Amazon Omics: A new age of clinical research is rising
Discover the latest clinical research trends in Amazon Omics, and explore how the phenomenon revolutionizes data analysis and precision/personalized medicine.
for AI. Special
Artificial Intelligence (AI) is a black meteoric star for most industries. Biotech and Pharmaceuticals are no exceptions. But, do you really know how pharma AI can affect the world with regards to the latest global developments?
Chinese technology whale Alibaba has created an AI system that can detect coronavirus (COVID-19) infections with 96% accuracy. It takes just 20 seconds for the AI system to determine whether it’s a coronavirus, which is impressively quick when compared with detection by a physical examination, that takes on average 15 minutes.
Moreover, just before 2020 New Year’s eve, an AI-powered disease-alert software in the US sent out the first global alert about the coronavirus outbreak in Mainland China. The system is called HealthMap and it monitors online news and social media posts that were talking about “unidentified pneumonia”. Early monitoring of social media activity and online publishing gives healthcare express a head start – when bureaucratic difficulties and language complications might get in the way.
With the help of coronavirus maps, biotechnology and pharma researchers have suggested a couple of possible scenarios to explore new treatments to combat COVID-19. For instance, creating antibodies that target concrete sites on angiotensin-converting enzyme (ACE2) protein can suggest methods to hold down the virus. This means that injecting proteins that incite an immune response could prepare the body to resist COVID-19 infection.
Gilead Sciences already issued a remdesivir treatment to treat one infected patient in the US from coronavirus. In late March 2020, a study on 1,000 patients in Asia will identify whether remdesivir can reverse the coronavirus infection.
Moderna Therapeutics developed a mRNA-1273 vaccine candidate against COVID-19 42 days after coronavirus was sequenced and set an industry record. If the mRNA-1273 vaccine is proved to be safe, thousands of patients will be enrolled in testing whether the vaccine protects against COVID-19.
The correct dose-finding is a highly important step for the clinical development of a new treatment. Detecting a no-effect dose and the mean or maximal effective doses requires a lot of effort, time, and a considerable number of patients to test it out. Especially when we take into account that during drug development, patients are exposed to a maximum tolerated dose (MTD), which is higher than they need.
A good example of precision medicine platforms is Curate.ai. It allows you to find an appropriate treatment combination and optimal drug dosing strategy over time, based on the scant data collected, specifically from the treated person. Such an approach shows excellent results over the current standards of dose-finding. Moreover, using a precision medicine platform minimizes the side effects and prolongs the quality and length of the treated person’s life.
Pharmaceutical companies often overestimate the pool of available patients that meet enrollment criteria. According to Lasagna’s Law, the patients’ availability dramatically lowers when the clinical trial starts and then returns back to the starting level as soon as the trial is finished.
What is more, clinicians don’t have any easy way to look for eligibility criteria for clinical trials, for their patients. Likewise, clinical trial companies do not have all the resources to effectively identify applicable potential patients for their trials.
IBM Watson, an intelligent AI assistant, solves this issue by finding all suitable clinical trials for the eligible patient. And for the clinical trial offices, IBM Watson helps to find the patients that are eligible for any of the current clinical trials. This improvement saves clinicians time spent on patient screening and eliminates the necessity of finding the appropriate patients manually. If you’d like to build a similar application for your company, get in touch with us.
The COVID-19 case has uncovered an interesting situation: quick expansion of the virus, which is extremely challenging to the world’s healthcare systems, and that it has a high mortality rate. But why? Remember that the Chinese government built new fully functional hospitals in a couple of days and a lot of people were wondering what’s happening, why they are doing it, and whether their hospitals were already full? What happened there is known by every software architect and it is called “scalability” – Chinese authorities scaled their healthcare system in order to prepare for handling more patients. But this is rather expensive and inflexible because even if the facilities can be built quickly, the doctors can’t be staffed in the blink of an eye and different types of medical equipment can’t be produced instantly. How do we change things? A solution is sitting right under our noses: digitalization. The more digital the system, the more scalable it is. To speed up your business digital transformation, get in touch with us.
