
CAR T cell based treatments have shown promise for many blood cancers, but have struggled against solid tumors. @deniswirtz and co. demonstrated a way to increase the tumor infiltration capabilities of CAR T cells using synthetic velocity receptors.

📄 https://doi.org/10.1101/2023.12.13.571595

These CAR TV (velociraptor) cells were tested against lung, ovarian, and pancreatic solid tumor models.

🔹 By varying velocity receptors, the researchers found a CAR TV type that infiltrated and inhibited each model (see pancreatic cancer model here).

Clever cells...

Not only do they show that the tumor volumes are markedly decreased, but also that survival is increased dramatically. In these charts you see 100% of mice surviving for the treatment length vs 0% controls. 🐭 💪

Other researchers are working in this domain as well – to extend CAR T to solid tumors by e.g., engineering them to secrete e.g., IL-10 to prevent exhaustion. This approach showed effectiveness against multiple tumor challenges, hinting at durability.

https://twitter.com/Tang_Lab_EPFL/status/1742919774647325103?s=20

We're not done yet! Researchers are also exploring the use of CAR T cells for treatment of autoimmune and rheumatic diseases.   https://x.com/MaxKonigMD/status/1744398691199222035?s=20

Let's give these researchers a hand, and a follow as a kudos for the work they are doing that is so important against these common scourges. We can't wait to see these various CAR T cells in trials and saving patients. @JohnsHopkins @EPFL_en

Also, don't forget to check out Denis' databases of postdoctoral fellowships, Ph.D. funding, and  funding opportunities for early career researchers and faculty.

🔸Early Career: https://twitter.com/deniswirtz/status/1743743214098739514?s=46&t=VBaTbz9G2S8Magx9V8T52A
🔹 Ph.D. Fellow: https://x.com/deniswirtz/status/1744355909747286353?s=20
🔸Postdoc: https://x.com/deniswirtz/status/1744000600738877865?s=20


Enhancing CAR T Cell Infiltration for Solid Tumors

We started #MaterialsDataMonth in Nov. with a goal of increasing visibility of materials data, #ML, #AI, and #OpenScience efforts and shining a light on some teams that are doing amazing work. We wanted to share with you our engagement stats for the month. Some highlights! 🧵

We showed you how to create a Google Scholar profile to track your multi-PI team's progress. https://twitter.com/BenBlaiszik/status/1589667758924386304?s=20&t=s43L5VteKuZrgZq_JjWXaw

We highlighted some sessions at #F22MRS @Materials_MRS https://twitter.com/BenBlaiszik/status/1596681798427316224?s=20&t=s43L5VteKuZrgZq_JjWXaw

We highlighted a new presentation and work by @RGBLabMIT @TatemRios https://twitter.com/BenBlaiszik/status/1588555487917662208?s=20&t=s43L5VteKuZrgZq_JjWXaw

We showed an example of materials discovery in action, with Gorilla Glass which is commonly used as screens in various electronic devices. https://twitter.com/BenBlaiszik/status/1597702342186782720?s=20&t=s43L5VteKuZrgZq_JjWXaw

We helped boost visibility of and add new content to the "Awesome List" of materials informatics resources.
https://twitter.com/BenBlaiszik/status/1588236869677961217?s=20&t=s43L5VteKuZrgZq_JjWXaw


We reminded folks of the great @MaterialismPod by @taylordsparks and co. 🎙️ https://twitter.com/BenBlaiszik/status/1587844501279674368?s=20&t=s43L5VteKuZrgZq_JjWXaw

We highlighted a new #FAIR foam database produced by @NIST colleagues @madgameC45 https://twitter.com/BenBlaiszik/status/1590359431945912320?s=20&t=s43L5VteKuZrgZq_JjWXaw

We introduced Foundry-ML a Python SDK to provide immediate access to structured ML-ready datasets. https://twitter.com/BenBlaiszik/status/1592924859054497792?s=20&t=s43L5VteKuZrgZq_JjWXaw

We made a banner for people to use on their profile. Thank you @CranfordMATTER! https://twitter.com/CranfordMATTER/status/1587816647489835008?s=20&t=s43L5VteKuZrgZq_JjWXaw

We highlighted just a small fraction of the excellent work being done in the @NIST Center for Hierarchical Materials and Design Center. https://twitter.com/BenBlaiszik/status/1587545321789235200?s=20&t=s43L5VteKuZrgZq_JjWXaw

and we shared a story of the research journey of several scientists from skeptics to believers in the power of data and models. https://twitter.com/BenBlaiszik/status/1593660746532913152?s=20&t=s43L5VteKuZrgZq_JjWXaw

Next year, we plan to have more content specifically focused around connecting people in the community, and we will be looking for partners to help get the word out. Until then, we will keep posting about the latest advances in materials science. ✨

You can read the unrolled version of this thread here: https://typefully.com/BenBlaiszik/yQhSrdc

🎙️ Stepping up to the podium next is @RGBLabMIT! Excellent talk on some of their latest work.

