Confluence

This page is to collect and save scientific citations and useful links.

At this time the research papers of interest are (1) quality reviews; (2) use cases in AI and ML use in drug discovery; (3) high-profile use cases of AI and ML in medical diagnostics; (4) regulatory and ethical aspects of AI application in drug discovery and medicine.

AI-powered drug discovery captures pharma interest https ://www.nature.com/nbt/journal/v35/n7/full/nbt0717-604.html
Great intro to ML and Big Data Demystifying Big Data and Machine Learning for Healthcare https://www.crcpress.com/Demystifying-Big-Data-and-Machine-Learning-for-Healthcare/Natarajan-Frenzel-Smaltz/p/book/9781138032637
The Next Era: Deep Learning in Pharmaceutical Research https ://www.ncbi.nlm.nih.gov/pubmed/27599991 - Ekins Nov 2016
https://www.pharma-iq.com/informatics/articles/is-big-pharma-really-on-cusp-of-ai-shake-out-0
Review of AI June 2017 http://www.biorxiv.org/content/early/2017/05/28/142760
Github discussions is exellent https://github.com/greenelab/deep-review/issues
Excellent Blog from Bharath Ramsundar http://rbharath.github.io/what-cant-deep-learning-do/
IBM Watson A Reality Check for IBM’s AI Ambitions https://www.technologyreview.com/s/607965/a-reality-check-for-ibms-ai-ambitions/
June 2017 Next Generation AI algorithms need to make most of AI chips optimisation
Greg Diamos, Head of Systems Research at Baidu Silicon Valley AI
https://www.forbes.com/sites/ciocentral/2017/06/21/we-need-next-generation-algorithms-to-harness-the-power-of-todays-ai-chips/#7f20f83a47a0
WSJ June 2017How AI Is Transforming Drug Creation https://www.wsj.com/articles/how-ai-is-transforming-drug-creation-1498442760
Forbes Aug 2017 https://www.forbes.com/sites/forbestechcouncil/2017/08/03/artificial-intelligence-in-drug-discovery-a-bubble-or-a-revolutionary-transformation/#26079df64494
Beyond the Hype: Deep Neural Networks 2 Outperform Established Methods Using 3 A ChEMBL Bioactivity Benchmark Set http://www.biorxiv.org/content/biorxiv/early/2017/07/28/168914.full.pdf
WSJ Article March 2017 - broader business space review https://www.wsj.com/articles/how-ai-is-transforming-the-workplace-1489371060
Top 10 Recommendations for the AI Field in 2017 Oct 2017
The challenge of AI technology adoption in healthcare https://www.linkedin.com/pulse/challenge-ai-technology-adoption-healthcare-jay-chyung-md-phd/
Oct 2017 Challenages of AI adoption in Healthcare https://www.linkedin.com/pulse/challenge-ai-technology-adoption-healthcare-jay-chyung-md-phd/?trackingId=Cpa346%2FuUurbWLcO%2FhUTxw%3D%3D
Jan 2017 First FDA Approved Cloud Based Deep learning https://www.forbes.com/sites/bernardmarr/2017/01/20/first-fda-approval-for-clinical-cloud-based-deep-learning-in-healthcare/#357cbfe1161c (*** added 2017)

Fundamental of Deep Learning https://www.analyticsvidhya.com/blog/2016/03/introduction-deep-learning-fundamentals-neural-networks/?utm_content=bufferf9225&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer
Drug Makers Guide to the Galaxy: Excentia and polypharmacology https://www.nature.com/news/the-drug-maker-s-guide-to-the-galaxy-1.22683
Data is Key ingredient for AI https://www.elsevier.com/connect/5-reasons-data-is-a-key-ingredient-for-ai-applications?sf175330640=1
HBR AI for the real world https://hbr.org/2018/01/artificial-intelligence-for-the-real-world
AI for Health and Healtcare report (Mitre Group) https://www.healthit.gov/sites/default/files/jsr-17-task-002_aiforhealthandhealthcare12122017.pdf
Deep Learning for Biology https://www.nature.com/articles/d41586-018-02174-z
An automated curation procedure for addressing chemical errors and inconsistencies in public datasets used in QSAR modelling https://www.ncbi.nlm.nih.gov/pubmed/27885862
The ELF Honest Data Broker: informatics enabling public–private collaboration in a precompetitive arena https://www.sciencedirect.com/science/article/pii/S1359644615004249?via%3Dihub
Feb 2018: AI Startups https://blog.benchsci.com/startups-using-artificial-intelligence-in-drug-discovery?lipi=urn%3Ali%3Apage%3Ad_flagship3_feed%3BmEHpBm1fQ%2BKkySB%2FSg8G%2Fw%3D%3D#step11
Opportunities And Obstacles For Deep Learning In Biology And Medicine https://www.biorxiv.org/content/early/2018/01/19/142760
J. R. Soc. Interface 15: 20170387. http://dx.doi.org/10.1098/rsif.2017.0387
Perspectives: Augmented intelligence https://cen.acs.org/articles/96/i14/Perspectives-Augmented-intelligence.html.html
Computer system predicts products of chemical reactions: Machine learning approach could aid the design of industrial processes for drug manufacturing http://people.csail.mit.edu/tommi/papers/Connor_etal_ACS_2017.pdf
ANN used for Image analysis - Novartis NIBR https://www.novartis.com/stories/discovery/machine-learning-poised-accelerate-drug-discovery and the original publication in Bioinformatics: https://academic.oup.com/bioinformatics/article/33/13/2010/2997285?searchresult=1
The failure of IBM Watson: https://www.statnews.com/2018/07/25/ibm-watson-recommended-unsafe-incorrect-treatments/
An optimistic editorial from Nature references success of Wuxi NextCode: https://www.nature.com/articles/d41586-018-05267-x
AI interpretation of radiology images is hard, harder than initially anticipated: https://www.technologyreview.com/s/610552/google-x-ray-project-shows-ai-wont-replace-doctors-any-time-soon/
PLASTER, Nvidia's methodology for assessment of AI performance (PDF): https://images.nvidia.com/content/pdf/plaster-deep-learning-framework.pdf
UK national platform for AI? http://news.top-consultant.com//New-White-Paper-Calls-for-Government-wide-AI-Platform-19336.html
How Technology can tame scientific literature https://www.nature.com/articles/d41586-018-06617-5
Oncology use case: http://gate250.com/rk/2018FLAG_UsingArtificialIntelligence_Whitepaper_digital-Final.pdf
Rules of Machine Learning http://martin.zinkevich.org/rules_of_ml/rules_of_ml.pdf - could be expanded
Google DeepMind and healthcare in an age of algorithms https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5741783/
"Winner's Curse" - How to do better science in AI: https://openreview.net/pdf?id=rJWF0Fywf
A list of biomedical start-ups that claim AI as a core technology: https://blog.benchsci.com/startups-using-artificial-intelligence-in-drug-discovery
AI used to predict CRISPR DNA cuts; in the future, this may be used for gene therapy: http://www.bio-itworld.com/2018/11/08/machine-learning-predicts-how-dna-breaks-under-crispr.aspx
Augmenting Medicinal Chemist with data https://doi.org/10.1016/j.drudis.2018.03.011
Review of Forrester Wave views on AI and Automation https://www.forbes.com/sites/gilpress/2018/11/06/ai-and-automation-2019-predictions-from-forrester/#788550a54cb5 Nov 2018
Ethics in AI: Rabbi Jonathan Sacks explores how we should respond to the ways in which AI is transforming our world https://www.bbc.co.uk/sounds/play/b0bgrw3k
What it means to open AI’s black box? http://usblogs.pwc.com/emerging-technology/to-open-ai-black-box/
Drug Discovery Maps, a Machine Learning Model That Visualizes and Predicts Kinome–Inhibitor Interaction Landscapes https://pubs.acs.org/doi/full/10.1021/acs.jcim.8b00640 (*** added 2018)

