Jonathan Irish

Interesting four-part series about cell therapy from Fierce Biotech. Definitely worth a full read. Provided the link and a summary for each part of the series below. 𝐏𝐚𝐫𝐭 𝟏: 𝐁𝐢𝐨𝐭𝐞𝐜𝐡 𝐟𝐚𝐜𝐞𝐬 𝐚 𝐫𝐞𝐜𝐤𝐨𝐧𝐢𝐧𝐠: '𝐖𝐞'𝐯𝐞 𝐥𝐨𝐬𝐭 𝐨𝐮𝐫 𝐥𝐮𝐬𝐭𝐞𝐫 𝐢𝐧 𝐜𝐞𝐥𝐥 𝐭𝐡𝐞𝐫𝐚𝐩𝐢𝐞𝐬' 𝐀𝐫𝐭𝐢𝐜𝐥𝐞 𝐥𝐢𝐧𝐤: https://lnkd.in/gGp9RCgr 𝐒𝐮𝐦𝐦𝐚𝐫𝐲 𝐨𝐟 𝐏𝐚𝐫𝐭 𝟏 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞𝐬 𝐢𝐧 𝐂𝐞𝐥𝐥 𝐓𝐡𝐞𝐫𝐚𝐩𝐲: -Decline in interest and performance of autologous CAR-T therapies. -Market challenges and manufacturing issues. -Shift in focus towards autoimmune conditions. 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐲 𝐒𝐡𝐢𝐟𝐭𝐬: -Major pharma companies like Sanofi and Bristol Myers Squibb (BMS) are restructuring and streamlining cell therapy efforts. -BMS closed a research center and signed deals focusing on autoimmune therapies. -Sanofi discontinued Kiadis, a business focused on natural killer cell therapies. 𝐌𝐚𝐫𝐤𝐞𝐭 𝐏𝐞𝐫𝐟𝐨𝐫𝐦𝐚𝐧𝐜𝐞: -Mixed results for approved cell therapies, with Gilead’s Yescarta as a notable success. -Bispecifics are outperforming cell therapies in cancer care. 𝐓𝐞𝐜𝐡𝐧𝐨𝐥𝐨𝐠𝐢𝐜𝐚𝐥 𝐚𝐧𝐝 𝐂𝐥𝐢𝐧𝐢𝐜𝐚𝐥 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞𝐬: -Slow progress in solid tumors and high relapse rates. -Emerging allogeneic (donor cell) therapies show promise but face setbacks. 𝐅𝐮𝐭𝐮𝐫𝐞 𝐃𝐢𝐫𝐞𝐜𝐭𝐢𝐨𝐧𝐬: -Companies like Allogene and Caribou Biosciences are exploring new gene-editing tools and next-gen therapies. -Focus on developing in vivo CAR and allogeneic products. -Continued commitment to advancing cell therapy despite current challenges. 𝐊𝐞𝐲 𝐈𝐧𝐬𝐢𝐠𝐡𝐭𝐬: -Industry needs to justify ongoing investment in cell therapy. -Potential for market acceptance of new technologies. -Ongoing development in gene-edited and allogeneic therapies may bring breakthroughs. Part 2: https://lnkd.in/gcpuXmMV

10

As promised, here's part 2 of a two-part article series by our executive editor Fran DeGrazio. In this installment of her Frankly Fran column, Fran takes a deep dive into particle testing to illustrate her take on drug product quality. #drugdelivery, #combinationproducts, #qualityassurance, #regulatory, #particleanalysis https://lnkd.in/dvkFUWeH

4

Pretty upbeat Illumina earnings call. Two positive things jumped out to me (not an investment analyst 🤪): NovaSeq X average pull-through hit $1M (money printer go brrr!) and margins are up (poor margins are long time analyst complaint). Illumina also announced an entry level NovaSeq X - a cheaper, single flow cell version of the X plus. At first I figured that was just a product strategy to increase the X's addressable market, but then this exchange happened, making me think this is all about fighting off Element Biosciences: Question by Puneet Souda/Leerink: "So just trying to understand the mid-throughput market, is cost per gigabase a lever that you can use to lower cost in the near term to compete more effectively [with Element😳] in that market?" "Or do you think the new NovaSeq X [single flow cell, coming out Q4] with pricing that is actually moving lower will address maybe a potential gap there with a competitor [Element😳] in the mid-throughput segment?" Answer by Jacob: "Yeah, I think both [are strategies we need to fight off Element😳]. With the new [cheaper] NovaSeq X you can run the 25B flow cell and get the full power of the pricing we have for the NovaSeqX plus [$2/Gb, thanks Element😳]. " "But I would also say on the mid-throughput we certainly have an opportunity to use our pricing [i.e. we've told sales to cut prices to whatever to win deals against Element😳] if we need to." "And while I certainly believe there are a lot of great members of the teams of our competitors [Element😳 is taking all our great people], I think sometimes we also get caught up in very too narrow definitions of claims." "I agree we could do a better job in communicating this [i.e. Element😳 is winning the PR battle] and we will spend more time next week also providing more insight on that."

