Happy Summer ๐ to Everyone!
And welcome back to another edition of MetaphysicalCells.
AI/ML Drug Development: podcasts and articles
๐ฃ Sean McClain and Joshua Meier, respectively the Founder/CEO and Chief AI Officer of Absci a public company harnessing generative AI to create more effective medicines faster and less expensively, are explaining during this podcast from ARK Invest (ARK is offering investment solutions to capture disruptive innovation in the public equity markets) how their approach to antibody production focused on data analysis really works. In particular, the key points from this episode are:
How Absci is making the process of drug creation much more efficient.
The most important aspects to consider when developing an antibody.
A high-level overview of how Absci generates data.
The number of AI-generated designs that Absci can validate in any given week. Metrics that Absci uses to evaluate their models.
A common problem in the AI field.
How the models Absci is developing can be applied in different contexts.
How AI is likely to change the way we approach scientific and technological developments.
๐ฃ A new brief overview of common AI models in the field of drug discovery, titled โArtificial Intelligence in Pharmaceutical Sciencesโ, was just published that summarises and discusses in depth: 1) their specific applications in various stages of drug R&D (such as target discovery, drug discovery and design, preclinical research, automated drug synthesis), 2) their influences in the pharmaceutical market and 3) the major limitations of AI in drug R&D as well as possible solutions are proposed.
In particular, you can find in this review a table regarding all pharmaceutical databases focusing on proteins, genes, drugs/drug targets, and diseases:
๐ฃ Astera Institute, that was established by an American programmer, entrepreneur and philanthropist Jed McCaleb, is a 501c3 non-profit dedicated to developing high leverage technologies that can lead to massive returns for humanity. Right now they target areas of latent potential and support programs in Artificial General Intelligence, Science and Climate without a natural home in the existing innovation ecosystem.
Michael Nielsen (working on metascience, programmable matter and tools for thought), is a Research Fellow at Astera Institute and just published a blog ๐งโ๐ป, titled โHow is AI impacting science?โ. The text is from a talk given at the Metascience 2023 Conference in Washington, D.C., May 2023, and is about AlphaFold acting as a bridge to a new era, opening up many scientific and metascientific questions.
Metascience (also known as meta-research and often found in the literature hyphenated as meta-science) is the use of scientific methodology to study science itself.
Metascience seeks to increase the quality of scientific research while reducing inefficiency.
AI Drug Discovery News ๐ฐ๐๏ธ
๐ Researchers at MIT and McMaster University have successfully identified by using AI a new ๐ antibiotic (abaucin) that can be used for many drug-resistant infections (AI CREATES KILLER DRUG).
In particular, the researchers used training data from measurements of known drugsโ action on the drug-resistant bacteria, and then they projected the effect of 6,680 compounds with no data on their effectiveness against the germ with the learning algorithm. Luckily ๐ค, the new antibiotic seems only to be effective against the target microbe, which is a plus, but it isnโt available for people yet.
๐ Google Cloud has just announced two new AI-powered life sciences solutions:
the Target and Lead Identification Suite, allowing researchers to better identify the function of amino acids and predict the structure ๐ of proteins. This solution also enables lead optimisation that can be used to discover novel, high quality candidates at low cost; for quantitative structure-activity relationship (QSAR) studies; or for free energy perturbation (FEP) calculations. And
the Multiomics Suite, accelerating the discovery and interpretation of genomic data for precision medicine (by storing, processing and analysing genomic data for analysis in a scalable, cost-effective and secure manner).
Combined the two novel solutions will accelerate drug discovery and precision medicine (Google's latest AI tackles long and costly drug discovery).
In particular, the Target ๐ฏ and Lead Identification Suite is using
AlphaFold and Vertex AI pipelines to accurately predict protein structure,
cost-effective high performance computing resources to accelerate target discovery and preparation of lead candidates, and
virtual high-throughput screening.
While the Multiomics Suite
ingests raw sequences files with genome-wide association study (GWAS) pipelines into Cloud Storage,
extracts variants using Batch API,
and accelerates the process of analysis by using Compute Engine to turn raw sequencing (DNA/RNA) data into actionable insights, and identify genes associated with a particular disease or traits to be integrated into multimodal datasets.
๐ Retro Biosciences is a longevity ๐ด๐ตcompany focused on cellular reprogramming, autophagy and plasma-inspired therapeutics, which will all be supported by AI since the company received last year a $180M in funding from an undisclosed source, revealed two months ago to be Sam Altman of OpenAI, and others.
