The COVID-19 pandemic has sent shockwaves around the globe and has very quickly presented challenges that the world has never experienced. The massive disruption is affecting billions of people and nearly every aspect of our daily lives. The private sector, as well as government laboratories and research universities are developing cutting-edge technologies to combat the COVID 19 crisis. In some of these efforts, technology developed at Lawrence Livermore National Laboratory (LLNL) is playing a key role in addressing challenges associated with this pandemic.
For example, LLNL developed a rapid PCR technology that enabled Cepheid to provide rapid detection of the current pandemic coronavirus in approximately 45 minutes with less than a minute of hands on time to prepare the sample. High performance computing capabilities available at LLNL allowed scientist to predict the shape of a protein on the surface of the virus so that development of potential drugs and antibodies that recognized the shape could begin quickly, saving precious time in the search for treatments and tests.
Visit LLNL’s COVID-19 website to learn more about how the lab is contributing to the fight against COVID-19.
To help the nation address the serious challenge that the COVID-19 pandemic represents, LLNL has identified the following technologies and capabilities that it is making available under special non-exclusive licenses:
Use of LLNL technology to combat the consequences of the COVID-19 pandemic:
LLNL has developed royalty free licenses to ensure expedited access to its technology and software in the fight against COVID-19. These non-exclusive, royalty free licenses are intended to be non-negotiable and designed to enable rapid adoption of technology by the private sector.
Chest computed tomography (CT)-scans are being used to diagnose both suspected and known COVID-19 cases in patients. CT scans produce digital images that are processed using image segmentation software which identifies and then classifies regions of interest within the images. Traditional segmentation is ineffective and unreliable due to source data typically being noisy, cluttered, or corrupted. LLNL scientists have developed a new strategy that combines classification with randomized segmentation ensembles to significantly improve the overall performance of image segmentation systems. Instead of relying on a single best-guess segmentation, a large collection of potential segmentations are used providing better accuracy and reliability. This patented technique can be used on any host of digital images such as CT scans, X-ray images, and ultrasound images, among others. Improving accuracy and reliably allows physicians to trust the data they are provided, potentially leading to faster and more accurate diagnosis of complex COVID-19 cases and thereby saving lives.
US patent 9,760,801 "Identification of uncommon objects in containers" (LLNL internal case # IL12944)
In current machine learning (ML) models, faulty or corrupt data can lead to incorrect predictions. If these predictions are about whether a patient has coronavirus or not, one could imagine how an incorrect prediction could lead to the spread of the disease or worse yet, the death of a patient. To combat these incorrect predictions, LLNL scientists have developed MimicGAN, a new generation of methods that can “self-correct” for unseen corruptions in the data. MimicGAN serves as a very powerful pre-processing step to ML models deployed in the field. By utilizing MimicGAN, ML models will train on clean and correct data, allowing for more accurate predictions of which could save lives.
US patent application 16/840,358 "Mimicking of corruption in images" (LLNL internal case # IL13386)
When analyzing a dataset, one must not only understand the relationship between the data points, but also the underlying structure of the set. The underlying structure of a dataset is generally estimated from the data on hand, leading to assumptions and less accurate predictions. In order to improve structure learning, LLNL scientists have developed an open source software suite called MTL. Multi-task learning (MTL) aims to improve generalization performance by learning multiple related tasks simultaneously. This software suite can handle any type of data and consists of multitask learning methods and a framework for easy experimentation with machine learning methods, leading to more accurate assumptions and predictions. The MTL software suite has been tested on earth system model outputs and shows that the proposed model outperforms several existing methods for the problem.
MTL-suite, Open-sourced software licensed under the MIT license (LLNL internal case # CP02174)
Drug discovery could be significantly sped up by the integration of in silico methods. To this end, LLNL scientists along with other ATOM Consortium members created the ATOM Modeling PipeLine (AMPL). AMPL is an open-source, modular, extensible, end-to-end software pipeline for building and sharing models. It extends the functionality of DeepChem and supports an array of machine learning and molecular featurization tools. AMPL has been benchmarked on a large collection of pharmaceutical datasets covering a wide range of parameters and has been shown to generate machine learning models that can predict key safety and pharmacokinetic-relevant parameters. By integrating in silico methods such as AMPL into the drug discovery process, patients can obtain safer, more effective drugs faster.
AMPL, Open-sourced software licensed under the MIT license (LLNL internal case # CP02227)
Some COVID-19 diagnoses are utilizing computed tomography (CT)-scans for triage. CT-scans produce immediate results with high sensitivity. The digital images produced by a CT-scan require physicians to identify objects within the image to determine the presence of disease. Object identification can be done using machine learning (ML) techniques such as deep learning (DL) to improve speed and accuracy of disease identification in CT images. Current techniques require images to be the same size and resolution in order to properly train DL algorithms. LLNL scientists have developed a technique which automatically samples across various views and backgrounds to pre-select possible objects of interest. This technique overcomes the limitations of current techniques and provides more efficient object identification, saving physicians time and potential patient lives.
