BioByte 059: self-amplifying RNA vaccine, off-target protein translation from mRNA vaccines, RAG agents for scientific research, immune cell type-cytokine atlas
Welcome to Decoding Bio, a writing collective focused on the latest scientific advancements, news, and people building at the intersection of tech x bio. If you’d like to connect or collaborate, please shoot us a note here or chat with us on Twitter: @zahrakhwaja @ameekapadia @ketanyerneni @morgancheatham @pablolubroth @patricksmalone. Happy decoding!
Some quick housekeeping: we’ll be taking the last couple of weeks of the year off from the newsletter, and will be back in the new year. Thanks for an incredible year of Decoding! Lastly, the Decoding team will be in SF for JPM in January. Reach out if you’ll be there!
What we read
Blogs
Self-copying RNA vaccine wins first full approval: what’s next? [Dolgin, Nature News, December 2023]
ARCT-154 developed by Arcturus Therapeutics and CSL is the first self-amplifying RNA (saRNA) vaccine that has been granted full regulatory approval anywhere in the world. saRNA can make copies of itself inside cells, and because it can be used at a lower dose (up to 10% of the dose compared to other RNA-based vaccines) it might have fewer side effects than mRNA vaccines and lower production costs.
ARCT-154 encodes for the spike protein expressed by SARS-CoV-2, and uses the replication machinery from a mosquito-borne virus known as Venezuelan equine encephalitis virus. The vaccine has been approved in Japan and the developers are now seeking authorization in Europe.
Some Thoughts on Biotech vs Pharma for Computational Chemists [Pat Walters, Practice Cheminformatics, December 2023]
In the dynamic landscape of drug discovery, a recent editorial in J Med Chem by Dean Brown and subsequent discussions have highlighted a significant shift in training computational chemists and cheminformaticians. Unlike the past, where most professionals in this field learned through hands-on experience within biotech and pharmaceutical companies, the current scenario presents a more complex picture. Today's computational chemists, often with academic backgrounds in fields like quantum chemistry or machine learning, enter the industry without direct experience in drug design. This lack of practical exposure necessitates a strong emphasis on mentorship and on-the-job learning, a challenge that is particularly pronounced in smaller biotech firms with limited mentoring resources.
The contrast between large pharma and smaller biotech companies is striking. In larger firms, computational chemists tend to become specialists in specific niches, such as molecular dynamics or AI, which while beneficial for becoming domain experts, can limit their exposure to the broader aspects of drug discovery. In contrast, smaller companies offer a wider range of experiences and skill development due to the varied roles these chemists play, driven by necessity and the organization's size. Moreover, the increasing reliance on both commercial and open-source tools, along with the shift to remote work, introduces new challenges in training and professional development. The industry is thus at a crossroads, needing to balance the traditional mentorship model with the demands of a technologically advancing and structurally diverse professional landscape, to ensure the effective training and development of the next generation of experts in drug discovery.
Seattle Hub for Synthetic Biology launched by Allen Institute, Chan Zuckerberg Initiative, and the University of Washington will turn cells into recording devices to unlock secrets of disease [Allen Institute, December 2023]
The Allen Institute, Chan Zuckerberg Initiative (CZI), and the University of Washington (UW) launched the Seattle Hub for Synthetic Biology, a pioneering research collaboration aimed at developing technologies to record cellular history over time. This initiative will allow scientists to understand the dynamic changes in cells and genes, going beyond end point measurements in health and disease.
The project, led by UW Medicine researchers Dr. Jay Shendure, Marion Pepper, Cole Trapnell, and Jesse Gray, will advance the technology initially developed at the Allen Discovery Center for Cell Lineage Tracing and UW Medicine’s Brotman Baty Institute. The goal is to transform living cells and genomes into devices for chronicling complex biological information over time. This approach could revolutionize our understanding of the cellular and genetic roles in human health, offering new insights into the progression from genetic mutations to clinical diseases. The research tool developed will initially focus on developmental and immunological cell changes, with potential for broader applications in research, diagnostics, and clinical fields. The Allen Institute and CZI are committed to open science, ensuring that the findings are shared globally to advance research in various labs.
Academic papers
PaperQA: Retrieval-Augmented Generative Agent for Scientific Research [Lála et al., December 2023]
Why it matters: The rate of papers published every year grows exponentially, leading to important discoveries going unnoticed. LLMs have been used to answer scientific questions, but off-the-self, pre-trained models can hallucinate plausible-sounding responses which can cause more harm than good, and can’t be used to retrieve information past the training cut-off date.