Retro Biosciences founded by Joe Betts-LaCroix (he previously cofounded Vium to accelerate the development of new medical therapies by automating in-vivo research, and acquired by Recursion in 2020) is a ML based computational biology company with lab automation working on aging mechanisms for which interventions have shown robust proofs of concept in mammals and have a feasible path to translation to humans.
And since we are talking about Sam Altman, he and other AI tech giants (all of them having no intention to live on planet Mars ๐), issued this week a warning of advanced AI as โextinctionโ risk.
๐ BenchSci, a Toronto based AI company that is offering a platform called Ascend to speed up the drug discovery process, just raised $70M to accelerate drug discovery with AI. Generation Investment Management was the lead investor โฝ while iNovia Capital, TCV, Golden Ventures and F-Prime Capital participated as well, bringing BenchSciโs total funding to $170 million.
BenchSci, by using AI and proprietary visual ML, has built ASCEND (Google-backed โmapโ) a revolutionary preclinical R&D platform that harnesses patented ML technology trained by scientists to extract experiment evidence from internal and external data sources. It has been estimated that, companies using ASCEND save up to $6 million per year in hard costs alone. Costs, that are based on proprietary customer spend analysis that depends on a number of factors, including a companyโs total annual reagent spend and its waste ratio.
So far, 16 top pharma companies are using BenchSciโs ASCEND, along with over 50,000 scientists working at more than 4,500 institutions worldwide.
๐ Aitia, the leader in the development and application of Causal AI and "Digital Twins", just Entered into Multi-Year AI-Driven Drug Discovery and Drug Simulation Collaboration with Servier for Pancreatic Cancer. The new project with the French pharma group Servier will focus on pancreatic cancer, and builds on a collaboration in the blood ๐ฉธcancer multiple myeloma that was started by the two companies last year.
Aitia Bio (ex GNS Healthcare) is known for creating Gemini Digital Twins โ by combining multi-omic and clinical patient data and by using Aitiaโs patented causal AI and simulation technology, named REFS. The โDigital Twinsโ Aitia is creating can link patientsโ characteristics to diverse drug treatments, in order to reveal complex genetic and molecular mechanisms and pathways driving clinical outcomes. For now, Aitia is advancing drug discovery programs in multiple myeloma, prostate cancer, Alzheimerโs Disease, Parkinsonโs Disease and Huntingtonโs Disease. @Aitiabio has 12 investors including Celgene and Cigna Ventures and has raised to date $77.3 million.
๐ CHARM Therapeutic a 3D deep learning biotechnology company has developed the proprietary DragonFold technology, an end-to-end algorithm that predicts protein/ligand co-folding to discover novel transformational medicines for hard-to-drug targets. DragonFold was inspired by the breakthroughs of CHARMโs co-founder, the Wiley Prize winner David Baker, an American biochemist and computational biologist who has pioneered methods to predict and design the three-dimensional structures of proteins. After predicting a 3D protein structure, at CHARM they use their state-of-the-art lab facilities to validate all their AI-derived medicines in an integrative loop, striving to bring them to patients as fast as possible.
On May 15, 2023, @CHARMTherapeutx announced ๐ข an Investment for Deep Learning-enabled Drug Discovery Research from NVIDIA. The investment will enable CHARM to continue to fund its organisation while harnessing the power of NVIDIAโs cutting-edge GPUs to rapidly identify potential new compounds using its proprietary DragonFold platform.
Moreover, just two months ago CHARM Therapeutics announced ๐ฃ๏ธ a strategic discovery collaboration with Bristol Myers Squibb for the identification and optimisation of compounds against Bristol Myers Squibb selected targets.
London-based CHARM has raised $70 million to date and it is supported by high quality international investors including OrbiMed, F-Prime Capital, General Catalyst, Khosla Ventures, Axial, Braavos, NVIDIA and grep -vc.
๐ VeriSIM Life (founded in 2017 by Dr. Jo Varshney, a veterinarian with a Ph.D. in genomics and cancer biology) is a company building AI enabled biosimulation models combining chemical and biological modelling along with AI and ML techniques and data in order to provide a Translational Indexโข, that can predict how drugs will affect patients and optimise portfolio ๐management as well as increase clinical success. This BIOSIM platform is offering:
scalable models from data from thousands of compounds across 7 species,
earlier insights since can be used before preclinical animal ๐ trial start (by replacing the need for R&D costs related to animal testing could save U.S. companies some $20 billion a year collectively and spare countless animals from testing),
AI and mechanistic models,
great translatability,
a continuous improvement process to iterate on data stability, improved models and an always up-to-date software service, and
flexibility of the platform, generating customer-ready tools that allow for the development of custom models and validations within weeks.