US patent 10,521,699 "Multi-scale deep learning system" (LLNL internal case # IL13158)
Clinical images have a wealth of data that are currently untapped by physicians and machine learning (ML) methods alike. Most ML methods require more data than is available to sufficiently train them. In order to obtain all data contained in a clinical image, it is imperative to be able to utilize multimodal, or various types of, data such as tags or identifications, especially where spatial relationships are key to identification of a clinical diagnosis. To this end, LLNL scientists have developed a method for embedding representations into an image for more efficient processing. Elements within an image are identified, and their spatial arrangement is encoded in a graph. Any machine learning technique can then be applied to the multimodal graph, as representations of the images. These representations can give information such as the proximity of one cell to another allowing the image viewer to obtain knowledge, informing their next decisions. By tapping into the wealth of data in a clinical image, a doctor can gain knowledge that they might not have known prior to applying this patented method, potentially saving time and lives.
US patent application 16/684388 "Universal image representation based on a multimodal graph" (LLNL internal case # IL13464)
Conventional dimension reduction methods aim to maximally preserve the global and/or local geometric structure of a dataset. However, in practice one is often more interested in determining how one or multiple user-selected response function(s) can be explained by the data. To intuitively connect the responses to the data, LLNL scientists developed function preserving projections (FPP), a scalable linear projection technique for discovering interpretable relationships in high-dimensional data. FPP constructs 2D linear embeddings optimized to reveal interpretable yet potentially non-linear patterns of the response functions. Using FPP on real-world datasets, one can obtain fundamentally new insights about high-dimensional relationships in large-scale data that could not be achieved using existing dimension reduction methods.
FPP, Open-sourced software licensed under the BSD license (LLNL internal case # CP02219)
Natural Language Processing (NLP) is a field of study which aims to program computers to process and analyze large amount of natural language data. In order to accurately and effectively utilize datasets in NLP systems, labeled datasets are a must. In cases like pathology reports, the sub-parts of the report are not programmatically labeled. To solve the unlabeled dataset problem, LLNL researchers have developed a software that implements an active learning framework for NLP systems called AL-NLP. It is intended to be applied on scenarios where a limited amount of labeled data is available to train a machine learning-based NLP classification system, but a large set of unlabeled documents exist such as is the case with pathology reports. AL-NLP identifies which unlabeled document should be labeled next so that the overall performance of the classifier is improved. This leads to better NLP systems and ultimately an understanding of the dataset at large. In the case of pathology reports, it helps to label the data within the report so that NLP systems can assist a doctor in analyzing all of a patient's information, leading to a more informed treatment decision.
AL-NLP, Open-sourced software licensed under the MIT license (LLNL internal case # CP02231)
While PCR has had tremendous success in molecular diagnosis, one of the shortcomings is the need to transporting samples to centralized laboratory while maintaining the viability the viability of the sample. There is an unmet need for a device to perform on-site, real-time PCR. LLNL scientists have designed a small and compact device that allows for easy transport of an otherwise not easily transportable assay device. Its real-time aspect lends its usefulness in time-sensitive scenarios. Potential applications of this technology include (i) rapid detection of disease or possible biological causes of injury/death in clinical and non-clinical settings, (ii) rapid law enforcement or battlefield detection of possible bio-warfare/bio-terrorism agents or other diseases, and (iii) further development of availability to be a ubiquitous research and educational tool in addition to its current applications.
US patent 6,699,713 "Polymerase chain reaction system" (LLNL internal case # IL-10517)
Epidemics and pandemics have taken a serious toll throughout history. An ideal method for mitigating the impact of the disease on a population is to identify and treat or isolate affected individuals prior to developing symptoms. This prompts the need for an adequate system for time-of-contact care and pre-contact disease prevention. LLNL scientists have developed a technology which fulfills this need. The technology is comprised of two elements which are to be embedded in a user's personal electronic device (e.g. cell phone, tablet device, pager, etc.). The first is a proximity monitor which transmits location and temporal data such as the distance between the user and a contagious individual and the duration of proximity. The second is a personal exposure notification which comes after the user has been positively diagnosed with a contagious disease by a healthcare provider. Information of their contagiousness is downloaded to a server and the user's device would then transmit exposure warnings to other individuals who have encountered the user and are deemed by the proximity monitor to be at high risk for contracting the disease. Two other factors are critical for the success of this LLNL technology: the accurate disease screening of afflicted individuals by healthcare providers and the early medical response taken by the individuals who have been notified by their device as being potentially afflicted by another individual.
US patent 7,993,266 "Early detection of contagious diseases" (LLNL internal case # IL-10967)
LLNL researchers have invented a system for identifying all known and unknown pathogenic or non-pathogenic organisms in a sample. This invention takes a complex sample and generates droplets from it. The droplets consist of sub-nanoliter volume reactors which contain the organism sized particles. A lysis device lyses the organisms and releases the nucleic acids. An amplifier then magnifies the quantity of available nucleic acids. Then, a fractionator liberates the nucleic acids from the droplets. Finally, a parallel analyzer identifies all the known and unknown pathogen or non-pathogenic organisms in the complex sample. This device functions with DNA or RNA samples. The method has advantages such as (i) speedy and accurate detection and identification of organisms, (ii) parallel analyzer is compact for clinical, field, or research purposes, and (iii) the device can function with small sample sizes.