In the coronavirus situation a simple chatbot for early diagnostics and information gathering can help thousands of people to understand the disease better. So far, anyone can obtain more information on COVID-19 from the following chatbots:
Biosensor devices, based on optical nanotechnology, can detect the coronavirus within 30 minutes; straight out of the patient’s biosensor sample without testing it in centralized clinical laboratories. The technology is capable of swiftly determining whether it’s COVID-19 or pneumonia. What is more, the technology behind the biosensor device will be used to identify various types of coronavirus in animals, so the possible coronavirus evolutions among humans can be monitored and prevented.
Additionally, the coronavirus AI surveillance monitoring technology, Biofourmis, allows clinicians to track the infected and suspected patients, and apply precision medicine to provide more efficient drug interventions. This AI platform will help to obtain more comprehensive epidemiological understanding of COVID-19 and assist in developing effective novel treatments.
Volunteer participants will wear built-in sensors on their upper arm for 24/7, so the device can track patients’ body temperatures, respiratory and heart rates, blood oxygen levels and transfer the data back to the Biofourmius digital platform for live monitoring, analysis and forecasting. With the help of AI and ML techniques, the platform identifies the key health transformations that would signify the COVID-19 progression.
According to Statista, right now in 2020, 356.83 million users put on wearable devices every day and this number is expected to grow. Additionally, 25% of the adult population in the US will be using wearable devices by 2022.
Many commonly use wearable devices to monitor their health, such as heart rate, steps walked, calories burned, etc. However, wearables can do much more than that. Thanks to the progressive AI algorithms, miniaturisation of biosensors, and elasticity of electronics, wearable devices can produce real-time data within the IoT.
Moreover, wearables can be set up to make tight contact with epidermal, intracochlear, ocular, and dental interfaces and are able to gather biochemical and electrophysiological signals. All these medical data and signals can be used for monitoring patients at risk, intervening in disease evolution at the early stages, and reducing healthcare expenses by predicting and restraining the diseases from further progress.
Personal connected devices and biosensors are used for various pharmaceutical purposes, starting from aiding patients with metabolic disorders to those with disabilities. Wearables can be customized to diffuse drugs through the skin, eyes or ears. Biosensors integrated into wearables formulate closed-circuit systems, where the dosing of drugs can be managed over a wireless network and communicated to the end-user.
Wearables and biosensors often function via close-proximity protocols, like Bluetooth Low Energy and share the data to smartphones and tablets and are hardly accessible for exploits and secure.
Deep learning methods can extract COVID-19 graphical features, based on radiographic changes in computer tomography scans of people infected by the coronavirus. These features can generate a clinical diagnosis prior to a pathogenic test, saving critical time for disease control. A deep learning algorithm has achieved 90% accuracy by analyzing 1,119 computer tomography scans of people diagnosed with COVID-19, along with the same scanned number of people diagnosed with pneumonia. This means that a deep learning algorithm has confirmed the POC to extract radiological features for prompt and precise coronavirus diagnosis.
Furthermore, the more inexpensive chest X-ray radiographs can also be used to detect COVID-19. Using X-rays rather than CTs is especially relevant for rural areas, where people can’t get a CT that easily. The trained model was capable of predicting, with 99% accuracy, that the lung x-rays are positive for coronavirus.
AI and neural networks algorithms can accelerate and drive drug discovery in 2 ways:
For the Coronavirus Disease, Google’s DeepMind released computational predictions of the unexplored protein structures of the SARS-CoV-2, the virus that leads to COVID-19. The larger the protein is, the harder it becomes to model it, as it contains a plethora of interactions between amino acids that need to be taken into account. According to Levinthal’s paradox, it will take longer than the age of the universe to randomly enumerate all potential configurations of a classic protein prior to getting to its 3D structure, because proteins fold sporadically, within several milliseconds. Once the protein structure is understood, the role it has within the cell can be estimated and researchers can develop medicines that work with specific protein structures.
