A5054571851- Computational Drug Discovery Methods
- Machine Learning in Materials Science
- Protein Structure and Dynamics
- Chemical Reactions and Isotopes
- Machine Learning in Bioinformatics
- vaccines and immunoinformatics approaches
- Synthesis and biological activity
- Machine Learning in Healthcare
- Machine Learning and Data Classification
- Trauma, Hemostasis, Coagulopathy, Resuscitation
- Medical Coding and Health Information
- Catalysis and Oxidation Reactions
- Software Engineering Research
- Cognitive Science and Education Research
- Innovative Microfluidic and Catalytic Techniques Innovation
- Topic Modeling
- Gene expression and cancer classification
- Genomics and Phylogenetic Studies
- Financial Markets and Investment Strategies
- Blood donation and transfusion practices
- Emergency and Acute Care Studies
- Heart Failure Treatment and Management
- SARS-CoV-2 and COVID-19 Research
- Sepsis Diagnosis and Treatment
- Click Chemistry and Applications
Johannes Kepler University of Linz
2020-2024
Microsoft Research (United Kingdom)
2024
Automated synthesis planning has recently re-emerged as a research area at the intersection of chemistry and machine learning. Despite appearance steady progress, we argue that imperfect benchmarks inconsistent comparisons mask systematic shortcomings existing techniques, unnecessarily hamper progress. To remedy this, present library with an extensive benchmarking framework, called SYNTHESEUS, which promotes best practice by default, enabling consistent meaningful evaluation single-step...
We introduce a modern Hopfield network with continuous states and corresponding update rule. The new can store exponentially (with the dimension of associative space) many patterns, retrieves pattern one update, has small retrieval errors. It three types energy minima (fixed points update): (1) global fixed point averaging over all (2) metastable subset (3) which single pattern. rule is equivalent to attention mechanism used in transformers. This equivalence enables characterization heads...
Due to the current severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic, there is an urgent need for novel therapies and drugs. We conducted a large-scale virtual screening small molecules that are potential CoV-2 inhibitors. To this end, we utilized ChemAI, deep neural network trained on more than 220M data points across 3.6M from three public drug-discovery databases. With screened ranked one billion ZINC database favourable effects against CoV-2. then reduced result...
Activity and property prediction models are the central workhorses in drug discovery materials sciences, but currently they have to be trained or fine-tuned for new tasks. Without training fine-tuning, scientific language could used such low-data tasks through their announced zero- few-shot capabilities. However, predictive quality at activity is lacking. In this work, we envision a novel type of model that able adapt inference time, via understanding textual information describing task. To...
The COVID-19 pandemic continues to pose a substantial threat human lives and is likely do so for years come. Despite the availability of vaccines, searching efficient small-molecule drugs that are widely available, including in low- middle-income countries, an ongoing challenge. In this work, we report results open science community effort, "Billion molecules against challenge", identify inhibitors SARS-CoV-2 or relevant receptors. Participating teams used wide variety computational methods...
A central task in computational drug discovery is to construct models from known active molecules find further promising for subsequent screening. However, typically only very few are known. Therefore, few-shot learning methods have the potential improve effectiveness of this critical phase process. We introduce a new method discovery. Its main idea enrich molecule representation by knowledge about context or reference molecules. Our novel concept enrichment associate both support set and...
Patient admission is a decision relying on sparsely available data. This study aims to provide prediction models for discharge versus ward observation or intensive care, and 30 day-mortality patients triaged with the Manchester Triage System.This single-centre, observational, retrospective cohort from data within ten minutes of patient presentation at interdisciplinary emergency department Kepler University Hospital, Linz, Austria. We trained machine learning including Random Forests Neural...
Automated Synthesis Planning has recently re-emerged as a research area at the intersection of chemistry and machine learning. Despite appearance steady progress, we argue that imperfect benchmarks inconsistent comparisons mask systematic shortcomings existing techniques, unnecessarily hamper progress. To remedy this, present synthesis planning library with an extensive benchmarking framework, called syntheseus, which promotes best practice by default, enabling consistent meaningful...
Finding synthesis routes for molecules of interest is an essential step in the discovery new drugs and materials. To find such routes, computer-assisted planning (CASP) methods are employed which rely on a model chemical reactivity. In this study, we single-step retrosynthesis template-based approach using modern Hopfield networks (MHNs). We adapt MHNs to associate different modalities, reaction templates molecules, allows leverage structural information about templates. This significantly...
In this paper, we apply machine learning models to execute certain short-option strategies on the S&P500. particular, formulate and focus a supervised classification task which decides if plain short straddle S&P500 should be executed or not daily basis. We describe our used framework present an overview of evaluation metrics for different models. Using standard techniques systematic hyperparameter search, find statistically significant advantages gradient tree boosting algorithm is...
Language models for biological and chemical sequences enable crucial applications such as drug discovery, protein engineering, precision medicine. Currently, these language are predominantly based on Transformer architectures. While Transformers have yielded impressive results, their quadratic runtime dependency the sequence length complicates use long genomic in-context learning proteins sequences. Recently, recurrent xLSTM architecture has been shown to perform favorably compared modern...
The COVID-19 pandemic continues to pose a substantial threat human lives and is likely do so for years come. Despite the availability of vaccines, searching efficient small-molecule drugs that are widely available, including in low- middle-income countries, an ongoing challenge. In this work, we report results community effort, “Billion molecules against Covid-19 challenge”, identify inhibitors SARS-CoV-2 or relevant receptors. Participating teams used wide variety computational methods...
Transfusion of packed red blood cells (pRBCs) is still associated with risks. This study aims to determine whether renal function deterioration in the context individual transfusions patients can be predicted using machine learning. Recipient and donor characteristics linked increased risk are identified.
Due to the current severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic, there is an urgent need for novel therapies and drugs. We conducted a large-scale virtual screening small molecules that are potential CoV-2 inhibitors. To this end, we utilized "ChemAI", deep neural network trained on more than 220M data points across 3.6M from three public drug-discovery databases. With ChemAI, screened ranked one billion ZINC database favourable effects against CoV-2. then reduced...
In this working paper we present our current progress in the training of machine learning models to execute short option strategies on S&P500. As a first step, is breaking problem down supervised classification task decide if straddle S&P500 should be executed or not daily basis. We describe used framework and an overview over evaluation metrics different models. preliminary work, using standard techniques without hyperparameter search, find no statistically significant outperformance simple...