102

8 Comments

Here are several points about AI model validation in the field of drug discovery: https://lnkd.in/gZF_DejR This posting was inspired by Oleg Trott's blog post, https://lnkd.in/gmCaUt6b. Since my writing doesn't directly address the core content of Oleg's article, please consider it as a supplementary discussion related to it. I think these are points that people who know them are very familiar with, and those who don't may never have thought about. I wrote this because I believe there are more people who don't know about these points than those who do.

13

I learned the basics of building RNNs for sequences and CNNs for images using PyTorch, as well as multi-input and multi-output architectures.

1

We’re excited to release gReLU (https://lnkd.in/ggN5Ugxp), a Python package for deep learning on DNA! gReLU covers all major steps in DNA sequence modeling: data processing, training, evaluation, interpretation, variant effect prediction, and ML-guided DNA sequence design. Deep learning models are powerful scientific oracles - we use them to perform in silico genome engineering, prioritize disease-associated genetic variants, interpret regulatory rules, and design new regulatory elements or edit endogenous ones for therapeutic applications. There are lots of great DNA models, design methods and software tools. But the ecosystem is fragmented in terms of deep learning frameworks and how downstream tasks such as interpretation and design are implemented. This makes cross-talk (e.g. mix & match of models) hard. gReLU isn’t the first attempt to provide a unifying framework for sequence modeling - but it has many unique and advanced features, and we hope to add many more. It comes with a curated model zoo (http://wandb.ai/grelu - thanks Weights & Biases) containing trained models that users can apply and fine-tune. It has state-of-the-art multimodal models like Borzoi by David Kelley, models of human snATAC (http://catlas.org) and more. It’s thoroughly documented and tested, and comes with detailed tutorials - follow them and learn how to process genomic data for deep learning, train your own models with customized architectures, and use these models for all kinds of interpretation and design tasks. Thank you to all the users and contributors! That includes our ReLU team Gökçen Eraslan Laura Gunsalus Surag Nair, collaborators David Garfield Oriol Fornés Crespo, all Biology Research | AI Development (BRAID) department members at gRED/Genentech. Also Jacob Schreiber, whose software we used extensively. Preprint soon!

573

20 Comments

I've read a few papers over the last couple of years that try to predict RNA-seq from H&E whole slide images. This article by Francisco Carrillo-Perez et al. does the reverse; they generate synthetic whole slide images from RNA-seq data. They used a diffusion model and demonstrated the results on five different cancer types. To validate the generated images, they used three different techniques. In the first, they applied a cell segmentation and classification model called HoverNet and compared the proportion of different cell types detected with that of real images for each type of cancer. They also used the synthetic images to train a tissue classification model. By manipulating the fraction of synthetic vs real images in the training set, they showed that including synthetic images did not hurt model performance. Further, by adding synthetic images to the real ones, they created a larger dataset for model pretaining with self-supervised learning; this improved model performance further. In an additional experiment, they used synthetic images to pretrain a model for colorectal cancer. They fine-tuned this model for classifying microsatellite instability status, outperforming a model trained from scratch. Generative AI is becoming increasingly powerful, but always needs to be validated when used in a medical application. Validating synthetic images is not an easy task, so it's great to see the creative approaches used here. https://lnkd.in/eir8GNQm To learn more about the latest research, subscribe to my Computer Vision Insights newsletter: https://lnkd.in/g9bSuQDP #Pathology #CancerResearch #PrecisionMedicine #MedicalImaging #MachineLearning #DeepLearning #ComputerVision

84

5 Comments

The FY2024 House National Defense Authorization Act (NDAA) is a bill of significant importance, and its implications for WuXi and the US biotech landscape are worth noting. According to CEO Rick Connell, "We're reading the context of the bill, and it's all about human genome sequencing data. We genuinely did not understand why we were inserted into the bill." It appears that Congress knows more about WuXi than their CEO! Let's delve into this together and read this article to learn more about the bill's impact on WuXi and the implications for the US biotechnology landscape. #FY2024 #NationalDefenseAuthorizationAct #WuXi #biotech #genomesequencing

2

Very insightful take on data analysis! I see both sides of this opinion. Flexibility in methods of analysis is OK for exploratory preclinical research but after a certain while you need a standardized modus operandi to make go-no go decisions. I would like to take the example of my favorite technique, Flow cytometry. Despite being one of the best and most versatile techniques it is not utilized to its full potential in a lot of clinical studies due to lack of standardized analysis techniques and lack of consensus on phenotypic markers of certain immune cells.