Thanks to the substantial reduction of the cost of genomic sequencing, deep neural networks are catalysing the progress in protein shape discovery. AlphaFold (Google DeepMind’s department) has successfully utilized deep neural networks to accurately predict the experimentally identified SARS-CoV-2 protein structure stored in the Protein Bank. AlphaFold has already provided the general research community with the predictions of under-studied proteins that require laborious efforts for template modelling and they are open for everyone now.
These protein structures are already helping researchers to cross-examine coronavirus functions, and aid the hypothesis generation system for future laboratory work in discovering therapeutics. So far only half of human proteins have been mapped. Discovering mutations in a single gene and uncovering the shape of malformed proteins can considerably contribute to effective drug development and help scientists to create new and efficient cures quickly.
→ Read our article on Pharma Manufacturing – Improving the risk-reward calculus for clinical trials: How natural language processing and machine learning can boost success in drug development by Michael DePalma and Igor Kryglyak
The typical drug development cycle lasts 15 years and costs 2.6 billion USD per drug. ML and especially deep neural networks are now disrupting the pharmaceutical industry by leveraging the power of algorithms to discover better treatment mechanisms, quicken the process of drug discovery and lower its cost.
Drug discovery is a complex mathematical problem as it involves hundreds of thousands of possible molecule combinations. Deep neural networks can be leveraged to speed up the entire drug discovery process by predicting molecular wave functions and molecular electronic properties. Resolving these tasks by conventional methods requires months of computing time, which is often an obstacle in drug discovery processes. Deep neural networks and AI algorithms enable the drastic acceleration of drug simulation efforts.
The biggest advantage deep learning provides for drug discovery is that it enables the design of a neural network architecture that is custom-tailored to a particular task.
The current issues with deep learning in pharma and drug discovery are that they require huge datasets for training. Learning with a small amount of data is a question of major significance for ML in drug discovery.
In addition, deep learning algorithms are capable of solving adverse problems in the pharmaceutical industry by assisting in bioactivity prediction, de novo molecular pattern identification, reaction prediction, and image analysis.
After the COVID-19 outbreak, a consortium of technology leaders developed an AI-enabled data set that has organized 30,000 articles on the coronavirus disease. The open research data set, named CORD-19, aims to integrate tens of thousands of research papers on the Coronavirus Disease 2019 in the common format. The data set is a new platform for open science, where biotechnologists, pharmacists, epidemiologists and big data professionals can find the data they need and create impact. The AI-powered dataset converts all the data into a machine-readable form and produces an agile feed that keeps scholars on the right page in the COVID-19 research areas that have been left untackled.
The AI database on the coronavirus is kept updated and refined as new research appears. Moreover, academic research is now linked with clinical trial data. The CORD-19 database is available to 4 million data scientists in the world, all eager to find answers to the behaviour of the virus.
Every pharmaceutical research generates genetic, metabolomic, phenotyping, proteomic, transcriptomic, spectroscopic, and other types of data. Further, the treatment development processes produce even more data on pharmacokinetics, toxicity, efficacy, etc. This data comes from different sources varying in scope and format. Integrating all this information into one system in order to receive the most exhaustive and all-inclusive results is an extremely time and effort consuming task.
However, having a data integration system in place enables an extensive search in the collected data subsets based on linkages between the data. Smart algorithms, including ML and AI technologies together with BI platforms, can assist in creation of automatic reports, raise red flags about safety, as well as generate insights.
The ultimate goal of the data integration system is to generate a clean, consistent and timely analytics for large scale pharmaceutical projects and help manage pharmaceutical data 24/7. Proper data integration serves as an essential tool to other processes such as forecasting and reporting.
Data integration resolves the following problems:
Having a data integration system in place can foster communication and collaboration between pharmaceutical departments, provide better customer engagement and increase productivity. Improved communication between internal and external stakeholders offers new opportunities and provides access to industry-leading clinical studies or simulations. This consequently leads to novel insights about clinical outcomes and offers precision medicine opportunities. Connect with us and get an expert consultation on how you can create a data integration system or revamp your collaboration systems.