In this first paper from Future House, the team presents PaperQA, a Retrieval-Augmented Generation (RAG)-based agent that can answer questions over scientific literature. RAG models retrieve text using vector embedding or keyword search and add the retrieved passage to the context window of the LLM; thus reducing hallucinations and improving performance. PaperQA has three components: “finding papers relevant to the given question, gathering text from those papers, and generating an answer with references.”
In the team’s evaluations, PaperQA performed better than GPT-4, Perplexity and other models (see below). PaperQA was used as a basis for WikiCrow, an automated system to synthesize Wikipedia-style summaries for technical topics from literature. As a demo, WikiCrow generated draft articles for 15,616 human protein-coding genes that lack Wikipedia articles; each article took 8 minutes to create (instead of ~4h) and made incorrect inferences or statements 9% of the time. We couldn’t be more excited about their vision of the ‘super-pedia’ where articles are generated about any topic in real time.
Dictionary of immune responses to cytokines at single-cell resolution [Cui et al., Nature, December 2023]
Why it matters: The immune system is underpins the body’s response to all types of disease. A more granular understanding of its complexity can enablement the development of novel therapeutics to ameliorate disease progression. The scale of the study described is the first of its kind and offered several interesting conclusions that can pave the way for future therapeutic research.
A new, large-scale single-cell resolution map of each major immune cell type responding to each major cytokine in vivo. The reference displays more than 1,400 cytokine-cell type combinations. The team also developed software for scientists, called Immune Response Enrichment Analysis (IREA), that they can use to identify the most active cytokines involved in a disease or drug response, and how different immune cells carry out different functions depending on which cytokine signal they receive. The authors found that both the immune responses to cytokines and the plasticity of immune cells were much more complex than we previously appreciated. Even ‘well-studied’ cytokines such as IL-1β have been discovered to induce a distinct responses in each cell type, which shows how a single cytokine can trigger a coordinated, multicellular immune response. Regarding plasticity, the study revealed that several other immune cell types have polarisation potential that was not known before.
Bridging the gap between chemical reaction pretraining and conditional molecule generation with a unified model [Qiang et al., Nature Machine Intelligence, December 2023]
Why it matters: Uni-RXN's ability to accurately simulate and predict chemical reactions can streamline the development of new pharmaceuticals and materials. By efficiently generating valid, diverse, and synthetically accessible molecules, it can reduce the time and cost associated with drug development, thereby accelerating scientific progress in medicinal chemistry and related fields. This model's success in handling complex chemical reaction data marks a substantial advancement in applying AI to practical and challenging problems in chemistry.
The paper discusses Uni-RXN, a comprehensive deep-learning model designed to address the complexities of chemical reactions in computational chemistry. Uni-RXN stands out by integrating two critical methodologies: self-supervised learning for effective data representation and conditional generative modeling for precise molecule generation. This integration allows for a more accurate and efficient simulation of chemical reactions, which is crucial in tasks like classifying reactions, retrieving specific reaction types, and generating molecular structures. The model shows remarkable performance in producing valid, diverse, and synthetically feasible drug-like structures, setting a new standard for molecule generation models.
This development holds immense significance for the scientific community, particularly in drug discovery and organic chemistry. By streamlining the prediction and simulation of chemical reactions, Uni-RXN can potentially expedite the development of new pharmaceuticals, reducing both the time and resources typically required.
N1-methylpseudouridylation of mRNA causes +1 ribosomal frameshifting [Mulroney et al., Nature, December 2023]
Why it matters: Anne Willis’ lab discovers the translation of unintended, off-target proteins in mRNA vaccines caused by the incorporation of modified ribonucleotides intended to decrease immunogenicity of mRNA.
mRNA vaccines typically contain modified ribonucleotides in order to decrease immunogenicity and increase mRNA stability. For example, the SARS-CoV-2 mRNA vaccines contain N1-methylpseudouridine to decrease undesirable immunogenicity. In a recent Nature paper, Anne Willis’ group at the University of Cambridge discovered that incorporation of these engineered nucleotides result in a +1 ribosomal frameshifting caused by ribosomal “slips” when it reads the modified base, resulting in the translation of unintended off-target proteins. These proteins elicited off-target cellular immune responses following mRNA vaccination. Since N1-methylpseudouridine is used in the COVID vaccines, the group analyzed humans that received the Pfizer vaccine and found an immune response to the frame-shifted peptides in some individuals. Importantly, there were no side effects noted as a result of these proteins.
