Reviews & Analysis

Deep learning models for sequential data can be trained to make accurate predictions from large biological datasets. New tools from computer vision and natural language processing can help us make these models interpretable to biologists.

Microscopy-based drug screens with fluorescent markers can shed light on how drugs affect biological processes. Without adding markers and imaging channels, which is cumbersome and costly, a new generative deep-learning method extracts new fluorescence channels from images, potentially improving the drug-discovery pipeline.

Neural networks can be implemented by using purified DNA molecules that interact in a test tube. Convolutional neural networks to classify high-dimensional data have now been realized in vitro, in one of the most complex demonstrations of molecular programming so far.

Unsupervised machine learning algorithms reduce the dependence on curated, labeled datasets that are characteristic of supervised machine learning. The authors argue that the developmental science of infant cognition could inform the design of unsupervised machine learning approaches.

Both proteins and natural language are essentially based on a sequential code, but feature complex interactions at multiple scales, which can be useful when transferring machine learning models from one domain to another. In this Review, Ferruz and Höcker summarize recent advances in language models, such as transformers, and their application to protein design.

Despite the growing number of initiatives that employ AI to counter corruption, few studies empirically tackle the political and social consequences of embedding AI in anti-corruption efforts. The authors outline the societal and technical challenges that need to be overcome for AI to fight corruption.

Machine reading and knowledge extraction methods can be used to mine the scientific literature and reveal the direction and robustness of discoveries. Such efforts now point to the importance of independent tests of reported claims.

Behavioural big data and algorithmic behaviour modification technologies controlled by commercial platforms have become difficult for academic researchers to access. Greene et al. describe barriers to academic research on such data and algorithms, and make a case for enhancing platform access and transparency.

It is an outstanding challenge to develop intelligent machines that can learn continually from interactions with their environment, throughout their lifetime. Kudithipudi et al. review neuronal and non-neuronal processes in organisms that address this challenge and discuss pathways to developing biologically inspired approaches for lifelong learning machines.

GPUs, which are highly parallel computer processing units, were originally designed for graphics applications, but they have played an important role in accelerating the development of deep learning methods. In this Review, Pandey and colleagues summarize how GPUs have advanced machine learning in the field of drug discovery.

Graph convolutional networks have become a popular tool for learning with graphs and networks. We reflect on the reasons behind the success story.

Predicting the performance of a tactile sensor from its composition and morphology is nearly impossible with traditional computational approaches. Machine learning can not only predict device-level performance, but also recommend new material compositions for soft machine applications.

Drug resistance in tropical diseases such as malaria requires constant improvement and development of new drugs. To find potential candidates, generative machine learning methods that can search for novel bioactive molecules can be employed.

Machine learning applications in agriculture can bring many benefits in crop management and productivity. However, to avoid harmful effects of a new round of technological modernization, fuelled by AI, a thorough risk assessment is required, to review and mitigate risks such as unintended socio-ecological consequences and security concerns associated with applying machine learning models at scale.

Machine learning performs well at predictive modelling based on statistical correlations, but for high-stakes applications, more robust, explainable and fair approaches are required. Cui and Athey discuss the benefits of bringing causal inference into machine learning, presenting a stable learning approach.

Although artificial reinforcement learning agents do well when rules are rigid, such as games, they fare poorly in real-world scenarios where small changes in the environment or the required actions can impair performance. The authors provide an overview of the cognitive foundations of hierarchical problem-solving, and propose steps to integrate biologically inspired hierarchical mechanisms to enable problem-solving skills in artificial agents.

Digitally recreating the likeness of a person used to be a costly and complex process. Through the use of generative models, AI-generated characters can now be made with relative ease. Pataranutaporn et al. discuss in this Perspective how this technology can be used for positive applications in education and well-being.

Geometric representations are becoming more important in molecular deep learning as the spatial structure of molecules contains important information about their properties. Kenneth Atz and colleagues review current progress and challenges in this emerging field of geometric deep learning.

Newly sequenced organisms present a challenge for protein function prediction, as they lack experimental characterisation. A network-propagation approach that integrates functional network relationships with protein annotations, transferred from well-studied organisms, produces a more complete picture of the possible protein functions.

Substantial advances have been made in the past decade in developing high-performance machine learning models for medical applications, but translating them into practical clinical decision-making processes remains challenging. This Perspective provides insights into a range of challenges specific to high-dimensional, multimodal medical imaging.