The efficient use of big data aggregation and analysis allows pharmaceutical companies to determine new drug candidates quicker and develop useful and approved treatments faster.
Thanks to the big data collection by Taiwan as it mitigated the spread of the coronavirus. Taiwan was ranked as the second most risky region for the COVID-19 infection after Mainland China, as it’s located just 130 kms away from China. By implementing big data analytics, Taiwan’s government monitored the spread of the coronavirus, tracked 14-day travel histories and symptoms of all its citizens, restricted the entry of foreign travelers, and issued special government-tracked smartphones to all the people placed under quarantine. The cross-department big data platform can track people’s movement and integrate the collected information from the police, health institutions and government, and alert the responsible people to take action where the COVID-19 spreads.
The AI can map COVID-19 infection clusters to help governments tame the disease. Early identification of infection areas can stop the coronavirus from spreading and slow it down substantially. The best option to map infection clusters would be conducting a huge number of tests. However, it’s not possible to conduct such intensive testing in third world countries.
Therefore, Israeli researchers compiled questionnaires that contained information on different symptoms and the places people stayed. The collected data will be analyzed by ML algorithms and then be provided to government and health agencies giving sufficient information about the spread of the coronavirus disease.
As soon as one COVID-19 infection cluster is identified, the ML algorithms can assist in forecasting any prospective infection areas before coronavirus symptoms become obvious, and assist in restraining the disease from spreading.
According to the McKinsey Global Institute the implementation of big data collection mechanisms and procedures could result in $100,000 savings across the pharmaceutical industry, enabling better efficiency of clinical trials and drug research.
What is more, the next-gen data modelling and sequencing techniques produce a lot of medical data instantly. The plethora of generated data and progressive analysis techniques can revamp the pharmaceutical industry and bolster the drug development process. For example, genomics enables pharma companies to run an increasingly precise approach to drug discovery and will incite safer and more efficient drug development.
Just a couple of years ago only a small number of human genomes had been sequenced. Right now it’s changing rapidly. Illumina, the leading DNA developer, claims to have sequenced the 65,000 human genomes and to have lower the cost of human genome sequencing to just $100.
Artificial Intelligence (AI) and Machine Learning (ML) technologies have greatly evolved in the last couple of years and are capable of performing tasks that were unimaginable for a computer to do. AI and ML are bringing a paradigm shift for the pharmaceutical industry, transforming how healthcare professionals discover new medications, diagnose, and treat diseases.
With the help of big data, AI, business intelligence and advanced analytics, pharma companies can achieve the following objectives:
At Avenga, we have extensive expertise in delivering enterprise solutions for pharma and life sciences. Our best-in-class consultancy, design and technology services can drive a comprehensive transformation of your medium-sized business into a multinational corporation. We design, develop and deploy top-notch technology platforms that reinforce your next breakthrough.
The Avenga’s team of certified Data Experts has successfully utilized the latest data warehousing and consolidation techniques and cutting-edge systems that boost ROI and take a business to a new level.
Get in touch with us using this contact form, and we’ll be happy to assist you in making sense of AI for better decisions.
Discover the latest clinical research trends in Amazon Omics, and explore how the phenomenon revolutionizes data analysis and precision/personalized medicine.
Explore an open-source approach to clinical reporting supported by leading industry companies.
Discover how to detect and stop pharmaceutical and healthcare fraud with advanced AI-based instruments so as to secure patient care and maintain the industries’ integrity.
Discover the latest and most promising healthtech innovations to pursue in 2023, while addressing the challenges preventing the adoption of these tech trends.
Get in-depth insights into a step-by-step process coupled with a real case scenario depicting data analytics frameworks handling vast amounts of data.
Explore the top 10 financial industry tech trends to obtain a competitive edge through harnessing the power of ongoing and upcoming innovations.
Learn what ChatGPT, the successor of the very successful GPT-3, is capable of when it comes to personal and corporate usage.
Learn how Stable Diffusion generates images and why it significantly outperforms other generative models.
Start a conversation
We’d like to hear from you. Use the contact form below and we’ll get back to you shortly.