US patent 8,338,166 "Sorting, amplification, detection, and identification of nucleic acid subsequences in a complex mixture" (LLNL internal case # IL-11652).
LLNL scientists have created a standalone automated diagnostic kiosk that can be placed in public settings, such as in stores or on street corners. Not unlike an ATM in physical size, this kiosk will accept biological samples from an individual for multiplexed analysis. The sample collection process will be sufficiently simple such that anyone could begin the diagnostic process after making the appropriate payment via cash, credit, or debit cards. After the customer signs the appropriate disclaimers concerning diagnosis and liability, a sterile swab or collection tool or vial, viral transport media, instructions on collecting a sample, gloves, and antiseptic wipes would be dispensed to them. The customer then would select what pathogens they want to be screened for before the assay begins. The multiplexed assay itself is comprised of multiple prior IP developed at LLNL and can detect the presence of bacterial or viral pathogens.
The instrument will have the capability to run multiple samples simultaneously. Once the assays are complete, the customer will be emailed or contacted via phone within three hours to be given their results as well as a notice that all results are not official until confirmed by a medical diagnostic laboratory. The results will include a description of the identified pathogen, the typical symptoms associated with it, and the expected course of the disease. Finally, epidemiological data collected over the previous several months by regional machines regarding the percentage of other individuals infected by the same pathogen will be provided to the customer.
US patent 8,647,573 "Automated diagnostic kiosk for diagnosing diseases" (LLNL internal case # IL-11738).
LLNL researchers have developed a new method for faster, more accurate, and precise thermal control for DNA amplification. This technology uses sensor-controlled nodes to monitor and cycle materials through a microfluidic heat exchanging system. Thermal energy travels from a power module through thermal electric elements to sample wells. Sensors coupled to each sample well monitor and respond to predetermined temperature thresholds allowing for the simultaneous directional transfer of thermal energy and therefore better thermal cycling controls. When using LLNL’s solid-state distributed node-based rapid thermal cycler, researchers can be assured that sample DNA is being amplified under optimal conditions. The constant communication between sensors and controllers within the system increases the efficiency of this apparatus’ thermoelectric elements resulting in higher quality samples for better data integrity in a shorter amount of time.
US Patent 8,720,209 “Solid state rapid thermocycling” (LLNL internal case # IL-12275).
LLNL, Sandia National Laboratories, UC Davis and NASA scientists have developed a portable device which analyzes one or multiple types of body fluids or gases to test for one or more medical conditions. A bodily fluid (such as blood, perspiration, saliva, breath, or urine) is put into a condenser surface and is then separated into both a primarily gas fluid component and a second one that is primarily liquid. These two samples from the same fluid or gas source are subjected to analysis by, in various combinations, five different instruments: a condenser, functionalized nanostructures, an optional volatilizer, a differential mobility spectrometer, and an optional biomarker analyzer. Each instrument provides a unique analysis of a physical or chemical element of the tested bodily fluid.
US Patent 9,824,870 "Portable medical diagnosis instrument" (LLNL internal case # IL-13001).
This technology describes a method for performing immediate in-line sample heating to promote the required chemical reactions for amplification, activation, or detection, depending on the thermodynamics of the particular assay involved. The basis of this technology is a method that employ microwave energy absorption to instantaneously heat fluidic partitions without heating the device itself or any oil, or entrapped air. With this invention little energy is wasted heating the device and instead is absorbed heating the aqueous solution within the microfluidic device’s chambers, channels, or reservoirs. This will allow the most efficient, fastest, and best method for energizing chemical reactions in microfluidics devices. Additionally, this system allows for optical addressability of the cavity or waveguide, which allows fluorescence detection of temperature, pH, nucleic acid amplification for PCR, or direct optical observation of cell lysis, sedimentation, and other signals and observations under test for the real-time microfluidic device.
LLNL researchers have developed a high-volume, low-cost diagnostic test that is easy to use and provides results in under an hour. The testing platform will provide emergency responders and other medical professionals with the ability to screen individuals using oral and nasal samples, and obtain results in approximately 30 minutes. This point-of-care testing approach will enable rapid triage of a high volume of patients, without needing to send a sample to a laboratory for testing and then waiting for results. The easy-to-use, compact testing kit consists of a single, disposable tube, which is used throughout the process to collect samples from patients and obtain a positive or negative test result. Using this approach, where all items are contained in a fully sealed, disposable tube, reduces the risk of cross contamination via testing equipment. The diagnostic platform will include a built-in sampling swab attached to the tube’s cap, as well as a buffer that is loaded through a port in the cap, using a syringe. Once the buffer is loaded, the syringe is removed, and the device is sealed and heated isothermally to amplify the pathogen-specific DNA/RNA. Each tube also includes a reagent that is pre-loaded in the tube as a stabilized bead, ready to initiate a chemical reaction once the buffer is loaded. Test results are easy to read. Due to the presence of a colorimetric dye, there is a visible color change that indicates any positive results.