A novel antidiuretic hormone governs tumour-induced renal dysfunction [Xu et al., Nature, December 2023]
Why it matters: Patients with cancer often have concomitant kidney dysfunction, which is often attributed to sequelae from chemotherapy (which can be eliminated by the kidneys), or by physical compression of the urinary tract. Kidney disease can often lead to significant worsening of quality of life, along with worsened outcomes in these patients. In this paper, Xu et al., identify a novel antidiuretic hormone that is secreted by certain tumor cells, that leads to renal tubular dysfunction, shedding light on a previously unknown biochemical pathway that leads to cancer-associated renal dysfunction.
The authors induced tumors in the adult fly gut by overexpressing the transcription factor Yorkie-Yap/Taz in intestinal stem cells, inducing cachexia (a wasting syndrome that causes severe weight loss) and bloating (as a function of accumulated abdominal fluid). They found that an an antidiuretic hormone (reduces urine), ITP, was able to potently induce tumor-associated bloating while having no affects on tumor growth or wasting. They found that an isoform of ITP, ITPf, was produced by stem-like cells in the gut that were distant from the tumor; ITPf was found to activate a cGMP pathway in stellate cells of the renal tubules, and that TkR99D (a GPCR, functionally equivalent to mammalian NK3R receptor) is a membrane-associated receptor for ITPf. Interestingly enough, TkR99D was required for bloating to occur in tumor bearing animals, although seemingly expendable in those without cancer.
They then injected NK3R inhibitors in mice bearing several cancers (colon, lung, etc.) and found a significant improvement in renal dysfunction in all apart from liver tumors. Of note, neurokinin B – the cognate ligand of NK3R – did not improve renal function, nor did it influence water imbalance in controls. This suggests that an unknown ligand for NK3R is responsible for tumor-associated kidney failure in mice, and that it may be a potential target for therapeutic intervention.
Notable Deals
Tome Biosciences: $213M for the fourth generation of gene-editing
Overview: Tome was founded based on 4th generation gene-editing science from Omar Abudayyeh and Jonathan Gootenberg with an all-star executive team. Tome’s has received funding from investors such as a16z, ARCH, GV, Longwood, Polaris and has been in stealth for two years.
Technology: PASTE — programmable addition via site-specific targeting elements - the ability to insert any DNA sequence of any size precisely into any location without making double-stranded breaks.
Why it matters: PASTE enables the insertion of large segments of DNA into the genome in parallel, potentially overcoming engineering challenges associated with earlier generation gene editing technologies. PASTE is also agnostic to the mutation in a given diseased gene, meaning that one medicine can be developed per gene rather than per mutation (which may be different in each patient with a given genetic disorder).
Infarm: From Billions to Bankrupt
Sifted recently covered a story that highlights a vertical farming company’s fall from a billion dollar valuation to bankruptcy in a few years. Infarm was a Berlin-based hydroponics company that rapidly grew to 1000 employees across 10 countries. The article details a classic deep tech start-up story where $500M was raised and multiple industry-partnerships were executed on an ‘idea’ or ‘promise’ that was not yet proven out by R&D. In this case, Infarm promised the development of vertical farmed strawberries to the supermarket giant Edeka. Alongside, it seems the company was burning money on utilities such as electricity and headcount without careful planning.
AbbVie bets on Psychiatry with $8.7B Cerevel buy-out
The acquisition (which came at a 22% premium) is further sign of renewed interest from large pharmaceutical companies in neuroscience (e.g. Pfizer / Biohaven).
This may be a risky acquisition for AbbVie: FTC scrutiny may be a risk considering AbbVie’s presence in neuroscience and the potential overlap with its bipolar drug Vraylar. Further, Cerevel’s two lead assets are yet to read out in their ph2 and ph3 trials, and so clear revenue potential is yet to be confirmed.
However, Cerevel and Immunogen both give AbbVie steady revenue growth towards the end of the decade, a strategy much anticipated by its investors given the Humira patent cliff.
Freya Biosciences raises $38M for Women’s Health
Another neglected area by biopharma is Women’s Health. Freya Bioscience is using microbial immunotherapies to shift inflammation that arises in reproductive tract. Key indications for Freya are infertility, endometriosis and preterm birth. Gut microbiome therapeutics have seen recent successes with FDA approvals for Seres Therapeutics and Ferring Pharma, both against clostridium difficile infections.
More:
Cambium closes $19M Series A to develop advanced biomaterials for next-gen hardware
After sickle-cell screwups, Bluebird Bio’s only option might be a company sale
What we listened to
What we liked on socials channels
Field Trip
Did we miss anything? Would you like to contribute to Decoding Bio by writing a guest post? Drop us a note here or chat with us on Twitter: @ameekapadia @ketanyerneni @morgancheatham @pablolubroth @patricksmalone