As humanity struggles to contain the global Covid-19 infection, prophylactic actions are grandly slowed down by the shortage of testing kits. Governments have taken several measures to work around this shortage: the FDA has become more liberal on the approval of Covid-19 tests in the US. In the UK emergency measures allowed to increase the daily number of locally produced test kits to 100,000. China has recently launched a massive test manufacturing program. However, all those efforts are very insufficient and many poor countries are still under threat. A popular method for reducing the number of tests consists in pooling samples, i.e. mixing patient samples and testing the mixed samples once. If all the samples are negative, pooling succeeds at a unitary cost. However, if a single sample is positive, failure does not indicate which patient is infected. This paper describes how to optimally detect infected patients in pools, i.e. using a minimal number of tests to precisely identify them, given the a priori probabilities that each of the patients is healthy. Those probabilities can be estimated using questionnaires, supervised machine learning or clinical examinations. The resulting algorithms, which can be interpreted as informed divide-and-conquer strategies, are non-intuitive and quite surprising. They are patent-free. Co-authors are listed in alphabetical order.

I'm pleased to announce the results of our open-source Coronavirus Drug Discovery Competition! In just 2 weeks, hundreds of developers from around the world signed up to join the fight against the novel coronavirus, using publicly available datasets and algorithms to come up with relevant solutions. The top 3 submissions, winning $3500 in prizes, stood out from the rest in terms of their algorithmic and reporting quality. In this episode, I'm going to announce each of their backgrounds, as well as dive into the various machine learning techniques they used to predict a suitable treatment for Coronavirus. The top submission identified a compound called Remdesivir as the the most promising treatment for COVID-2019, due to its high scoring inhibitory potential when docked against the Coronavirus main Protease.

The federal government has received plenty of well-deserved flack for slow-rolling the national launch of diagnostic tests for Covid-19. First came the flawed swab-based tests from the Centers for Disease Control and Prevention, followed by a chaotic, lost month of regulatory tango that prevented independent tests from getting scaled and out the door. So when interest arose in a different kind of testing--antibody blood tests, which are used to find evidence of past infection, not a current diagnosis--the US Food and Drug Administration was under pressure to hurry things along. In mid-March, the agency loosened its rules, declaring via an update to its emergency use guidance that antibody tests could be sold without seeking the agency's approval, provided that manufacturers did their own validation. Now FDA officials are walking back that decision.

Effective property prediction methods can help accelerate the search for COVID-19 antivirals either through accurate in-silico screens or by effectively guiding on-going at-scale experimental efforts. However, existing prediction tools have limited ability to accommodate scarce or fragmented training data currently available. In this paper, we introduce a novel approach to learn predictors that can generalize or extrapolate beyond the heterogeneous data. Our method builds on and extends recently proposed invariant risk minimization, adaptively forcing the predictor to avoid nuisance variation. We achieve this by continually exercising and manipulating latent representations of molecules to highlight undesirable variation to the predictor. To test the method we use a combination of three data sources: SARS-CoV-2 antiviral screening data, molecular fragments that bind to SARS-CoV-2 main protease and large screening data for SARS-CoV-1. Our predictor outperforms state-of-the-art transfer learning methods by significant margin. We also report the top 20 predictions of our model on Broad drug repurposing hub.

COVID-19 has created a global pandemic with high morbidity and mortality in 2020. Novel coronavirus (nCoV), also known as Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV2), is responsible for this deadly disease. International Committee on Taxonomy of Viruses (ICTV) has declared that nCoV is highly genetically similar to SARS-CoV epidemic in 2003 (89% similarity). Limited number of clinically validated Human-nCoV protein interaction data is available in the literature. With this hypothesis, the present work focuses on developing a computational model for nCoV-Human protein interaction network, using the experimentally validated SARS-CoV-Human protein interactions. Initially, level-1 and level-2 human spreader proteins are identified in SARS-CoV-Human interaction network, using Susceptible-Infected-Susceptible (SIS) model. These proteins are considered as potential human targets for nCoV bait proteins. A gene-ontology based fuzzy affinity function has been used to construct the nCoV-Human protein interaction network at 99.98% specificity threshold. This also identifies the level-1 human spreaders for COVID-19 in human protein-interaction network. Level-2 human spreaders are subsequently identified using the SIS model. The derived host-pathogen interaction network is finally validated using 7 potential FDA listed drugs for COVID-19 with significant overlap between the known drug target proteins and the identified spreader proteins.

Nines, a Palo Alto-based teleradiology startup, received FDA 510(k) clearance for a system that can detect and triage two serious conditions from CT scans. The company's NinesAI system uses machine learning to identify potential cases of intercranial hemorrhage and mass effect, both of which are associated with strokes. The system then flags those cases for expedited review by radiologists. For both of these conditions, every hour counts toward a patient's survival. "This has been a long time coming. A lot of hard work went into it," CEO David Stavens said in a phone interview.

If you're a hospital exec, departmental lead, or run a clinical service, you've likely been approached by a gazillion AI vendors with all sorts of shiny new tech that's just bursting with promise. If so, I know the feeling. My inbox is full of the stuff every single day. To the uninitiated it can be hard to discern which ones are actually going to help (if at all), which are fads and which are just plain dangerous. The wheat needs careful separating from the chaff.

In the fight against COVID-19, several artificial intelligence labs are turning to an unexpected piece of evidence that might help diagnose the illness: people's voices. A team of researchers from Harvard and MIT is using machine learning to comb through voice recordings from COVID-19 patients and healthy people in an attempt to identify specific vocal signatures that could indicate someone is carrying the virus. A similar project is underway at Carnegie Mellon University's CyLab. Research is still in early stages, but the teams aim to develop AI tools that could tell people whether they have coronavirus based on an audio recording of their voice. If proven successful, the tools could allow more people to choose to self-isolate even if they don't have access to a COVID-19 test.

April 22, 2020 -- Siemens Healthineers has received clearance from the Food and Drug Administration (FDA) for its AIDAN artificial intelligence technologies on the Biograph family of positron emission tomography/computed tomography (PET/CT) systems, which includes the Biograph Horizon, Biograph mCT, and Biograph Vision. AIDAN is built on a foundation of patient-focused bed design and proprietary AI deep-learning technology to enable four new features – FlowMotion AI, OncoFreeze AI, PET FAST Workflow AI, and Multiparametric PET Suite AI. Siemens Healthineers PET/CT systems with AIDAN offer enhanced protection against cyber threats via syngo Security – a security package for general regulatory security rules that enables compliance with the Health Insurance and Accountability Act (HIPAA). FlowMotion AI Because each patient's body habitus and presentation of disease is different, tailoring PET/CT protocols to produce the highest-quality diagnostic imaging information possible for each patient can be difficult and time-consuming. The standard one-size-fits-all protocol lacks personalization and is often of suboptimal quality.

The Charisma Machine chronicles the life and legacy of the One Laptop Per Child project and explains why--despite its failures--the same utopian visions that inspired OLPC still motivate other projects trying to use technology to "disrupt" education and development. Announced in 2005 by MIT Media Lab cofounder Nicholas Negroponte, One Laptop Per Child promised to transform the lives of children across the Global South with a small, sturdy, and cheap laptop computer, powered by a hand crank. In reality, the project fell short in many ways, starting with the hand crank, which never materialized. Yet the project remained charismatic to many who were enchanted by its claims of access to educational opportunities previously out of reach. Behind its promises, OLPC, like many technology projects that make similarly grand claims, had a fundamentally flawed vision of who the computer was made for and what role technology should play in learning.