Last month, they published a blog about AI technologies to use in clinical trials (โHow to Invest in AI to Improve Clinical Trialsโ) such as:
AI-based technologies to create better structured, standardised and normalised data elements across a variety of inputs and sources.
AI-based technologies to digitise and automate clinical trial processes, helping wrap studies faster and getting medicines and treatments to patients sooner.
AI can also be used to create patient centred study designs, managing complexities to reduce patient burden, increasing compliance, and creating a range of new efficiencies across study processes.
AI can be used alongside with wearables to generate real-time clinical insights, reduce trial drop out rates, maximise patient adherence and minimise the need for in-person trial sites.
Moreover, the ability to conduct many parts of the clinical trial experience remotely enables organisations to accelerate recruitment of patients from different populations, and even precision matching patients for specific trials (AI-enabled patient stratification helps guarantee the equitable allocation of participants or subgroups.)
Finally, AI powered automation and data collection helps to generally lower cost and reduce the amount of time required to process clinical trial data. For example, ML algorithms can help predict the success of molecules that are used in clinical trials, identify new compounds, and draw novel insights. Biosimulation engines can accelerate drug development insights and de-risk R&D decisions by predicting the clinical benefit of a drug or molecule before it gets to human trial.
VeriSIM, that was named โPredictive Analytics Solution of the Year" in last yearโs BioTech Breakthrough Awards program conducted by BioTech Breakthrough, just partnered with Clarivate (a global leader in connecting people and organisations to intelligence they can trust) to de-risk drug development. The collaboration combines VeriSIM Life's unique approach with Cortellis Drug Discovery Intelligenceโขย by Clarivate, to provide pharmaceutical and biotech companies with R&D insights to de-risk and accelerate drug discovery and avoid late-stage failures during clinical trials.
VeriSIM Life raised a total of $20.77M.
๐ The diva company of AI Drug Discovery Insilico Medicine, a clinical stage generative AI-driven drug discovery company, announced in a study just published that it successfully combined generative AI and quantum computing to accelerate drug discovery in order to explore lead candidate discovery in drug development. In particular, they demonstrated the potential advantages of quantum generative adversarial networks in generative chemistry, by an implicit GAN for small molecular graphs, with a variational quantum circuit (VQC) as the noise generator.
Generative Adversarial Networks (GANs) are one of the most successful generative models in drug discovery and design and have shown remarkable results for generating data that mimics a data distribution in different tasks. The classic GAN model consists of a generator and a discriminator. The generator takes random noises as input and tries to imitate the data distribution, and the discriminator tries to distinguish between the fake and real samples. A GAN is trained until the discriminator cannot distinguish the generated data from the real data.
Building on these findings, Insilico scientists plan to integrate the hybrid quantum GAN model into Chemistry42, the company's proprietary small molecule generation engine, to further accelerate and improve its AI-driven drug discovery and development process.
The company has also announced that the FDA recently approved their initial investigational new drug (IND) application for ISM3091 for the treatment of patients with solid tumours, that is Insilico's first oncology program to advance to the clinical validation stage.
Insilico Medicine has raised a total of $401.3M in funding over 10 rounds.
๐ Xtalpi, an AI-powered drug discovery company based in China, announced this week it has partnered with Eli Lilly in a deal worth up to $250 million.
Xtalpi is offering a closed loop of AI and quantum physics algorithms working in sync with the data factory of large-scale robotics experiments. In particular, XtalPiโs superior AI algorithms (and so far more than 200+) combined with physics-based methods, and XtalPiโs advanced automation robotics can accelerate small molecule discovery, macro-molecular discovery, amplified chemical synthesis and solid state formulation.
XtalPi has raised a total of $786.4M in funding over 10 rounds.
Until next time ๐,
PS: More AI biotech news ๐โ๏ธ from Substack
๐โ๏ธFor more about AI revolution in chemistry:
BioByte 031: the AI for chemistry revolution
By Decoding Bio
๐โ๏ธFor more about UX of Biological R&D:
The Disparity in UI/UX between Biotech & Tech
Why is the UX of Biological R&D so bad?
By Vega Shahย