Brought to you by: Luis Armando Gonzalez-Ortiz, @lissnoriega.bsky.social, Filiberto Ortiz-Chi, Gabriela Vidales-Ayala, Emmanuel Soberanis-Cáceres, Amilcar Meneses, @aspuru.bsky.social and Gabriel Merino
[6/6]

We also release SmilX, an open-source Python tool that produces grammar-consistent molecular strings for researchers and practitioners, along with a web interface.

Code: github.com/LuisOrz/SmilX
Web interface: smilx-isogenerator.streamlit.app
[5/6]

The benefits of TokenSMILES are:
➡️ Reduces representational duplication, ensuring that every generated string follows clear syntactic rules.
➡️ Supports more reliable molecular editing, enumeration, and downstream AI tasks, improving consistency across datasets and models.
[4/6]

Our team created TokenSMILES, introducing a grammar-based view of molecular strings, treating them not as loose sequences of symbols, but as structured “sentences” governed by syntactic rules.
[3/6]

In AI-driven molecular design, one recurring challenge is that many valid SMILES strings exist for a single molecule. This redundancy complicates exploring chemical databases, training models, and traversing chemical space.
[2/6]

This project was the joint efforts of many talented individuals: @chertianser.bsky.social, Han Hao (co-first), Sergio Pablo-García, @valencekjell.com, Shangyu Li, @robpollice.mstdn.science.ap.brid.gy, and @aspuru.bsky.social
[5/5]

Using density functional theory, we show that bulky ligands preferentially transmetalate onto palladium to form unstable intermediates that undergo facile protodeboronation, while compact ligands form stable palladium-boron complexes via a sterically-demanding process.
[4/5]

We demonstrate that the ligand choice is the dominant factor for palladium-catalyzed protodeboronation using high-throughput experimentation, where the role of base and substrate choices, alongside Pd pre-catalysts and ligand loadings were also investigated.
[3/5]

While bulky ligands are designed to enable difficult cross-couplings by promoting the main Suzuki-Miyaura catalytic cycle, we reveal that paradoxically the same structural motifs can impede cross-coupling product formation.
[2/5]

I stand by my statement. All sandwiches are tacos.

Our work in AI-enabled chemistry, self-driving labs, and quantum algorithms reflects a belief that the future of materials and molecular design will be driven by the convergence of AI, automation, and quantum computing.

The article captures many of the challenges and opportunities ahead.

Nice joke but not for long :)

Wait for the release of #elagente Pre-signup at elagente.ca Like a #Toronto subway line we are building and testing the scalable infrastructure but it is coming closer and closer every day.

@variniabernales.bsky.social @thematterlab.bsky.social #chemsky #compchemsky

EGMOF demonstrates how hybrid architectures can bridge data scarcity and structural complexity, marking a step toward universal, data-efficient AI for materials discovery.
[6/6]

This two-step, interpretable design allows EGMOF to:

- Achieve 95% validity and 84% hit rate for hydrogen uptake targets
- Work robustly even with just 1,000 training samples
- Generalize across 29 computational and experimental datasets: including CoreMOF, QMOF, and even text-mined datasets
[5/6]

Instead of directly generating complex structures, EGMOF introduces a descriptor-based modular workflow:

- Prop2Desc: a diffusion model that maps target properties to chemically meaningful descriptors

- Desc2MOF: a transformer that reconstructs full MOF structures from those descriptors.
[4/6]

To address this, EGMOF (Efficient Generation of Metal–Organic Frameworks): a hybrid diffusion–transformer framework is developed that rethinks how AI approaches materials generation.
[3/6]