US Patents 9,315,858 “Apparatus for point-of-care detection of nucleic acid in a sample” and 9,222,126 “Methods for point-of-care detection of nucleic acid in a sample” (LLNL internal case # IL-12201A and IL-12208A respectively).
Viral pathogens are the most common cause of respiratory infection including rhinoviruses, respiratory syncytial virus, influenza virus, parainfluenza virus, measles, mumps, adenovirus, and coronaviruses. It also includes viruses responsible for acute infections that often occur as epidemics and pandemics such as Severe Acute Respiratory Syndrome (SARS), Middle East Respiratory Syndrome (MERS) and current COVID-19 pandemic caused by SARS Coronavirus-2 (SARS CoV-2). A method is needed to detect and identify the many types of respiratory pathogens that may affect a patient admitted to a hospital, allowing them to be properly and quickly treated. This LLNL-developed invention is multiplexed and utilizes the Luminex bead-based liquid array, which contains 100 different unique beads. Oligonucleotide probes with sequences complementary to the target sequences are covalently coupled to these unique beads. These capture beads are mixed with viral samples obtained from the patient via cheek swabbing or a throat wash and subjected to PCR in a conventional thermocycler. The amplified target sequence is then hybridized to complementary capture oligonucleotide probes via forward biotinylated primers. If this bead-probe-amplicon unit contains the target nucleic acid, it will be bound by the reporter molecule and fluorescence will be detected by a flow cytometer. This multiplexed assay would thus be able to detect and identify respiratory pathogens present in hospital and clinical settings. It allows rapid, accurate, and convenient point-of-care testing in hospitals and other clinical settings and is also amenable to high-throughput future applications and designs.
US patent 8,232,058 "Multiplex detection of respiratory pathogens" (LLNL internal case # IL-11577).
Global and regional epidemics have progressed from rampant diseases as seen with the Bubonic plague in the 14th century Europe, to the currently ongoing COVID-19 pandemic caused by Severe Acute Respiratory Syndrome Corona Virus-2 (SARS C0V-2). There is an unmet need for a technology to provide fast and accurate identification of unknown pathogens from an often convoluted clinical sample. LLNL scientists have developed a rapid parallel genetic profiling technology that can be used to detect an array of pathogens from a small, complex sample. The device works by first splitting a given sample into millions of emulsified, encapsulated microdroplets each of which are then split once more and run through a parallel analysis consisting of both a genomic and a proteomic assay. The genomic assay includes performing a PCR in droplets followed by agarose gel electrophoresis to identify unique genetic signature of the unknown pathogen. The microdroplets split for physical/proteomic analysis contain DNA and RNA from the original emulsified sample. The droplets are subjected to in vitro transcription and translation to yield proteins which may then be analyzed by mass spectrometry, ELISA, and two-dimensional differential gel electrophoresis, and other proteomic assays. The combined proteomic and genomic analysis results from this LLNL invention allow for heightened specificity in identification of the unknown pathogen sample.
US patent 9,422,586 "Method for genetic identification of unknown organisms" (LLNL internal case # IL-11599).
LNLL scientists have invented a method for multiplexed detection of PCR amplified products which can be completed in a single step. Highly validated species-specific primer sets are used to simultaneously amplify multiple diagnostic regions unique to each individual pathogen. Resolution of the mix of amplified products is achieved by PCR product hybridization to corresponding probe sequences, attached to unique sets of fluorescent beads in liquid. The hybridized beads are processed through a flow cytometer, which detects presence and quantity of each PCR product. The assay is optimized to allow for maximum sensitivity in a multiplexed format. A background PCR product is formed via background multiplex PCR amplification reaction using a control DNA sequence. Comparing the fluorescence of the sample hybridization product with the background product identifies the target DNA sequence. The method is quick and accurate. All-in-one reactions save on labor, reagents, and consumable costs. Multiplexing allows high throughput. Liquid bead array assay can accommodate up to 100 different diagnostic reactions, and fluorophore tagging allows for easy visual detection of DNA
US patent 7,972,818 "Flow cytometric detection method for DNA samples" (LLNL internal case # IL-11701).
Identifying emerging biological threats and epidemics rapidly, particularly concerning viruses, has historically been a challenging task. LLNL researchers have developed a method to quickly and accurately identify the family of a virus infecting a vertebrate via PCR. Universal primer sets consisting of short nucleic acid strands of 7 to 30 base pairs in length were created to amplify target sequences of viral DNA or RNA. These primers can amplify certain identifying sequences of all viral genomes sequenced to date as well as numerous virus subgroupings. The PCR products are separated on a gel by gradient electrophoresis to identify the virus. Altogether, these primers can identify all 28 known virus families that affect vertebrates. Each strain or species of virus produces its own unique electrophoretic banding signature, allowing for easy and quick virus identification. Primer libraries will be updated as new virus families and subgroupings are discovered and classified.
US patent 9,434,997 “Methods, compounds and systems for detecting a microorganism in a sample” (LLNL internal case # IL-11773).