Tuning artificial intelligence on the de novo design of natural-product-inspired retinoid X receptor modulators https://www.nature.com/articles/s42004-018-0068-1
Artificial intelligence and its potential in oncology https://doi.org/10.1016/j.artmed.2018.08.008
Deep convolutional neural networks for brain image analysis on magnetic resonance imaging: a review, by Bernal, J., Artificial Intelligence In Medicine,
https://doi.org/10.1016/j.artmed.2018.08.008
Transforming Computational Drug Discovery with Machine Learning and AI ACS Med. Chem. Lett. 2018, 9, 11, 1065-1069
The convergence of artificial intelligence and chemistry for improved drug discovery (AZ) https://www.future-science.com/doi/full/10.4155/fmc-2018-0161
AI in Pharma, by LEK Consulting: https://www.lek.com/insights/artificial-intelligence-life-sciences-formula-pharma-success-across-drug-lifecycle#.XAvu0zCNM7M.linkedin
Are Ontologies relevant in a Machine Learning-centric world? By Lee Harland, based on our own Boston AI workshop: https://www.scibite.com/are-ontologies-relevant-in-a-machine-learning-centric-world/
Computers turn neural signals into speech:http://science.sciencemag.org/content/363/6422/14?utm_campaign=wnews_sci_2019-01-03&et_rid=17144995&et_cid=2581434
How is automated text summarization done? https://arxiv.org/pdf/1707.02268.pdf
FREE ML courses (the title says 10, but the text contains links to over 30): https://www.kdnuggets.com/2018/12/10-more-free-must-see-courses-machine-learning-data-science.html
Eric Topol: High-performance medicine: the convergence of human and artificial intelligence
AI and Neural Net Summary in 21 pages:https://www.linkedin.com/feed/update/activity:6492067351771574272/ (*** added 01/2019)

Dark Secret at heart of AI - Black box https://www.technologyreview.com/s/604087/the-dark-secret-at-the-heart-of-ai/
Classification of ligand-binding pockets in proteins with a convolutional neural network: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1006718 (but only works well for nucleotide and heme binding sites, so only in a couple of highly specialized cases)
Overfitting dangers with small datasets: https://www.chemistryworld.com/news/dispute-over-reaction-prediction-puts-machine-learnings-pitfalls-in-spotlight/3009912.article
Pathologist-level classification of histologic patterns on resected lung adenocarcinoma slides with deep neural networks (Nature): https://www.nature.com/articles/s41598-019-40041-7.pdf
AI recognizes speech‐based markers for posttraumatic stress disorder in US veterans, with 89% classification rate: https://onlinelibrary.wiley.com/doi/abs/10.1002/da.22890
A machine learning approach for somatic mutation discovery for diagnostics: https://stm.sciencemag.org/content/10/457/eaar7939 and the full-length research article: https://stm.sciencemag.org/content/11/489/eaat6177?utm_source=STAT+Newsletters&utm_campaign=50dc0470e5-MR_COPY_01&utm_medium=email&utm_term=0_8cab1d7961-50dc0470e5-149702353
Learn the basics of statistics: https://www.statisticshowto.datasciencecentral.com/probability-and-statistics/ Well, not only the basics! Some advanced topics too.
On the interpretability of Neural Net models: https://appsilon.com/please-explain-black-box/
Predict risk of lung cancer by a Google AI model? Claimed 94% AUC, but refused to release the code: https://www.nature.com/articles/s41591-019-0447-x
What is the difference between AI, ML, and Deep Learning models? See here: https://www.edureka.co/blog/ai-vs-machine-learning-vs-deep-learning/ and here: https://www.geospatialworld.net/blogs/difference-between-ai%EF%BB%BF-machine-learning-and-deep-learning/ (*** added 05/2019)

Review: AI in Drug Design: https://www.mdpi.com/1420-3049/23/10/2520/htm
Is China way ahead of the West in AI use in medicine? See here (may require registration for full-length paper): https://www.statnews.com/2019/06/05/china-leapfrogging-us-using-ai-in-medicine/
ML Flow, a system for managing the AI model life cycle, and deployment to the cloud: https://mlflow.org/
An interesting lecture in the history of science. Bernard Widrow, a Stanford professor, talks about the Least Mean-Squares (LMS) adaptive algorithm used to train first neuron models. The device shown in the lecture dates back to 1959 and is actually an electro-mechanical switch box. It has parts that broke due to age, but still functions, and can be trained to tell apart letters of the alphabet. Part 1 and part 2.
Clinical validity and technical validity of AI methods in medicine: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6526255/#r10
In a review of 516 research papers in the diagnostic analysis of medical images with AI, only 31 (or 6%) performed external validation, and none used best practices in clinical trial design: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6389801/ The papers reviewed were published between Jan 1st and Aug 18th 2018.
FDA: Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback: https://www.regulations.gov/document?D=FDA-2019-N-1185-0001 (download the actual document in PDF format from this page)
Example of QSAR in 3D using Convolutional Neural Network: https://www.jstage.jst.go.jp/article/cpb/67/5/67_c18-00757/_html/-char/en
Nature Review: Technologies to watch in 2019: https://www.nature.com/articles/d41586-019-00218-6
Machine Learning for Medical Ultrasound: Status, Methods, and Future Opportunities. This review includes organ segmentation, an apparently hard topic. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5886811/
Predict Adverse Drug Reactions (ADRs) through drug-gene interactions: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5473874/ This review might be of interest in view of our recent drug repurposing datathon, where prediction of ADRs played a large role. (*** added 06/2019)

Privacy of patients and data donors in the age of genomics and big data mining: https://genomebiology.biomedcentral.com/articles/10.1186/s13059-019-1741-0
Automated de novo molecular design by hybrid machine intelligence and rule-driven chemical synthesis. By Alexander Button, Daniel Merk, Jan A. Hiss & Gisbert Schneider, Nature Machine Intelligence, volume 1, pages307–315 (2019)
Review: Concepts of Artificial Intelligence for Computer-Assisted Drug Discovery: https://pubs.acs.org/doi/full/10.1021/acs.chemrev.8b00728
Machine and deep learning meet genome-scale metabolic modeling: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1007084
Disease gene prediction for molecularly uncharacterized diseases: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1007078
Not quite ML, but curious: Identifying determinants of persistent MRSA bacteremia using mathematical modeling: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1007087
A popular piece on using AI to mine Pubmed data in an attempt to diagnose a rare disease: "How an AI expert took on his toughest project ever: writing code to save his son’s life", but what is really interesting, is that this project has a public github page, from which one can get one's own copy of the software
Artificial intelligence for assisting diagnostics and assessment of Parkinson's disease-A review. Main point is that after 48 completed studies, there is still not enough data.
Putting benchmarks in their rightful place: The heart of computational biology. Editorial
Ten simple rules for writing and sharing computational analyses in Jupyter Notebooks. Best practices
Three pitfalls to avoid in machine learning (Nature, Google) (*** added 07/2019)