4

Those interested can check out this most recent addition of BioWire Weekly where we discuss cutting-edge therapies targeting chronic conditions like osteoarthritis and obesity, drug delivery systems for neurological disorders, and an interesting anti-aging study in dogs. https://lnkd.in/gGh9ZAy2

2

Biotech weekly news roundup! All of the weeks biggest topics in one place: Illumina enters into $750M agreement with JP Morgan to fund Grail Divestment! https://lnkd.in/gFwA-2Dp Ultragenyx gets the green light to seek FDA approval of Gene Therapy! https://lnkd.in/gTTwKezi Follow / connect with me to keep up to date with all the latest developments in the space! #biotech #bioinformatics #news

10

ISoP Community. As the co-chair of this years communications committee, alongside Eleni Karatza, we have some pretty big updates so please read through this post. First and foremost, ACoP 2024 registration is open. Please see the following link (also found below): https://lnkd.in/eNZ2PAcq Note hotel reservations are open and the cut off date for group rate is October 6th. However, as always, if you plan to attend I would suggest booking sooner than later. You can reserve a room by calling 877-800-4888 or through the Reservation site here: https://lnkd.in/ea95yZQS If you haven't checked out the new ISoP website, it looks great! There is a brand new members only section with various benefits. There is a main home feed which contains content/posts from various "communities". Honestly, it feels/looks like an internal Facebook exclusive for our society (I don't mean this in a negative way). You can update your profile, connect with others, and keep up to date with ISoP events/communications. Here is the link the updated website and members section (you must have an active membership): https://lnkd.in/eDVHwgSZ Right now there is only one main community called "Members Connect Community", which contains all ISoP members and acts as a way to make a broad post to everyone. In the future, we could create ISoP SIG specific groups, alumni groups, local ISoP chapters, and others. Even individual working groups within ISoP (e.g., QSP SIG WGs) could create their own channel. Here is the link to the Member Connect Community channel: https://lnkd.in/eBZWrscu If there are any questions feel free to reach out to me directly or comment below! Happy Father's Day!

36

Team of Physics, Math and CS engineers are solving complex molecular biology problems in AI. Almost no biologists 👏 Chai Discovery team is excited to release Chai-1, a new multi-modal foundation model for molecular structure prediction that performs at the state-of-the-art across a variety of tasks relevant to drug discovery. Chai-1 enables unified prediction of proteins, small molecules, DNA, RNA, covalent modifications, and more. Chai-1 model achieves a 77% success rate on the PoseBusters benchmark (vs. 76% by AlphaFold3), as well as an Cα LDDT of 0.849 on the CASP15 protein monomer structure prediction set (vs. 0.801 by ESM3-98B). #Math #AI #Biology https://lnkd.in/gTJEgVBY

4

Excellent short essay by Nathan Frey and Eric Dai on the value of AI/ML in the #drug #development process emphasizing the decision making process as the one that offers greates value. I prefer to use the term "drug development" as I believe "drug discovery" is selling the opportunity of AI/ML short. Discover is imho not the bottleneck, deciding on which #hits to mature to #leads and eventually clinical candidates is where true value is generated - by reducing #uncertainty in the decision making process. At iuvantium we offer exactly this to our partners: #certainty for immunobiology. With our technology platform we can predict quality and quantity of an immunological response, have the tool box to program immunobiology, and eventually build a digital twin of the immune system to move toward in silico clinical trials and personlized medicine.