Researchers at LLNL have designed a new technology that allows the integration of a large bench-top thermal cycling instrument onto a miniaturized instrument. This instrument is powered and controlled by portable thumb-drive systems such as an USB. USB thumb-drives are commonly used to transfer data from the instrument onto a PC, however, in this new technology the thumb drive becomes the instrument itself! LLNL researcher’s technology includes thermocycling configured for low power and efficiency, miniaturization of components and controllers, fabrication on a solid-state thumb drive, and integration with USB data and supplied power. This system uses bus power for thermal cycling and bus data lines for data transmission and programming, which allows for portable power. The system can handle 4 wells of 5uL volume cycling with 5 seconds ramp time which is high performance by compassion to benchtop cycles. Additional wells can also be added. The integration of a thermocycle onto a programmable device allows for a portable power, faster and better use friendly software and the thumb drive can be used on the user’s PC.
US Patent Application 2020/0047184 “Integrated solid-state rapid thermal cycling system” (LLNL internal case # IL-12840).
Researchers at LLNL have developed an instantaneous sample heating method to efficiently deposit thermal energy into a continuous stream or segmented microdroplets on a MOEMS device using an optimally low energy, commercially available CO2 laser. The device uses an ideal wavelength (absorption in the far infra-red (FIR) region (λ=10.6 μm)) to instantaneously heat fluidic partitions. The wavelength is absorbed by water molecules and waste little energy because, unlike typical PCR heating elements, the device itself is not heated by the laser. Instead the aqueous solution directly absorbs the heat. This technology is a major improvement over current microfluidic channel heating methods. The use of efficient optical heating elements increases precision and provides users with a cost effective, high-throughput PCR device. It cuts processing time by an order of magnitude resulting in truly real time PCR, in vitro protein translation, immunoassay analysis, etc.
LLNL has developed a new technology that provides a method for near-instantaneous heating of aqueous samples in microfluidic devices. The technology relates to a heating method that employs microwave energy absorption from a coincident low power Co-planar waveguide or microwave microstrip transmission line embedded in a microfluidic channel to instantaneously heat samples. The method heat samples in a focused area within a microfluidic channel on miniaturized chips. Aqueous solution microwave heating allows extremely fast heat transfer for both heating and cooling. This method/device provides a major advantage over current heating methods such as joule-heating from trace resistors which are time-consuming and provide an associated whole device heat build-up. The LLNL microwave heating method provides focused-area heating on an instantaneous time scale. It also provides cost incentive by cutting processing times by an order of magnitude. This method is applicable to on-chip processes such as PCR, in vitro protein translation, immunoassay analysis, etc.
US Patent 10,123,380 “Instantaneous in-line heating of samples on a monolithic microwave integrated circuit microfluidic device” (LLNL internal case # IL-11981).
LLNL scientists have designed a rapid PCR technology that incorporates the use of microfluidic thermal heat exchanger systems and is comprised of a porous internal medium, with two outlet channels, two tanks, and one or more exchanger wells for sample receiving. The wells and their corresponding inlet channels are coupled to two tanks that contain fluid with cold and hot temperatures. A controller is used to dictate the position of the fluid pump’s valves, which directs fluid flow between tanks. The fluid passes though the system’s porous medium, heating or cooling the samples being housed in the wells. When the fluid passes through the matrix, it provides extremely fast heat conduction that enables rapid thermal transfer between the fluid, matrix, and sampler holder. This technique has considerably higher heating/cooling temperatures ramps and it can produce very uniform temperatures utilizing lower input power than prior equipment. By keeping samples at optimal temperatures, researchers can expect a higher sample throughput and better quality, resulting in more reliable data.
In 2005, LLNL researchers won a R&D 100 award for developing advanced technologies to rapidly detect the airborne release of biological threat agents. The Biological Aerosol Mass Spectrometry (BAMS) system is an instrument about the size of three podiums that can analyze individual aerosol particles in real time and at high rates to almost instantly identify the presence and concentration of harmful biological particles in air samples. BAMS was designed for operation in office buildings or at ports of entry such as airports or train stations to monitor for potential epidemic diseases. Biomedical applications could include rapid detection of respiratory diseases such as tuberculosis and coronaviruses.
Multiple issued patents and copyrights (LLNL internal cases # CP01358, CP01357, CP01356, IL10944, IL11492)
LLNL has a successful history of developing instruments for detecting and characterizing airborne pathogens. Often, aerosol characterizing instruments require highly focused particle beams with little or no transmission losses. In addition, they need to interface to the sampling environment with a very high sampling rate so that more aerosol particles can be collected and sensitivity can be improved. The pressure-flow reducing nozzle was originally part of LLNL’s R&D100 award winning Biological Aerosol Mass Spectrometry (BAMS) system. The patented design can be used with other aerosol analysis instruments to perform high-flow, atmosphere-pressure sampling.