Review: Deep learning in drug discovery: opportunities, challenges and future prospects
Learning Health System for Breast Cancer: Pilot Project Experience. "It is possible to extract, read, and combine data from the EHR to view the patient journey. The agreement between NLP and the gold standard was high, which supports validity."
WE-E-213CD-06: A Locally Adaptive, Intensity-Based Label Fusion Method for Multi- Atlas Auto-Segmentation. Comparison of proprietary anatomy segmentation methods for medical images.
Best practices of AI use in medicine: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6599029/
A collection of papers on use of AI and ML in diagnostic imaging: The use and performance of artificial intelligence applications in dental and maxillofacial radiology: A systematic review; State-of-the-Art Deep Learning in Cardiovascular Image Analysis; Artificial intelligence in cardiovascular imaging: state of the art and implications for the imaging cardiologist; Radiomics and Artificial Intelligence for Biomarker and Prediction Model Development in Oncology
The combination of computational chemistry and computational materials science with machine learning and artificial intelligence This is relevant to AI in CADD
"Learning rates of state-of-the-art artificial learning algorithms can be improved by adopting fundamental principles that govern the dynamics of the brain" Biological learning curves outperform existing ones in artificial intelligence algorithms
Artificial intelligence in digital pathology
An Analytical Review of Computational Drug Repurposing
Automated Diagnosis of Lymphoma with Digital Pathology Images Using Deep Learning
Yearbook of Medical Informatics, published by Thieme Journals in August 2019, contains many reviews on the AI applications in medical research and practice, for instance, "Recent Advances in Using Natural Language Processing to Address Public Health Research Questions Using Social Media and ConsumerGenerated Data", "Artificial Intelligence in Health: New Opportunities, Challenges, and Practical Implications", and many more. (*** added 08/2019)

Exploring microRNA Regulation of Cancer with Context-Aware Deep Cancer Classifier
PLATYPUS: A Multiple–View Learning Predictive Framework for Cancer Drug Sensitivity Prediction
Tumor classification based on image analysis: A Novel Method for Classifying Liver and Brain Tumors Using Convolutional Neural Networks, Discrete Wavelet Transform and Long Short-Term Memory Networks
Drug repurposing in oncology: Compounds, pathways, phenotypes and computational approaches for colorectal cancer
Towards trustable machine learning "Clinical implementations of machine learning that are accurate, robust and interpretable will eventually gain the trust of healthcare providers and patients"
Gartner on ML in business (very general and not focused on pharma, but could be a good primer): https://www.gartner.com/binaries/content/assets/events/keywords/catalyst/catus8/preparing_and_architecting_for_machine_learning.pdf
NP-Scout: Machine Learning Approach for the Quantification and Visualization of the Natural Product-Likeness of Small Molecules
Review: Looking beyond the hype: Applied AI and machine learning in translational medicine
Special issue of the Journal of the American College of Radiology on AI and Data Science, with many application cases of AI to clinical image processing
Statistical considerations for testing an AI algorithm used for prescreening lung CT images. Although a highly specialized use case is employed here, the statistical concepts are broadly valid.
Deep learning enables rapid identification of potent DDR1 kinase inhibitors. Report from A. Zhavoronkov's group to Nature Biotechnology.
Machine Learning to Predict, Detect, and Intervene Older Adults Vulnerable for Adverse Drug Events in the Emergency Department
Extracting biological age from biomedical data via deep learning: too much of a good thing? An attempt to create a CNN predictor of the natural "biological clock"
Computational prediction of diagnosis and feature selection on mesothelioma patient health records. Random forest outperformed all models tried, but MCC is low for all model categories. Data is available via the Irvine depository.
Modelling the prevalence of diabetes mellitus risk factors based on artificial neural network and multiple regression. Logistic regression outperformes a neural net. (*** added 09/2019)

AI used to make sonogram movies of living heart. STAT news report.
User-centric design of AI products: People+AI Guidebook from Google
AAIH whitepaper describes the basic AI techniques and discusses some applications in healthcare: https://www.theaaih.org/pdf/1571334853.pdf
AI in health care delivery: https://catalyst.nejm.org/health-care-ai-systems-changing-delivery/
WanDB, a package to log hyperparameters and output metrics from your runs, explore model architectures, and compare results. (FYI only. We at Pistoia Alliance have NOT tested this).
Bradshaw molecular design tool GSK Oct 2019 https://link.springer.com/article/10.1007/s10822-019-00234-8
Looking beyond the hype: Applied AI and machine learning in translational medicine
Characterizing Artificial Intelligence Applications in Cancer Research: A Latent Dirichlet Allocation Analysis. Trends in AI publications in medicine.
Patient similarity networks for precision medicine
How far have decision tree models come for data mining in drug discovery?
Ethics of Artificial Intelligence in Radiology: Summary of the Joint European and North American Multisociety Statement
An Online Calculator for the Prediction of Survival in Glioblastoma Patients Using Classical Statistics and Machine Learning
A collection of psychiatry-themed use cases of AI came out this month:
1. Complexity in mood disorder diagnosis: fMRI connectivity networks predicted medication‐class of response in complex patients
2. Exploring the Utility of Community-Generated Social Media Content for Detecting Depression: An Analytical Study on Instagram
3. Machine learning in major depression: From classification to treatment outcome prediction
4. Prediction Models of Functional Outcomes for Individuals in the Clinical High-Risk State for Psychosis or With Recent-Onset Depression Apparently, ML-based prognostic engine out-performs human psychiatrists
The plan to mine the world’s research papers. Nature on a controversial plan to subvert the copyright
Discovery of Novel Conotoxin Candidates Using Machine Learning. De-novo design of toxic peptides. Interesting but unfortunately does not include experimental validation of the pharmaceutical properties of these sequences
Why deep-learning AIs are so easy to fool DNNs are brilliant at highly specific tasks and brittle when unexpected input arrives. Nature editorial on adversarial learning in DNNs.
Rethinking Drug Repositioning and Development with Artificial Intelligence, Machine Learning, and Omics. Review
Artificial Intelligence, Responsibility Attribution, and a Relational Justification of Explainability. Responsibility to explain AI results is not only for the regulators, but mainly for patients.
“A patient like me” – An algorithm-based program to inform patients on the likely conditions people with symptoms like theirs have.
Key challenges for delivering clinical impact with artificial intelligence. "Robust clinical evaluation, using metrics that are intuitive to clinicians and ideally go beyond measures of technical accuracy to include quality of care and patient outcomes, is essential. Further work is required (1) to identify themes of algorithmic bias and unfairness while developing mitigations to address these, (2) to reduce brittleness and improve generalisability, and (3) to develop methods for improved interpretability of machine learning predictions." (*** added 10/2019)

