prevention is oftentimes difficult. In Japan, this condition is one of the major causes of death, which is associated with high medical cost, especially among the society’s aging population. Therefore, stroke prediction and treatment is important. Stroke incidences can be avoided by a preventive treatment based on the risk of onset. However, since judgment of the onset risk largely depends on the individual experience and skill of the doctor, a highly accurate prediction method that is independent of the doctor’s experience and skill is the focus of this study. The target of prediction for this research is subarachnoid hemorrhage that is part of stroke. Logistic regression and support vector machine that predict cerebral aneurysm rupture by machine learning using combined medical data and cerebral blood-flow-simulation data were employed to analyze 338 cerebral aneurysm samples (35 ruptured, 303 unruptured). SMOTE algorithm solved the imbalance of data, while the SelectKBest algorithm was used to extract important features from the total 70 features obtained from both data. Out of the 27 important features extracted, 40% belonged to the medical data and the remaining 60% were from the blood-flow-simulation data. Using logistic regression as a classification model, we found the sensitivity of 0.64 and the specificity of 0.85. The results validated the possibility of a highly accurate method of cerebral aneurysm rupture prediction by machine learning using engineering information obtained from mechanical simulation.

This paper explores machine learning application in the case of drug discovery. We apply extreme gradient boosting and K-nearest neighbor to biomedical data and it signiflcantly outperform former studies using feature selection and proper tuning parameters. The novel application motivated by a recent circumstance that there is a need for rapid development of radio-protectors. It mainly targets the DNA of growing cancer cells, whereas it has adverse side effects, including p53-induced apoptosis of normal tissues and cells. It considered that p53 would be a target for therapeutic and mitigated radioprotection to escape from the apoptotic fate. On the other hand, many types of compounds contain several level of toxicity, so it is important to consider not only radiation protection but also the level of toxicity of candidate compounds for radioprotectors. Compounds of radio-protectors that have low toxicity and high radiation protection are expected. It is possible to do efficiently the compounds discovery using machine learning. This study predicts compounds of radioprotectors using plural machine learning methods, Extreme Gradient Boosting, K-nearest neighbor, SVM and Random Forest. We compare these methods and suggest proper methods to predict radioprotectors.

An overview of notable ILP areas, focusing on three invited talks at ILP 2015, two best student papers and the panel discussion on "ILP 25 Years".

Machine learning has been often used for drug discovery in recent years. Inductive logic programming (ILP), that can express common features of each data in a qualitative, is one of the machine learning methods. The advantage of ILP is that the classification model is clear compared with other machine learning methods such as Support Vector Machine (SVM) because ILP classifies inhibitors using generated rules. ILP is allowed to learn structure of the compounds so that we can draw the common structure of the ligands from the generated rules. In this method, the data of ligands and decoys are collected from “A Database of Useful Decoys: Enhanced” (DUD-E). ILP provides classification model called a rule by learning these data. This study applies the ILP algorithm to the virtual screening of inhibitors of carbonic anhydrase II (CAH2). We demonstrate its performance by classifying ligands and decoys which aren't included in DUD-E. Our results show that ILP has the performance equivalent to SVM which is known for its high classification performance. In addition, this paper shows that ILP can derive rules explaining structural features of CAH2 ligands. Several of these rules are consistent with known property of CAH2 ligand. This paper demonstrate that ILP has high classification performance and clear classification model. Our method is useful for generating rules for ligand design.

Radiation therapy and some chemotherapeutic agents mainly target the DNA of growing cancer cells, whereas these therapies have adverse side effects, including p53-induced apoptosis of normal tissues and cells. It is considered that p53 would be a target for therapeutic and mitigative radioprotection to escape from the apoptotic fate. So far, only three radioprotective p53 inhibitors have been reported, namely, pifithrin-α (PFTα), pifithrin-μ (PFTμ), and sodium orthovanadate (vanadate), which protect mice from acute lethality due to hematopoietic syndrome, indicating that pharmacologically temporary suppression of p53 effectively minimize the radiation damage. In this study, we examined the inhibitory activity of some zinc(II) chelators against radiation-induced apoptosis of MOLT-4 cells, based on the assumption that the binding of these compounds to zinc(II) in p53 proteins or removal of zinc(II) from the protein would temporally inhibit the function of p53. However, we have had some problems. The development of drug has been slow, due to the time required and the high cost of screening candidate compounds. It is possible to efficiently search for drugs by using machine learning. So we predict compounds that radioprotectors using Random Forest to study compound futures and using other machine learning methods for comparison with Random Forest. Procedure of learning is as follows: First, compounds were divided into several groups based on the toxicity and protection capability. Next, it was performed classification using machine learning. These results may contribute to discover of new radioprotectors.

In this paper we describe a Web-based method for communication with and control of heterogeneous robots in a unified manner, including mobile robots and vision sensors. We base this method on recent Web technologies such as browsers, Java language and socket communication. It allows users to connect to robots through a Web server, using their hand-held computers, and to monitor and control the robots via various input devices. As a model of an everyday working environment, we configured mobile robots and cameras that work autonomously and cooperatively. The Web technologies provide a general framework for integrating robot control, communication between robots and humans, and Web-page access. This framework can extend the robot-based intelligent office as a new direction of the intranet.

The paper describes a concurrent logic specification language for developing programs that control multiple robots. This language is based on guarded Horn clauses and specifications are atomically transformed into concurrent logic programs, yielding executable specifications. They makes it possible to easily implement complex tasks such as concurrency control, communication and multiagent-type profile, solving for multiple robots. An experiment on program development for multiple manipulators was undertaken to show advantages of the specification language. The resulting observations are twofold. First, specifications are more abstract and natural than conventional procedure-oriented programs. Second the amounts of specifications are very shorter than that of the programs developed by the underlying concurrent logic programs are then compiled into C programs with little overhead the proposed language can be useful as a multiple robot programming language with the efficiency for both program execution and developed.