U.S. Patent No. 7,361,891, “Pressure Flow Reducer for Aerosol Focusing Devices” (LLNL internal case # IL11478)
To understand complex problems using machine learning it is generally necessary to have large amounts of data. In order to generate these large amounts of data, researchers utilize simulation. Simulations are best run on High-Performance Computers (HPCs) which require various complex processes. To simplify running machine learning based workflows on HPCs, LLNL scientists developed Merlin. The goal of Merlin is to make it easy to build, run, and process the kinds of large scale HPC workflows needed for cognitive simulation. At its heart, Merlin is a distributed task queuing system, designed to allow complex HPC workflows to scale to large numbers of simulations. The workflow software is applicable to any application space and uses other LLNL developed open source software, Maestro. Merlin has been used to study inertial confinement fusion, extreme ultraviolet light generation, structural mechanics and atomic physics, to name a few.
Merlin, open-sourced software licensed under the MIT license (LLNL internal case # CP02250)
The global COVID-19 pandemic and the ensuing shelter-in-place orders has created a unprecedented demand for internet bandwidth taxing the full capacity of existing telecommunications networks. LLNL researchers have developed a practical solution to the challenge that leverages the existing fiber optic telecommunications infrastructure. The lab’s breakthrough enables the potential for installed optical fibers to operate in an untapped spectral region known as the E-band, a nascent capability in all existing installed optical fibers, in addition to the C- and L-bands where they currently operate -- effectively doubling a single optical fiber's information-carrying potential. The innovation is enabled by the successful development of a practical fiber-optic amplifier that generates operationally significant optical gain from 1,390 nanometers (nm) to 1,460 nm with relatively good efficiency. The LLNL E-band amplifier design is based on a novel waveguide, also developed by lab researchers, with resonant leakage elements that frustrate guidance at well-defined and selectable wavelengths. Based on this waveguide, the LLNL team were able to demonstrate an E-band amplifier design with comparable performance and form factor to current commercial C- and L-band amplifiers, thereby validating its commercial viability.
US patent 10,348,050 "Nd3+ fiber laser and amplifier" (LLNL internal case # IL-13120).
US patent 10,033,148 "Waveguide design for line selection in fiber lasers and amplifiers" (LLNL internal case # IL-13058).
US patent pending “Wavelength selective filtering with non-radial array of microstructure elements“ (LLNL internal case # IL-13512).
Complex problems, such as COVID-19, are being studying computational, prior to be tested experimentally. These complex computational problems require HPC resources, of which must be understood and allocated properly. This requires the user to waste valuable computational time just setting up a job on the HPC system. In order to allow computational scientists to focus on the science, LLNL scientists created Maestro. Maestro is an open-source HPC software tool that automates execution of software by defining required multi-step workflows on HPC resources. The core design of Maestro focuses on encouraging clear workflow communication and documentation, while making consistent execution easier to allow users to focus on science. Maestro’s specifications helps users think about complex workflows in a step-wise, intent-oriented, manner that encourages modularity and tool reuse. These principles are becoming increasingly important as computational science is continuously more present in scientific fields and has started to require a similar rigor to physical experiment. Maestro is currently in use for multiple projects at Lawrence Livermore National Laboratory, has been used to run existing codes including MFEM, and other simulation codes, as well as to train machine-learned models.
Maestro, open-sourced software licensed under the MIT license (LLNL internal case # CP01969)
Analyzing the performance and efficiency of complex facilities with modern instrumented components – and the performance of regional networks of such facilities - is a daunting task. Increasingly, facilities collect data from manually input systems as well as diverse Internet of Things sensors and monitoring tools for specialty equipment, storage systems, computing networks, and power/cooling infrastructure. Analyzing the disparate collected data can be intractable. Similar to data from complex hospital facilities, LLNL’s high performance computing center data comprises different formats, granularities, and semantics. Handwritten data processing scripts no longer suffice to transform the data into a digestible form. To aid in solving this issue, LLNL developed Scrubjay, an open source, intuitive, scalable framework for automatic analysis of disparate data. Users can describe the datasets (files, formats, database tables), then describe the integrated dataset(s) desired, and then let ScrubJay derive it in a consistent and reproducible way. Scrubjay may be useful for COVID-19 recovery efforts as an infrastructure analysis tool – such as for performance/availability analysis of large-scale medical facilities.
ScrubJay is distributed under the terms of both the MIT license and the Apache License (Version 2.0). Users may choose either license, at their option. The source code is available at https://github.com/LLNL/ScrubJay (LLNL-CODE-759938)
Protective equipment has always been central to keeping our military, first responders and medical personnel safe. The COVID-19 pandemic is a sobering reminder of the importance of such equipment and the need for improvements. A team lead by LLNL has developed a smart, breathable fabric designed to be incorporated into a garment in small patches to protect the wearer against biological and chemical warfare agents. Material of this type could also be used in clinical and medical settings.
Existing protective garments do not allow for high breathability and thus increases the risk of heat-stress and exhaustion, especially for highly active military personnel and first responders. Dubbed, Second Skin, this new material is both breathable and protective by successfully combining two key elements: a base membrane layer comprising trillions of aligned carbon nanotube pores and a threat-responsive polymer layer grafted onto the membrane surface. The carbon nanotubes easily transport water molecules through their interiors while also blocking all biological threats, which cannot fit through the tiny pores.