A good primer, An Ophthalmologist’s Guide to Deciphering Studies in Artificial Intelligence
Reliable Prediction of Human Cytochrome P450 Inhibition Using Artificial Intelligence Approaches
Translational AI and Deep Learning in Diagnostic Pathology
Artificial‐Intelligence‐Driven Organic Synthesis—En Route towards Autonomous Synthesis?
Has Drug Design Augmented by Artificial Intelligence Become a Reality? By Ola Engkvist
Transfer learning for biomedical named entity recognition with neural networks
Machine Learning in Drug Discovery
Fréchet ChemNet Distance: A Metric for Generative Models for Molecules in Drug Discovery
Machine learning predicts individual cancer patient responses to therapeutic drugs with high accuracy
AI algorithm detects melanoma in skin images with accuracy matching that of a specialist MD
Artificial intelligence applications for pediatric oncology imaging. A review of a multitude of AI application in medical imaging in many modalities. Contrary to the title, case studies are not limited to juvenile patients, so the review is broadly applicable.
Design of metalloproteins and novel protein folds using variational autoencoders. Automated protein engineering. Follows the best practice of making code fully and freely available (see ref).
Superior skin cancer classification by the combination of human and artificial intelligence. What is interesting about this paper (yet another description of yet another clinical classifier), are: (1) the size of the collaboration effort; and (2) the combined approach of using AI and human input.
Protein engineering with Machine Learning driven CAD: "One-shot optimization of multiple enzyme parameters: Tailoring glucose oxidase for pH and electron mediators"
Pathways to breast cancer screening artificial intelligence algorithm validation. This paper focuses on external validation of diagnostic AI algorithms. How much validation is enough? Is the US FDA standard too permissive?
The Last Mile: Where Artificial Intelligence Meets Reality discusses the challenges to the practical use of AI in the clinical setting
A governance model for the application of AI in health care
Clinical-grade computational pathology using weakly supervised deep learning on whole slide images (Nature Medicine)
This may be important for all innovators, but the full text is behind a pay wall: Early Identification of Patentable Medical Innovations Use AI to browse descriptions of medical discoveries and flag those that are likely to be patentable
BIPSPI: a method for the prediction of partner-specific protein–protein interfaces Alas, leave-one-out validation method is prone to overfitting
Automatic Organ Segmentation for CT Scans Based on Super-Pixel and Convolutional Neural Networks Mentions some benchmark datasets for the organ segmentation problem ("Silver 7", (ref))
An Efficient Implementation of Deep Convolutional Neural Networks for MRI Segmentation
Computational Protein Design with Deep Learning Neural Networks While interesting, the resulting accuracy of 38% indicates that there is much room for improvement
Current trends in AI drug discovery start-ups, by our member Simon Smith
167 Startups Using Artificial Intelligence in Drug Discovery, by Simon Smith. This list was started in 2017 and is continuously updated.
62 Drugs in the Artificial Intelligence in Drug Discovery Pipeline, by Simon Smith
By this moment (11.2019) the US FDA has approved at least 26 medical devices that use AI. Link Link to the original Eric Topol post from the middle of 2019 and to his Nature paper
Statement from FDA Commissioner Scott Gottlieb, M.D. on steps toward a new, tailored review framework for artificial intelligence-based medical devices: "We are exploring a framework that would allow for modifications to algorithms to be made from real-world learning and adaptation, while still ensuring safety and effectiveness of the software as a medical device is maintained. A new approach to these technologies would address the need for the algorithms to learn and adapt when used in the real world"
DeepWAS: Multivariate genotype-phenotype associations by directly integrating regulatory information using deep learning
Machine Learning Models Based on Molecular Fingerprints and an Extreme Gradient Boosting Method Lead to the Discovery of JAK2 Inhibitors
Artificial intelligence in clinical and genomic diagnostics. Review (*** added 11/2019)

Real-time multi-model ensemble forecasts for seasonal influenza in the U.S.
A mathematical-descriptor of tumor-mesoscopic-structure from computed-tomography images annotates prognostic- and molecular-phenotypes of epithelial ovarian cancer Nature Communications
Are we doing a good job in validation of medical devices that are powered by AI? Lifecycle Regulation of Artificial Intelligence– and Machine Learning–Based Software Devices in Medicine, JAMA
How to Read Articles That Use Machine Learning. JAMA 2019 Users’ Guides to the Medical Literature
Adversarial Controls for Scientific Machine Learning. ACS Chemical Biology 2018 "Machine learning algorithms readily exploit confounding variables and experimental artifacts instead of relevant patterns, leading to overoptimistic performance and poor model generalization."
CDSeq: A novel complete deconvolution method for dissecting heterogeneous samples using gene expression data. A machine-learning technique
Algorithms on regulatory lockdown in medicine. Science paper on the regulation of AI-driven medical technologies.
Rethinking drug design in the artificial intelligence era. Nature review with an extensive reference list. Gisbert et al..
Attitudes Of Chinese Cancer Patients Toward The Clinical Use Of Artificial Intelligence. Cancer patients trust human doctors more than AI when AI recommendations and human MD opinions differ. Going forward this topic of trust will be gaining importance
A Bayesian graph convolutional network for reliable prediction of molecular properties with uncertainty quantification.
Visualizing structure and transitions in high-dimensional biological data. An algorithm to present highly complex, multi-dimensional data in a form with reduced dimensionality, suitable for human review.
PathFlowAI: A High-Throughput Workflow for Preprocessing, Deep Learning and Interpretation in Digital Pathology. Using AI to analyze liver microscopy images
Pacific Symposium in Biocomputing announcement of the workshop in ethics of AI.
DeepConv-DTI: Prediction of drug-target interactions via deep learning with convolution on protein sequences
Ten quick tips for effective dimensionality reduction
Random forest prediction of Alzheimer’s disease using pairwise selection from time series data. Not AI but machine learning
Converging a Knowledge-Based Scoring Function: DrugScore2018. DrugScore2018 is a competitive scoring and objective function for structure-based ligand design purposes.
Practical Model Selection for Prospective Virtual Screening. Random Forest wins over complex NN methods.
Unifying machine learning and quantum chemistry with a deep neural network for molecular wavefunctions Nature Communications 2019 Gasteiger et al
Rule of thumb: Which AI / ML algorithms to apply to business problems. Has a small section on problems in health sciences. Ideally we'd be able to produce a similar guide but for problems specific to life science research.
Conditional Molecular Design with Deep Generative Models
The global landscape of AI ethics guidelines. Nature 09/2019
Artificial Intelligence in medical imaging practice: looking to the future. A short review that covers the major application advances in the field as well as the ethical part of the AI use in medical decision-making. 11/2019
Drug combination sensitivity scoring facilitates the discovery of synergistic and efficacious drug combinations in cancer. Use ML to cluster cancer cell lines based on their drug combination sensitivity profiles. PLOS Comp Bio 05/2019
Implementation of machine learning algorithms to create diabetic patient re-admission profiles using public data sources at UC Irvine Machine Learning Repository
Machine Learning Technical Landscape picture
Bayesian Nonparametric Models tutorial
Comparison Study of Computational Prediction Tools for Drug-Target Binding Affinities
From What to How: An Initial Review of Publicly Available AI Ethics Tools, Methods and Research to Translate Principles into Practices
How the FDA Regulates AI
Ethical considerations about artificial intelligence for prognostication in intensive care "Respect for patients’ autonomy during decision-making requires transparency of the data processing by AI models to explain the predictions derived from these models." (*** added 12/2019)
International evaluation of an AI system for breast cancer screening. This is a Google-developed tool which is claimed to achieve a higher accuracy in diagnosis of breast cancer than that by human oncologists. However, code and data used to build the classifier are not released.
Applications of Machine Learning Approaches in Emergency Medicine; a Review Article. In this paper the authors compare performance of a different ML and AI methods for prediction (diagnosis, to be precise) of various diseases from EHR records. Performance is variable and the methods used for different disease types are vastly different.
Identification of pharmacodynamic biomarker hypotheses through literature analysis with IBM Watson
Opportunities for Artificial Intelligence in Advancing Precision Medicine. Review of ML methods used to analyze "-omics" data.
Towards A Rigorous Science of Interpretable Machine Learning. Arxiv
Applications of Deep-Learning in Exploiting Large-Scale and Heterogeneous Compound Data in Industrial Pharmaceutical Research
The MELLODDY Consortium. Pharma Companies Join Forces to Train AI for Drug Discovery Using Blockchain
Addressing Bias in Artificial Intelligence in Health Care. Abstract only.
AI in biomedical sciences attracted over $5B in VC investment. Market report
Deep Learning-driven research for drug discovery: Tackling Malaria. PLOS
Contract offers unprecedented look at Google deal to obtain patient data from the University of California. UCSF allowed Google to use patient data for AI research, which raises all kinds of ethical questions.
A comparative study of deep learning architectures on melanoma detection
Gartner Hype Cycle for Artificial Intelligence, 2019
A Deep Learning Approach to Antibiotic Discovery (*** added 02/2020)

