Accelerate Science pursues research at the interface of AI and the sciences, generating new scientific insights and developing AI methods that can be deployed to advance scientific knowledge. This research is carried out in partnership with a community of scientists and AI specialists passionate about the use of AI to benefit science and society. Recent publications from the team are listed below.
Solving Schrödinger Bridges via Maximum Likelihood
Vargas, F., Thorodoff, P., Lamacraft, A. and Lawrence, ND. (2021) Solving Schrödinger Bridges via Maximum Likelihood. Entropy, 23 (9) 1134
This paper proposes a numerical procedure to estimate Schrödinger bridge problems using Gaussian processes.Read more
Deep learning for bioimage analysis in developmental biology
Hallou, A., Yevick, H., Dumitrascu, B. and Uhlmann, V. (2021) Deep learning for bioimage analysis in developmental biology. Development (2021)148 (18), dev199616.
This review introduces key concepts in deep learning and its application to bio-image analysis, exploring how researchers can integrate these techniques into their work.Read more
Ranking the information content of distance measures
Glielmo, A., Zeni, C., Cheng, B., Csanyi, G. and Laio, A., S. Ranking the information content of distance measures. arxiv (2021)
Researchers can assess the similarity of different data points by creating measures of distance between them. This paper assesses the amount of information contained in different distance measures.Read more
Predicting phase behaviors of superionic water at planetary conditions
Cheng, B., Bethkenhagen, M., Pickard, C.J. and Hamel, S. Predicting phase behaviors of superionic water at planetary conditions. arxiv (2021)
Understanding the thermodynamic and transport properties of water is crucial for planetary science. This paper uses machine learning to characterise the phase behaviours of water at extreme conditions.Read more
Assessing psychosis risk using quantitative markers of disorganised speech
Morgan, S.E., Diederen, K., Vértes, P.E. et al. Assessing psychosis risk using quantitative markers of disorganised speech. Translational Psychiatry (2021)
Disorganised speech can help predict later psychotic illness. This paper assesses the performance of twelve automated Natural Language Processing markers in differentiating transcribed speech excerpts from subjects at clinical high risk for psychosis, first episode psychosis patients and healthy control subjects.Read more
Understanding the behaviour of water
Monserrat, B., Brandenburg, J.G., Engel, E.A. et al. Liquid water contains the building blocks of diverse ice phases. Nat Commun 11, 5757 (2020).
This paper investigates the structural similarities between liquid water and a comprehensive set of 54 simulated ice phases. Its findings shed light on the phase behaviour of water.Read more
Predicting the phase diagram of water
Reinhardt, A., Cheng, B. Quantum-mechanical exploration of the phase diagram of water. Nat Commun 12, 588 (2021).
This paper analyses the phase diagram of water at three hybrid density-functional-theory levels of approximation. It demonstrates that it is possible to predict the phase diagram of a polymorphic system from first principles and provides a thermodynamic way of testing the limits of quantum-mechanical calculations.Read more
Optimising marker gene selection
Dumitrascu, B., Villar, S., Mixon, D.G. et al. Optimal marker gene selection for cell type discrimination in single cell analyses. Nat Commun 12, 1186 (2021).
This paper presents scGeneFit, an approach for marker selection in single cell RNA-seq. This method selects gene markers that jointly optimize cell label recovery using label-aware compressive classification methods.Read more