17

THIS WEEK -----> Multi-Omics News! 👇🧬🧬 SeqSeries: 💰 PrognomiQ Inc closed a $34M Series D to advance their blood based lung cancer test... first as an LDT, then as an IVD. 🧬 Dovetail Genomics received a $2M SBIR Grant from NHGRI to further develop and validate its LinkPrep tech for oncology, focusing on automation and high-throughput analysis. --------------------------------- SeqSecrets: ❤️ CS Genetics introduced SimpleCell, a streamlined approach in single-cell genomics! 💙 GenomOncology partnered with Genomenon, Inc to improve germline variant classification. 💚 The UK will launch a 'world-first' genomics-based pandemic warning system! 💛 Ionpath launched MIBIscope 3.0 for high-definition spatial proteomics imaging. 💜 PacBio introduced a new benchtop sequencer to make HiFi long-read sequencing accessible beyond large genomics labs. 🖤 Caris Life Sciences’ MI Cancer Seek test gained FDA approval as a companion diagnostic for multiple therapies. 💖 EVOBYTE Digital Biology launched ReadStore, a platform for managing and integrating genomic data. 🧡 Microbiologics acquired German quality control company SensID GmbH. 💗 Inocras Inc. signed an MOU with Tokyo's Summit Pharmaceuticals to incorporate WGS into Japan’s healthcare. 💓 Oxford Nanopore Technologies is collabing with Genomics England, UK Biobank, and NHS England on genomic and epigenomic research for cancer, genetic disorders, and infectious diseases... with possible expansion to newborn screening and pharmacogenomics! 💕 QIAGEN's QiaStat-Dx panel received FDA clearance for detecting meningitis and encephalitis pathogens. 💔 MOBILion Systems, Inc. is applying its SLIM ion mobility tech to proteomics! --------------------------------- SeqShift: 👨‍💼 Ribbon Bio appointed John Luckey as CTO. 💼 Biocartis appointed Randy Polonsky as CFO. 👔 Volition, an epigenetics company, appointed Timothy I. Still as chairman of the board. 🌟 SOPHiA GENETICS promoted CFO and COO Ross Muken to president, and appointed George Cardoza as CFO! --------------------------------- Follow me for multi-omics news like this every week! #MoxeeMinute

88

5 Comments

This week’s recap highlights a new method for gene-level alignment of single-cell trajectories, an R package for integrating gene and protein identifiers across biological sequence databases, characterization of SVs across humans and apes, universal prediction of cellular phenotypes, a method to quantify cell state heritability versus plasticity and infer cell state transition with single cell data, and a new AI-driven, natural language-oriented bioinformatics pipeline assists with automatic and codeless execution of biological analyses. Others that caught my attention include pangenome-informed privacy preserving synthetic sequence generation, a paper showing generative haplotype prediction outperforms statistical methods for small variant detection, a metadata standardizer for genomic region attributes, a web-based platform for reference-based analysis of single-cell datasets, a new taxa-specific normalization approach for microbiome data, a deep-learning-based splice site predictor, models for human metabologenomics, a Snakemake pipeline to automatically generate pangenomes from metagenome assembled genomes and a new paper on COVID-19 origins.

20

Very important discussion Parashar Dhapola! We need tools to classify cells and we need them now, but we must live with uncertainty, because there is an underlying problem: we lack a definition for a Cell Type. As we discussed in our editorial on single cell profiling in MAM. Cells respond to changes in environment and some changes may be transient, while other may be more permanent. Some transient changes may require the environment conditions to be maintained, while other respond to a stimuli. Some changes may require cell division. Also, turnover of biomarkers (mRNA, microRNA, proteins, metabolites, methylation, chromatin modifications etc.) is very different. which biomarker is the determinant? We have fundamental questions to answer, before we can set the perfect bioinformatics criteria. But until then, we do our best :-) Thanks for initiating this! https://lnkd.in/dsNf9TrU

10

Designing powerful antibodies is tricky! New model #AntiFold tackles this by predicting sequences that fold into desired antibody structures while maintaining antigen binding. This could revolutionize #antibody design and development! 🎯 By synergizing inverse folding capabilities with protein language models, AntiFold represents a significant advancement in computational antibody design for therapeutic and diagnostic applications. Quick Read: https://lnkd.in/gV6agQzp #bioinformatics #immunotherapy #drugdiscovery #ai #sciencenews

21

The Fred Hutch Innovation Lab has conducted several benchmarking studies on single-cell technologies. Their most recent study evaluates the performance of Pipseq V4Plus T20 kit by Fluent Biosciences, which was recently acquired by Illumina. “PIPseq’s T-20 attractive feature is a low kit cost at $3600 and required one-time equipment purchase of $2500. Cost at $0.06 per cell, excluding sequencing, is competitive but reduced recovery, sequencing inefficiency and workflow demands raised this cost in our experience.” The study's conclusions align with the findings of the recently published single-cell benchmarking study I mentioned in an earlier post. Both studies demonstrate that read utilization and resulting sequencing cost should be included in any single cell technology evaluation. https://lnkd.in/eq_ncHiy

38

2 Comments