Machine learning in chemoinformatics and drug discovery. Review, may be valuable due to multiple references cited.
Machine learning with random subspace ensembles identifies antimicrobial resistance determinants from pan-genomes of three pathogens. PLOS Comp Bio
Systematic Review of Privacy-Preserving Distributed Machine Learning From Federated Databases in Health Care. Review
COVID-19 Open Research Dataset (CORD-19) of scholarly literature has been released. Suitable for ML.
In-Silico Molecular Binding Prediction for Human Drug Targets Using Deep Neural Multi-Task Learning. A comparison of many single-task and multi-task ML methods. Contrary to earlier observations, single-task models perform better. Multi-task models work well for groups of highly similar targets (as expected, more data results in better predictive performance).
Predicting or Pretending: Artificial Intelligence for Protein-Ligand Interactions Lack of Sufficiently Large and Unbiased Datasets. More data, better, less biased benchmarks are needed.
Large-scale comparison of machine learning methods for drug target prediction on ChEMBL.
A deep learning framework for automatic diagnosis of unipolar depression. A classifier for diagnosis of major depression from EEG data.
ARPNet: Antidepressant Response Prediction Network for Major Depressive Disorder. Not an ideal study set-up, likely resulting in overfitting. But the overall proposal to automate selection of drugs for personalized therapy is good.
DDI-PULearn: a positive-unlabeled learning method for large-scale prediction of drug-drug interactions. Absence of a reported interaction does not equal a negative case due to sparsity of experimental data.
Human–machine partnership with artificial intelligence for chest radiograph diagnosis. Collective intelligence has a long history in medicine, and here is a high-tech twist on it.
Improving Oral Cancer Outcomes with Imaging and Artificial Intelligence. Abstract only.
Computer science versus COVID-19
The ethics in AI research. Nature. The battle for ethical AI at the world’s biggest machine-learning conference. "Bias and the prospect of societal harm increasingly plague artificial-intelligence research — but it’s not clear who should be on the lookout for these problems."
Using AI to create potentials for protein folding: AlphaFold (Nature) and DeepECA.
AI is not really artificial reasoning, but a sophisticated classifier. "AI isn't" in response to "Deconstructing the diagnostic reasoning of human versus artificial intelligence"
Best practices for AI in medicine: "Machine learning and artificial intelligence research for patient benefit: 20 critical questions on transparency, replicability, ethics, and effectiveness". BMJ. A lot of the best practices identified in this paper overlap with those identified by us, in an R&D application. A related paper: "Artificial intelligence versus clinicians: systematic review of design, reporting standards, and claims of deep learning studies"
Real world study for the concordance between IBM Watson for Oncology and clinical practice in advanced non‐small cell lung cancer patients at a lung cancer center in China. Compare this report to the best practices described in the previous link. This study focuses on concordance between AI and human MD recommendations (in reality, for IBM Watson, cohort recommendation of US-based MDs who provide input to IBM Watson, versus recommendations by individual overseas doctors), and not on clinically relevant metrics (improved outcomes).
Deconstructing the diagnostic reasoning of human versus artificial intelligence. How do human doctors and AI tools arrive at diagnosis, and how do they mis-diagnose?
Large-scale comparison of machine learning methods for drug target prediction on ChEMBL. Chemical Science
In light of multiple announcements (see Kaggle and Decentralized.ai sites) of AI challenges to help combat COVID-19, a skeptical outlook: is there enough data? "Debate flares over using AI to detect Covid-19 in lung scans" StatNews (*** added 03/2020)

Ethics in the AI world. Exclude the "black box" from making of important decisions. Science.
Irreproducible results in biomarker research. This is not Ai strictly speaking, but this is a review of some data sources that we mine.
An Ai model that predicts COVID-19 patient decline is pushed into the clinic without sufficient testing (STAT newsletter). This is an indicator of a troubling trend.
The need for a system view to regulate artificial intelligence/machine learning-based software as medical device. A somewhat controversial viewpoint paper.
Ensuring Trustworthy Use of Artificial Intelligence and Big Data Analytics in Health Insurance. "Unless an enabling ethical environment is in place, the use of such analytics will likely contribute to the proliferation of unconnected data systems, worsen existing inequalities, and erode trustworthiness and trust."
Asilomar ethical principles for AI. Published in 2017.
Design of Natural‐Product‐Inspired Multitarget Ligands by Machine Learning. An interesting fully-automated ligand discovery that combines conventional molecular docking, chemoinformatics and AI approaches. The results of the computational molecular design were validated in-vitro.
Primer on an ethics of AI-based decision support systems in the clinic.
P7003 - Algorithmic Bias Considerations. A proposal for a standard by IEEE. (*** added 04/2020)

What ML techniques are most frequently used to make diagnostic and therapeutic care models for diabetes? "Artificial Intelligence Applications in Type 2 Diabetes Mellitus Care: Focus on Machine Learning Methods"
Diagnostic models based on epigenetic data. Review. Machine learning and clinical epigenetics: a review of challenges for diagnosis and classification.
On the interpretability of machine learning-based model for predicting hypertension
Predicting effective drug combinations using gradient tree boosting based on features extracted from drug-protein heterogeneous network. Predictive model is limited by availability of negative examples (failed drug combinations) in the training set.
Predicting synthetic lethal interactions in human cancers using graph regularized self-representative matrix factorization. Same issue with the data: no negative examples. A reference to a very peculiar database of synthetic lethal gene interactions.
Gender imbalance in medical imaging datasets produces biased classifiers for computer-aided diagnosis. Even small deviations from the 50/50 in the training data result in degradation of model performance on the under-represented gender. (ChestXpert data set, used in one of the reported studies, is 60% male and 40% female). Effectively it is a data consistency problem, similar to the two papers above.
Improving reproducibility in computational biology research. This is not a full paper, but a declaration of intent to run a study.
The Affective Ising Model: A computational account of human affect dynamics. Is Ising Model really ML? Not quite sure. But this paper makes a curious attempt to compute human emotions. (*** added 05/2020)