High Efficiency Particulate Air (HEPA) filters remove airborne particulates from a gas stream or ventilation system. The life span of the filters is determined by filter design and materials. Existing HEPA filters are made from glass or paper fiber, which are fragile, easily damaged, and have limited shelf life; and in limited cases, metal or Teflon©, which are subject to corrosion, or encounter other regulatory issues.
The LLNL ceramic HEPA filter is designed to be nonflammable, corrosion resistant, and compatible with high temperatures and moisture. The ceramic HEPA filters and materials are tailorable. They can be manufactured to replace existing filters with more durable and longer-lived ceramic versions to minimize retrofit problems, waste and costs while meeting regulatory requirements. Industries utilizing fume hoods and glove boxes can benefit from longer-lived, nonflammable and corrosion resistant ceramic filter. The ceramic HEPA filter was designed to meet commercial and DOE requirements, as well as to minimize upgrade installation logistics for use in existing facilities. Current performance requirements are described in DOE Standard 3020. The ceramic filter will significantly increase filter life span and reduce life cycle costs of standard filtration systems, but also can enable new overall process gas system and ventilation system design. Ceramic HEPA filters open the doors for new applications of HEPA filtration in numerous industries, such as biotechnology. Other applications are nuclear reactors, radiological facilities and other hazardous material processing facilities encounter nontrivial contamination issues and life cycle costs (both operational and waste disposal) for filters and affiliated support systems.
Cardiotoxicity is one of the major toxicity concerns when developing new drugs. However, these cardiotoxicity tests aren’t done until a drug has gone through years of development. LLNL scientists have developed a software suite called Cardioid that simulates the electrical current running through the heart tissue, triggering cells to contract like cascading dominoes and causing the heart to beat. These simulations, which generate copious amounts of virtual data, can be used to train a patented machine learning system. Once trained, actual clinical results could be used as a ground truth to develop a more accurate ML system that can determine how the heart is functioning. Cardioid along with the patented ML system have the potential to make drugs safer, save time, money and lives.
Cardioid, open-sourced software licensed under the MIT license (LLNL internal case # CP02134) and US patent application 62/950885 "Machine learning based reconstruction of intracardiac electrical behavior based on electrocardiograms" (LLNL internal case # IL13207)
Computed tomography (CT) is one of the most common imaging modalities used in industrial, healthcare, and security settings. During a CT scan, a narrow beam of x-rays is used to produce signals that are processed by a computer to generate cross-sectional images of a part of the human body, such as the lungs of a suspected COVID-19 patient or a patient in recovery needing long-term rehabilitation. With multiple tomographic images, they can be digitally stacked together to form a 3D image. However, when the number of image projection is small, streak artifacts can pollute the reconstructed image. Correcting the images takes more time to process and may not improve the image presented to the healthcare provider for diagnosis.
To solve these issues, LLNL has developed an innovative software package for CT reconstruction called Livermore Tomography Tools (LTT). LTT implements advanced algorithms to build 3D images of an object using just a few views, compared to the thousands of views that are typically necessary for traditional CT scans. LTT is platform independent and capable of processing data on one or more graphical processing units (GPUs) or other hardware accelerators. It can be used as a stand-alone application, accessed as a library from existing applications, or used with a separate graphical user interface. LTT provides quantitively accurate results independent of the system, and its flexibility allows data to be processed from any CT geometry, independent of the computing platform. By reducing the time needed to properly reconstruct a 3D image, LTT can increase throughput for medical screening using CT scanning and may significantly reduce the number of scans required—which further reduces both patients’ exposure to radiation and the operational costs and time demands on healthcare providers.
US patents pending, LLNS copyrights asserted
One of the biggest challenges in many fields of studies, such as COVID-19, is to analyze a complex mix of experimental and simulation data, which relies primarily on the intuition of trained experts. Many advanced analysis techniques are often difficult to integrate, leading to a confusing patchwork of analysis snippets too cumbersome for data exploration. To simplify data analysis, LLNL scientists developed a web-based software system that consists a combination of techniques from statistics, machine learning, topology, and visualization. The NDDAV (N-dimensional data analysis and visualization) framework integrates traditional analysis approaches with state-of-the art techniques and custom capabilities which are linked into an interactive environment that enables an intuitive exploration of a wide variety of hypotheses while relating the results to concepts familiar to the users. NDDAV’s modular design provides easy extensibility and customization for different applications, like COVID-19. The NDDAV framework allows for integration of trained expert intuition while analyzing complex mixtures of experimental and simulation data leading to faster, more easily understandable data analysis and visualization.