👪 Multi-fidelity predictions
💊 Drug discovery for COVID and Alzheimers treatments
📈 Graphs for macromolecules

#MaterialsDataMonth #AI #opendata #openscience
https://m.youtube.com/watch?v=oq6J26lZR-k&list=PLlMMtlgw6qNh-MH4-dsl0griJ-Bh1dtWJ&index=12

The 3D molecular geometry dataset (GEOM) is available via @DataFacility.

https://twitter.com/BenBlaiszik/status/1557098597677776904?s=20&t=wyUbeexTvs3WJvZCHrI7AQ

Do you find yourself wanting to learn about materials during your commute. Check out the Materialism podcast. Some topics include:
✨Materials Informatics @generalelectric
✨Recycling
✨Autonomic self-healing materials
Follow @MaterialismPod #MaterialsDataMonth @taylordsparks

Autonomic materials is a topic of great interest to reducing waste, increasing material lifetime, and generally imbuing materials with capabilities usually reserved for biological materials (e.g., self-healing/cooling). Read more about their history here!
https://twitter.com/BenBlaiszik/status/1530268300193812480?s=20&t=9U_vLM61qRTdolVvyb_MYw


Also from @taylordsparks and @SterlingBaird1 on a minimum viable self-driving lab (MV-SDL) demo. More on a number of SDL efforts in a future thread. https://twitter.com/SterlingBaird1/status/1583963490376556544?s=20&t=f6n6HU_wVC4O0TVTjB_4ig

We start off the month with a quick profile of the Center for Hierarchical Materials and Design (CHiMaD). CHiMaD is a @NIST-funded flagship Materials Genome Initiative (MGI) research center, strongly embracing MGI AI/ML/data approaches #MaterialsDataMonth

CHiMaD research leverages physical theory, advanced computational methods, AI/ML and materials databases to accelerate creation of novel materials with properties for particular applications. This is in contrast to the traditional trial-and-error method of materials discovery

CHiMaD is arranged into use cases in electronic inks, thermoelectrics, polyelectrolyte complexes, self-assembly, impact mitigation materials, composites, additive manufacturing, cobalt superalloys, phase field methods, uncertainty quantification, #AI/#ML, and materials databases.

An examplar of an MGI-inspired approach broadly seen across CHiMaD, Wei Chen of @NorthwesternU developed a data-centric design framework for complex systems. Here, applied to property prediction and optimization of polymer nanocomposites. https://doi.org/10.1016/j.eng.2021.05.022

CHiMaD is the center that has steadfastly supported our team's  efforts in materials informatics, and led to the creation of @DataFacility. We will hear more about MDF software, services, and impact this month, but learn about MDF capabilities now at https://www.materialsdatafacility.org

https://twitter.com/BenBlaiszik/status/1585996664526905344?s=20&t=UJEd03_K3js8lPSgWsJ-9w

The #CHiMaD annual conference starts next Monday (Nov 7), so we will feature even more great content from them soon!

Until then, you can find more information here: https://chimad.northwestern.edu

Call to Action: What centers do you find embracing the new "Jedi" power of AI/ML/data to speed their research and make their teams more effective? Quote tweet this thread and let the community know!

Follow me and @kjschmidt for more #MaterialsDataMonth content.

You can read the unrolled version of this thread here: https://typefully.com/BenBlaiszik/buI83nw

The latest in #AI4Science:
✨Diffusion based antibody design
✨New NN architecture for tabular data modeling
✨Data driven stable crystalline hydrate prediction
✨@QMOF_Database
✨A new "Informatics Help Desk"

and more #AI/#ML/data for science!🤖🧵

Diffusion is definitely a hot topic these days. @luost26 release a diffusion-based model for antibody design described as a “Swiss Army Knife capable of sequence-structure co-design, sequence design for given backbone structures, and antibody optimization". #NeurIPS

https://twitter.com/luost26/status/1582419287360172036?s=21&t=BI-YFzTFIBuw4JL5hdAYWw

Ahmad Sheikh an co. publish a method to predict stable crystalline hydrate materials. Their models enable prediction of hydrated crystal structures starting from an anhydrous framework without requiring specification of the water content. @PNASNews
https://www.linkedin.com/posts/ahmad-sheikh-30bb511_a-data-driven-and-topological-mapping-approach-activity-6989078550628757504-bhbf


@Andrew_S_Rosen and co.s @QMOF_Database is an excellent resource for for ML-guided discovery of MOFs. A new paper describing how it can be leveraged to speed MOF discovery is here.

https://twitter.com/andrew_s_rosen/status/1583303656350236673?s=21&t=oEwVROPyjaXsmKmBUeZZzw

@hspark1212, @SmitBerend and others combine the latest in transformers with MOF property prediction resulting in models that capture global and local influences better, and generalize more effectively across predicted MOF properties (e.g., H2 uptake, H2 diffusivity, band gap).

https://twitter.com/hspark1212/status/1584518986448134144?s=20&t=x0U8nc9wYGTlKjJ8-_phAg

Dane Morgan and co at @Wisconsin started a science and engineering informatics help desk. Come with questions or problems (e.g., model related, domain specific, or even software challenges), and get feedback from expert grad students and postdocs. @NSF
https://skunkworks.engr.wisc.edu/science-and-engineering-informatics-help-desk/

@FrankRHutter described TabPFN a new NN-based tabular data classification method that takes 1 second & yields SOTA. @rasbt has a nice thread and discussion on the strengths and weaknesses of this new technique.

https://twitter.com/rasbt/status/1584210875455791104?s=21&t=B-N5GaoI14outquuR5lIXQ

@shaneguML, Jiawei Han, and others describe a process that enables improvement of LLM performance on benchmark tasks through self training (e.g., Chain-of-Thought prompting and self-consistency), achieving SOTA on a number of tasks. @Google @UofIllinois

https://twitter.com/shaneguML/status/1584668991372464129?s=20&t=cna9P1nnHU26BkFwudHVgQ

Work from our team on segmentation of XCT data using UNet models.


https://twitter.com/BenBlaiszik/status/1583115892593868800?s=20&t=b_ta1tpK-e-dP94tlBhhbA

If you found this thread helpful:

1.  Follow me @BenBlaiszik  for more threads like this each week.

2.  Bounce back to the top of the thread and retweet it so you can find it later & others can find it too!

3.  Interested in AI/ML for science? Follow our group at
@LabsGlobus.


You can read the unrolled version of this thread here: https://typefully.com/BenBlaiszik/KiiM62p