Prioritizing and Analyzing the Role of Climate and Urban Parameters in the Confirmed Cases of COVID-19 Based on Artificial Intelligence Applications. Ultimately, this study resulted in a simple regression model, confirming what would seem already known: higher population density and higher relative humidity levels result in higher incidence of the COVID-19.
An Open-Source, Vender Agnostic Hardware and Software Pipeline for Integration of Artificial Intelligence in Radiology Workflow. This publication is notable because it makes the source code accessible; see https://github.com/bdrad/clinical_ml_integration
Artificial intelligence in chemistry and drug design. Editorial. Section that discusses relative performance of ML models over other QSAR approaches may be noteworthy.
Medicinal Chemists Versus Machines Challenge: What Will It Take to Adopt and Advance Artificial Intelligence for Drug Discovery? Quote: " To ensure continued evolution of AI technologies, we propose a series of challenges of increasing complexity by comparing and combining the machine and human intelligence in medicinal chemistry."
Application of explainable ensemble artificial intelligence model to categorization of hemodialysis-patient and treatment using nationwide-real-world data in Japan
Progressive learning: A deep learning framework for continual learning. An attempt to step forward from complex yet standardized classifiers into real machine learning with an explicit codification of transfer learning.
SEVERITAS: An Externally Validated Mortality Prediction for Critically Ill Patients in Low and Middle-Income Countries.
Predicting Parameters in Deep Learning. CNNs are often created too complex with too many parameters. In this study up to 95% of neural net's parameters could be predicted from the values of other parameters, suggesting that these extra parameters have no real predictive power. This phenomenon is already well-known to the practitioners of simple machine learning models, who routinely review input data sets for co-linearity between independent variables.
SD-UNet: Stripping down U-Net for Segmentation of Biomedical Images on Platforms with Low Computational Budgets. Unfortunately, you get what you pay for, and this is very much visible in figure 9 that illustrates performance of the simplified method on a microscopy dataset.
Regulation of predictive analytics in medicine: Algorithms must meet regulatory standards of clinical benefit. Science 02/2019
Predicting translational progress in biomedical research. PLOS Biol. An 84% accurate machine learning system that detects whether a paper is likely to be cited by a future clinical trial or guideline.
Mechanism of Baricitinib Supports Artificial Intelligence-Predicted Testing in COVID-19 Patients.
Advancing Drug Discovery via Artificial Intelligence. Review in Cell, with a vast collection of references to computational methods.
Artificial Intelligence-Powered Search Tools and Resources in the Fight Against COVID-19.
Measuring the Quality of Explanations: The System Causability Scale (SCS). Comparing Human and Machine Explanations
Accuracy of Artificial Intelligence-Assisted Detection of Upper GI Lesions: A Systematic Review and Meta-Analysis. AI is accurate in the detection of upper GI neoplastic lesions and HP infection status. However, most of these studies were based on retrospective review of selected images, which would require further validation in prospective trials.
Human Gut Microbiome Aging Clock Based on Taxonomic Profiling and Deep Learning.
Deep learning in mental health outcome research: a scoping review.
Sex and gender differences and biases in artificial intelligence for biomedicine and healthcare.
Transitional Mesothelioma and Artificial Intelligence: Do We Need One More Subtype? and Do We Need Computers to Identify Them? A contrarian viewpoint: instead of training AI to make differential diagnosis of more vs less aggressive tumor sub-types, train pathologists. (*** added 07/2020)

Artificial intelligence in chemistry and drug design. Review with multiple references to methods, but without critical assessment.
Artificial Intelligence in Drug Discovery: Into the Great Wide Open. Editorial in the special issue of the J Med Chem on AI. (The direct link to this special issue is not yet available).
The upside of being a digital pharma player. Review. Deep Dive Into Big Pharma AI Productivity: One Study Shaking The Pharmaceutical Industry is a commentary on the previous paper, with interviews with the authors, and some interesting statistics of business activity and publishing in the AI field in big pharmaceutical firms.
Reflections on sharing clinical trial data.
Research summary: Principles alone cannot guarantee ethical AI. (*** added 08/2020)

2020 Special Section on Ethics in Health Informatics
Artificial Intelligence and Suicide Prevention: A Systematic Review of Machine Learning Investigations
Digitizing clinical trials
Medical Information Extraction in the Age of Deep Learning
Recommendations for machine learning validation in biology. A preprint of a paper submitted by our members from Elixir EU to Nature Methods.
Ten simple rules to power drug discovery with data science. One more best practices paper. (*** added 09/2020)

REINVENT 2.0: An AI Tool for De Novo Drug Design
A collection of papers on ethics in biomedical use of AI came out this month:
1. The ethical adoption of artificial intelligence in radiology. Ownership, trading in patients' data, and anonymization are among the topics addressed.
2. Identifying Ethical Considerations for Machine Learning Healthcare Applications. With dozens of public comments, such as Respect and Trustworthiness in the Patient-Provider-Machine Relationship: Applying a Relational Lens to Machine Learning Healthcare Applications and Where Bioethics Meets Machine Ethics, and even An Ethical Framework to Nowhere
3. These papers talk to the ideas expressed in Do no harm: a roadmap for responsible machine learning for health care, published in Nature a year ago
Transfer learning enables prediction of CYP2D6 haplotype function
Inconsistency in the use of the term "validation" in studies reporting the performance of deep learning algorithms in providing diagnosis from medical imaging reminds one of the importance of standards in description and reporting of AI models.
A review on drug repurposing applicable to COVID-19
Global gene network exploration based on explainable artificial intelligence approach (*** added 11/2020)

A High Recall Classifier for Selecting Articles for MEDLINE Indexing
McKinsey The State of AI in 2020
Gartner Hype Cycle for AI in 2020
Artificial Intelligence of COVID-19 Imaging: A Hammer in Search of a Nail. A critical review of a flood of similar publications reporting binary classifiers of unknown quality and clinical utility.
Stochastic Channel-Based Federated Learning With Neural Network Pruning for Medical Data Privacy Preservation: Model Development and Experimental Validation
Artificial intelligence in the early stages of drug discovery. Review
Multi-objective optimization methods in novel drug design
Incorporating biological structure into machine learning models in biomedicine. Review
Spectrum of deep learning algorithms in drug discovery. Review
Identifying transcriptomic correlates of histology using deep learning
Repurposing therapeutics for COVID-19: Rapid prediction of commercially available drugs through machine learning and docking
Medical Information Extraction in the Age of Deep Learning
Artificial Intelligence (AI)-Based Systems Biology Approaches in Multi-Omics Data Analysis of Cancer
Deep metabolome: Applications of deep learning in metabolomics
MSpectraAI: a powerful platform for deciphering proteome profiling of multi-tumor mass spectrometry data by using deep neural networks
Predictive article recommendation using natural language processing and machine learning to support evidence updates in domain-specific knowledge graphs. IBM PARSe program.
Why We Need to Bust Some Myths about AI. And the site referenced: https://www.aimyths.org/
Superethics Instead of Superintelligence: Know Thyself, and Apply Science Accordingly. "We don't need superintelligence, we need superethics"
The clinical artificial intelligence department: a prerequisite for success. A critical outlook on the clinical utility of AI.
Radiomics for precision medicine: Current challenges, future prospects, and the proposal of a new framework. Review.
A novel virtual screening procedure identifies Pralatrexate as inhibitor of SARS-CoV-2 RdRp and it reduces viral replication in vitro.
Application of Computational Biology and Artificial Intelligence Technologies in Cancer Precision Drug Discovery. Review. (*** added 12/2020)