NDDAV, Open-sourced software licensed under the BSD license (LLNL internal case # CP02056)
Communicating complex scientific information is a critical activity in responding to today’s COVID-19 pandemic. Many sources exist now to present trustworthy and timely information in ways that decisionmakers and the general public can understand. One way to communicate scientific information is to show the technical data in visual forms that users can easily relate to and interact with. LLNL’s VisIT software has long been a leading tool for enabling scientists to visually show complicated scientific data. VisIt is an open source, interactive, scalable, visualization, animation and analysis tool. From Unix, Windows or Mac workstations, users can interactively visualize and analyze data ranging in scale from small (<101 core) desktop-sized projects to large (>105 core) leadership-class computing facility simulation campaigns. Users can quickly generate visualizations, animate them through time, manipulate them with a variety of operators and mathematical expressions, and save the resulting images and animations for presentations. VisIt contains a rich set of visualization features to enable users to view a wide variety of data including scalar and vector fields defined on two- and three-dimensional (2D and 3D) structured, adaptive and unstructured meshes. Owing to its customizable plugin design, VisIt is capable of visualizing data from over 120 different scientific data formats
VisIT is distributed under a BSD 3-Clause License and is available at https://wci.llnl.gov/simulation/computer-codes/visit. (LLNL-CODE-793424)
Understanding how proteins interact with membrane surfaces is important for drug discovery studies in which a drug may target a membrane protein. One of the main proteins of interest for COVID-19 antibodies is the ACE2 protein that binds to the neutral amino acid transporter B0AT1. B0AT1 sits in the membrane and understanding how movement or perturbation of that membrane might after the binding sites available on the ACE2 protein is important in creating effective antibodies to prevent viral infection. To study the dynamics associated with membranes, LLNL scientists created MemSurfer. MemSurfer is an efficient and versatile tool to compute and analyze membrane surfaces found in a wide variety of large-scale molecular simulations. MemSurfer works independent of the type of simulation, directly on the 3D point coordinates, and can handle a variety of membranes as well as atomic simulations. MemSurfer provides many in-built analysis tasks, such as computing the membrane curvature, density and normals of lipids, and area per lipid. More importantly, MemSurfer provides a simple-to-use Python API that may be easily used/extended to perform other types of analysis. To get a complete picture associated with COVID-19 antivirals, computational studies should investigate the involvement of the membrane and MemSurfer can be used for those computational analyses.
MemSurfer, open-sourced software licensed under the GPL-3.0 license (LLNL internal case # CP02099)
Automating protein classification via structural similarity has been a technique employeed by researchers for a while. The current methods generally only assess structure similarity using a single metric (e.g., Z-score) and only evaluate similar conformations of secondary structure elements. In order to accurately access structure similarity, LLNL scientists created a method called STRucture ALigment-based Clustering of Proteins (STRALCP). STRALCP is a structure alignment-based approach invented for the purpose of automated protein structure classification. For a given set of proteins, STRALCP generates detailed information about global and local similarities between pairs of protein structures, identifies the fragments that are structurally conserved among the proteins, and uses these fragments to classify the structures accordingly. This new method overcomes the limitations of previous methods and more accurately classes proteins which is a key step in studying proteins that might lead to vaccine development.
US patent 8,467,971 "Structure based alignment and clustering of proteins (STRALCP) " (LLNL internal case # IL11696)
Understanding proteins, their structures, and how they may be similar is necessary for many applications from basic science to developing vaccines for COVID-19. Most computational models that predict protein structure similarity consider certain features at the expense of others. To get a holistic picture of protein structures, LLNL scientists developed the Local-Global Alignment (LGA) model. The model works by predicting protein structures by considering both local and global structures without compromising either feature. To do this, LGA can use data corresponding to clusters and/or fragments of proteins. The data can be inputted manually or for ease of use, uploaded from the Protein Data Bank (PDB). Not only can LGA predict protein structure similarity, it can also predict sequence similarity. It has been reduced to practice and shows better performance in comparison to known computational models for predicting protein structures such as DALI, CE, and ProSup.
US patent 8,024,127 "Local-global alignment for finding 3D similarities in protein structures " (LLNL internal case # IL11160)
In order to identify new, unknown proteins associated with viruses, such as COVID-19, it is easiest to start by identifying structurally related proteins. LLNL scientists have created tools that identify structurally related proteins and their relevant residues, called cSpan. The cSpan (sequence conservation in structurally conserved “span” regions) calculation is a quantitative measure of residue conservation in local structure context. It is used to identify residues on a protein that are conserved with respect to a set of structurally related proteins. A set of protein structures (consisting of a reference protein, and any number of related proteins) is aligned using the LGA (local-global alignment) software (see Local-Global Alignment: A Method for Finding 3D Similarities in Protein Structures). A multiple structure-based residue-residue correspondence (or “multiple structure-based sequence alignment”, MSSA) is extracted from the structural alignments and corresponding (structurally aligned) residues are compared. Each residue in the reference protein is scored (assigned cSpan value) according to how similar it is to each corresponding residue in the set of related proteins. The reference protein’s cSpan values can be plotted vs. the residue number to identify conserved sub-sequences, consisting of high-cSpan residues. Additionally, they can be projected onto a 3-D structure or model to assist in identification of features conserved in sequence and structure. Our cSpan algorithm (combined structure- and sequence-based analyses) can be used to identify and characterize surface features of interest in development of diagnostic reagents, therapeutics, or vaccines, and to functionally annotate pathogen proteins.
US patent 8,452,542 "Structure-sequence based analysis for identification of conserved regions in proteins" (LLNL internal case # IL11776)