Prognostic gene expression signatures of breast cancer are lacking a sensible biological meaning Correlation is not causation, and molecular signatures are not unique
Prediction of the confirmed cases and deaths of global COVID-19 using artificial intelligence
Approval of artificial intelligence and machine learning-based medical devices in the USA and Europe (2015–20): a comparative analysis
From 'molecules of life' to new therapeutic approaches, an evolution marked by the advent of artificial intelligence: the cases of chronic pain and neuropathic disorders. Review at DDT
Cognitive analysis of metabolomics data for systems biology. Nature Protocols
Use of artificial intelligence to enhance phenotypic drug discovery. Review. DDT
Implementation of model explainability for a basic brain tumor detection using convolutional neural networks on MRI slices. "In conclusion, the early implementation of model explainability features can greatly benefit the development of neural networks in radiology as they can help to detect biases introduced with the training data and to assess the potential of an approach. In later stages of the development of a neural network, model explainability features become mandatory to provide physicians with the information they need to integrate the model’s predictions into a meaningful clinical decision." However the paper lacks detail on how exactly the features used in explaining the results were selected.
Regulatory considerations for artificial intelligence technologies in GI endoscopy. Clear description of established and emergent regulatory approval procedures for the AI software as a medical device.
STAT’s database of FDA-cleared AI tools. Requires subscription to read the full text. Excerpt: "Of 161 AI products cleared by the FDA in recent years, only 73 disclosed the amount of patient data used to validate the performance of their devices in public documents. Only seven reported the racial makeup of their study populations, and just 13 provided a gender breakdown" (*** added 01/2021)

SAveRUNNER: A network-based algorithm for drug repurposing and its application to COVID-19
DeepMIB: User-friendly and open-source software for training of deep learning network for biological image segmentation
Artificial intelligence in drug discovery: what is realistic, what are illusions? Part 1 and Part 2 (***added 04/2021)

Deep into Laboratory: An Artificial Intelligence Approach to Recommend Laboratory Tests
Application and Performance of Artificial Intelligence Technology in Oral Cancer Diagnosis and Prediction of Prognosis: A Systematic Review
De novo molecular design and generative models. Review
Galaxy-ML: An accessible, reproducible, and scalable machine learning toolkit for biomedicine
WHO: Ethics and governance of artificial intelligence for health
Exploring the phenomenon and ethical issues of AI paternalism in health apps
Ethical Issues in Patient Data Ownership
AI in the Healthcare Enterprise — Prashant Natarajan
AI High Growth Markets
FDA: Artificial Intelligence and Machine Learning (AI/ML)-Enabled Medical Devices
Machine-learning methods for ligand–protein molecular docking. DDT Review
Evaluation of artificial intelligence on a reference standard based on subjective interpretation. The Lancet (*** added 06-09/2021)
AI in medicine must be explainable Nature Medicine
A benchmark for neural network robustness in skin cancer classification
Accelerating antibiotic discovery through artificial intelligence
Papers on prediction of protein 3D fold using AI:
1. Applying and improving AlphaFold at CASP14
2. Accurate prediction of protein structures and interactions using a three-track neural network
3. Highly accurate protein structure prediction for the human proteome
4. AI revolutions in biology - an editorial on the protein fold prediction papers above
Papers on the critical assessment of AI tools in the clinic (and in particular, for the diagnosis of COVID-19 and disease prognosis):
1. COVID-19 special issue: Intelligent solutions for computer communication-assisted infectious disease diagnosis. Special issue of Expert Systems (May 2021)
2. Prediction models for diagnosis and prognosis of covid-19: systematic review and critical appraisal
3. Hundreds of AI tools have been built to catch covid. None of them helped. MIT Technology Review
4. Google’s medical AI was super accurate in a lab. Real life was a different story. MIT Technology Review
5. Evaluation and Real-World Performance Monitoring of Artificial Intelligence Models in Clinical Practice: Try It, Buy It, Check It.
6. Introducing Artificial Intelligence Applications at Our Community Hospital: A Contrarian Approach
7. Common pitfalls and recommendations for using machine learning to detect and prognosticate for COVID-19 using chest radiographs and CT scans. Nature Machine Intelligence
8. Data Science and AI in the Age of COVID-19. A report by the Turing Institute (PDF)
Diagnosis of Acute Poisoning using explainable artificial intelligence. AI performs marginally worse than human doctors in diagnosing simple cases of poisoning, but fails on moderately complex ones.
Protocol for development of a reporting guideline (TRIPOD-AI) and risk of bias tool (PROBAST-AI) for diagnostic and prognostic prediction model studies based on artificial intelligence
Guidance overview: Good Machine Learning Practice for Medical Device Development: Guiding Principles
UK, USA and Canadian regulators identify 10 guiding principles to be addressed when medical devices use AI or machine learning software
DOME (Data Optimization Model Evaluation) group guidelines for the AI model evaluation and reporting
Epic’s AI algorithms, shielded from scrutiny by a corporate firewall, are delivering inaccurate information on seriously ill patients
Crowdsourcing biocuration: The Community Assessment of Community Annotation with Ontologies (CACAO) (*** added 11-12/2021)

A highly accurate metadynamics-based Dissociation Free Energy method to calculate protein–protein and protein–ligand binding potencies
AI Ethics: Algorithms are not neutral
Development and validation of a prognostic and predictive 32-gene signature for gastric cancer. In and by itself, a patient stratification model based on gene expression profiling is not new; hundreds of these are published every month and such reports are generally not tracked here. But this publication makes a particular effort to validate the model, and it apparently does have prognostic power.
Data Mining Meets Machine Learning: A Novel ANN-based Multi-Body Interaction Docking Scoring Function (MBI-Score) based on Utilizing Frequent Geometric and Chemical Patterns of Interfacial Atoms in Native Protein-Ligand Complexes
Deep learning—a first meta-survey of selected reviews across scientific disciplines, their commonalities, challenges and research impact. Review of reviews.
From computer-aided drug discovery to computer-driven drug discovery. Review. (*** added 02/2022)

Using deep learning to annotate the protein universe. Full text
Ten quick tips for deep learning in biology
Artificial intelligence–enabled virtual screening of ultra-large chemical libraries with deep docking
In Silico Trial Approach for Biomedical Products: A Regulatory Perspective
In-silico trials network in the UK: InSilicoUK network launch (***added 04/2022)

The ATOM Modeling PipeLine (AMPL) is an open-source, modular, extensible software pipeline for building and sharing models to advance in silico drug discovery. Github
Artificial intelligence in virtual screening: models versus experiments
A theory of consciousness from a theoretical computer science perspective: Insights from the Conscious Turing Machine
AI recognition of patient race in medical imaging: a modelling study
Best Practices on Big Data Analytics to Address Sex-Specific Biases in Our Understanding of the Etiology, Diagnosis, and Prognosis of Diseases
Biological activities of drug inactive ingredients
“Many roads lead to Rome and the Artificial Intelligence only shows me one road”: an interview study on physician attitudes regarding the implementation of computerized clinical decision support systems
View the recording of the webinar “Reimagining drug discovery: the significance of artificial intelligence” produced by Charles River Laboratories and Pistoia Alliance member company Valo Health in the Nature Research Custom Media series
View the recording of the webinar “The ATOM molecular design approach for accelerated drug discovery” (*** added 05/2022)

Quality Models for Artificial Intelligence Systems
Quantitative systems pharmacology: Landscape analysis of regulatory submissions to the US Food and Drug Administration. Review of recent submissions. Use of mathematical models in drug approval submissions to the FDA is steadily increasing, but the types of models and the quality of descriptions of models are highly variable.
Using deep learning to annotate the protein universe. Predict protein function by ML.
Genome-wide identification of the genetic basis of amyotrophic lateral sclerosis (using ML and GWAS)
Gender Bias in Artificial Intelligence: Severity Prediction at an Early Stage of COVID-19
Independent evaluation of 12 artificial intelligence solutions for the detection of tuberculosis
Artificial intelligence challenges for predicting the impact of mutations on protein stability. "The accuracy of predictors of changes in protein thermodynamic stability upon mutations has remained surprisingly constant over the past decades." Lack of data is the most significant reason for the slow progress. (*** added 06/2022)

ChemCrow: Augmenting large-language models with chemistry tools. Pre-print
Rethink reporting of evaluation results in AI. Science
FAIR: Making Data AI-Ready Book Chapter in Artificial Intelligence for Science
An implementation framework to improve the transparency and reproducibility of computational models of infectious diseases
Evolutionary-scale prediction of atomic-level protein structure with a language model. Science
In Minnesota, researchers explore how to audit health AI models on a patient level
How Well Do Large Language Models Support Clinician Information Needs?
How Foundation Models Can Advance AI in Healthcare
The Shaky Foundations of Foundation Models in Healthcare
Benefits, Limits, and Risks of GPT-4 as an AI Chatbot for Medicine
The Shaky Foundations of Clinical Foundation Models: A Survey of Large Language Models and Foundation Models for EMRs. Pre-print (*** added 03/2023)

Large Language Models: an update for the perplexed by the Pistoia Alliance member company Zifo R&D
Drug repurposing with large language models
GeneGPT: Augmenting Large Language Models with Domain Tools for Improved Access to Biomedical Information
An accessible infrastructure for artificial intelligence using a Docker-based JupyterLab in Galaxy
The Pipeline for the Continuous Development of Artificial Intelligence Models
Machine learning for small molecule drug discovery in academia and industry. Review with references to federated learning use cases that were previously reported in our webinars
Supporting regulatory compliance with AI
A series of papers on LLMs in healthcare by STAT (may require subscription): ChatGPT in Medicine, “What does generative AI mean for health care?” and “Making sense of AI research in medicine, in one slide”
“Building the algorithm is the easiest part. Integration of AI in healthcare systems is hard”
And a caution from the same publisher about attempts to use AI to ration medical care (*** added 05/2023)
Knowledge graphs and LLMs: https://www.youtube.com/watch?v=WqYBx2gB6vA
Pistoia Alliance Lab of the Future survey reveals AI is the top tech investment in life sciences
A Framework for Estimating The Value Of Patient Data. A LinkedIn post by Chris Wayman, Director, Life Sciences, EY
“AI and Science: What 1600 Researchers Think” A survey by the Nature magazine
“First fully-automated AI/ML virtual screening cascade implemented at a drug discovery centre in Africa”. Nature
“Hospital bosses love AI. Doctors and nurses are worried” by Washington Post
“We need a Weizenbaum test for AI”. Science. A Weisenbaum test defines utility of technology from the viewpoint of the common citizen
“Drug Repurposing - Advances, Scopes and Opportunities in Drug Discovery”. An on-line book
“How to Build Trustworthy AI Systems” by DataIKU (*** added 04-10/2023)

AI gone astray: How subtle shifts in patient data send popular algorithms reeling, undermining patient safety
Implications of the Use of Artificial Intelligence Predictive Models in Health Care Settings
Pharma AI Use Cases collection. Very basic, and with vendor advertising, but it is nice to have the majority of use cases in one place
EPAM AI Solutions Library. Many products are open-source
Chan Zuckerberg Initiative to build largest biomedical research computing system
AI’s potential to accelerate drug discovery needs a reality check. Nature
Clinical landscape of AI. An interesting figure from a Twitter post
What is new in Artificial Intelligence from the 2023 Gartner Hype Cycle
Highlights in AI for Drug Discovery. Funding deals and business advancements
Generative artificial intelligence for de novo protein design
AI drug discovery companies are now mature enough to show examples of successes and failures. This clearly indicates that AI as a technology is not a magic bullet many thought it would become during the hype days. This LinkedIn post delivers some of the sobering statistics. Lack of reliable data is clearly identified as the main bottleneck
Predetermined Change Control Plans for Machine Learning-Enabled Medical Devices: Guiding Principles, by the US FDA; and a LinkedIn post on the same topic with some historic data and statistics on the approvals of AI medical devices
White House orders up a plan to vet AI in health care
EU AI Act
AI Standards Hub
On the LLMs: Google's Gemini AI model beats ChatGPT4 in various tests. The part relevant to biotechnology R&D is at 19:19
Advances in genetics and artificial intelligence have helped scientists to crack the second “code of life” – gene regulation
Library of digital endpoints
LoRAX: The Open Source Framework for Serving 100s of Fine-Tuned LLMs by Predibase
Deciphering RNA splicing logic with interpretable machine learning
Detection of AI-Created Images Using Pixel-Wise Feature Extraction and Convolutional Neural Networks
Lightweight Multi-Class Support Vector Machine-Based Medical Diagnosis System with Privacy Preservation
Virtually Possible: Enhancing Quality Control of 3D-Printed Medicines with Machine Vision Trained on Photorealistic Images
The Trends and Future Prospective of In Silico Models from the Viewpoint of ADME Evaluation in Drug Discovery
Trends in Medicinal Chemistry: KNIME Workflows, QSAR Models, LLMs and Chemical Search Strategies
“There is an AI for that” A list of AI models for many tasks (not biology research, or not only)
Deep convolutional and conditional neural networks for large-scale genomic data generation
Recognition and reconstruction of cell differentiation patterns with deep learning
Google AI models for healthcare (added 12/2023)

UK AI Standards Hub
LoRA Land
Ten simple rules for using large language models in science, version 1.0
Generative AI Tracker: A guide to the health systems and companies driving adoption
AMIE: A research AI system for diagnostic medical reasoning and conversations
WHO's 6 Principles For An AI Regulatory Framework For Medical Product Development
Roche AI Ethics Principles
Roche Data Ethics Principles
Langtest: Safe and Effective Language Models
Account of AI in drug discovery in 2023 and projections for 2024 on Tech Meets Bio
More predictions for AI in 2024
Pistoia Alliance launches freely available IDMP Ontology 1.0, covering all five standards set by ISO - Pistoia Alliance
Google's Gemini AI model beats ChatGPT4 in various tests. YouTube link. The relevant part is at 19:19
SCOPE: FDA facilitates AI and ML use for clinical trials and drug development
Exploring chemical reaction space for ML (added 02/2024)