Albe van der Merwe

Musician, scientist, whacky lover of all cool things.

[Albe van der Merwe @intrepidsci]

@fidelwachoo @EricLDaugh @grok check this for factual accuracy.

[Bitcoin Teddy @Bitcoin_Teddy]

For all the morons who say capitalism doesn’t work

[Albe van der Merwe @intrepidsci]

https://t.co/jJcZ4gxs2K

[Toni Gabaldón @toni_gabaldon]

We are looking for a Doctoral Candidate with a #bioinformatics background to use metagenomics to study microbial pathogens in non-clinical niches as part of #SPOILCONTROL @MSCActions DN DC5 - Pathogenic Flora https://t.co/GBNHsqjhfW

[Broad Institute @broadinstitute]

A new spatial transcriptomics approach eliminates the need for time-intensive imaging with specialized equipment, making the technique accessible to more researchers around the world. https://t.co/Yvki9deWE8

[Biology+AI Daily @BiologyAIDaily]

Identifying 14-3-3 interactome binding sites with deep learning

1. The study introduces a deep learning framework designed to predict protein binding sites to the 14-3-3 family, central hub proteins involved in various cellular signaling processes.

2. The researchers created an ensemble model combining multiple deep learning architectures, achieving 75% balanced accuracy in predicting 14-3-3 binding sites, even for challenging intrinsically disordered proteins.

3. This model was applied to 300 sequences of medically relevant proteins, including those linked to diseases like Alzheimer’s (Tau) and cancer (Myc, p53), identifying promising binding sites for further validation.

4. Experimental validation confirmed the model's predictive power, with five out of eight top predictions showing significant binding to 14-3-3 proteins, demonstrating the model's real-world applicability.

5. The study further supports the integration of deep learning with experimental methods such as X-ray crystallography and molecular dynamics to gain insights into protein-protein interactions and their structural dynamics.

6. The model was tested on various 14-3-3 client proteins like FOXO3, BAD, and Tau, with the top predictions providing new avenues for drug design targeting these key interaction sites.

7. By making the model accessible online (https://t.co/8AQrEt13Wg), the authors offer a tool for researchers to explore 14-3-3 interactions in their own studies, advancing our understanding of cellular signaling mechanisms.

💻Code: https://t.co/8AQrEt13Wg
📜Paper: https://t.co/IDk6p2nwUP

#DeepLearning #Proteomics #Bioinformatics #ProteinInteractions #Phosphorylation #Alzheimers #Cancer

[Jorge Bravo Abad @bravo_abad]

A 100-billion-parameter unified transformer for interpreting and generating protein sequences

Proteins underpin vital processes in biology, from cellular signaling to enzyme catalysis. Their amino acid sequences determine complex structural and functional properties, making it essential to capture the informational patterns embedded in these chains. In recent years, large-scale computational methods have emerged to learn meaningful representations of protein sequences, assisting researchers in areas such as protein structure prediction, activity engineering, and novel sequence generation.

Chen et al. introduce xTrimoPGLM, which unifies two core learning strategies—masking random tokens for bi-directional context encoding and autoregressively predicting continuous spans—within a single 100-billion-parameter transformer model trained on one trillion amino acids. The authors leverage a dual-stage schedule that first emphasizes masked language modeling, then integrates both the masked and autoregressive objectives. This enables the transformer to excel in residue-level understanding while generating biologically plausible sequences. Quantitatively, the model achieves perplexities as low as 6.70 on out-of-distribution tests, surpassing similarly large baselines. It also scales according to a power-law relationship, showing steady gains in representation quality when trained on increasing amounts of data.

The researchers demonstrate that xTrimoPGLM substantially outperforms prior large-scale protein language models across 18 tasks spanning structural classification, stability prediction, and molecular interactions. The model supports rapid folding calculations, with a specialized module yielding high TM-scores on challenging benchmarks. It also generates de novo sequences with diverse structures, which suggests strong potential in customizable protein engineering. By uniting advanced bi-directional encoding with autoregressive generation, this framework offers a versatile platform to advance both understanding and design of proteins at scale.

Paper: https://t.co/gJ3y8Lw6Ad

[Fritz Sedlazeck @sedlazeck]

Long reads @nanopore @PacBio can provide insights into epigenetics. We (@fuyilei96 @timp0) summarize the latest #Bioinformatics approaches to analyze methylation & why this matters! Out now @NatureRevGenet
https://t.co/rbjJcDQOrP
@BCM_HGSC @bcmgenetics @RiceCompSci @JohnsHopkins

[fajie yuan @duguyuan]

🚀 Update! Our latest Pinal bioRxiv now includes wet lab results. More proteins with diverse text prompt on the way.

Design proteins with just text. Everyone can do protein design!

Demo: https://t.co/4XxGI0Hp9x
paper: https://t.co/mXSygLEa0w
GitHub: https://t.co/mm8mynEuxL

[bioRxiv Bioinfo @biorxiv_bioinfo]

GeneWhisperer: Enhancing manual genome annotation with large language models https://t.co/NWrJZHR0bg #biorxiv_bioinfo

[Science Magazine @ScienceMagazine]

In a new Science study, researchers introduce GAIA, a statistical approach that seeks to learn the geographic position of every genetic ancestor of individuals included in a genome dataset.

Learn more ⬇️
📄: https://t.co/cD9oSZuEoE
#SciencePerspective: https://t.co/ikQfv84nYq

[Stephen Turner 🦋 @stephenturner.us @strnr]

Beyond the black box with biologically informed neural networks https://t.co/NDRMYMpPMA (read free: https://t.co/OTMVPM2Zhg) 🧬🖥️🧪

[Albe van der Merwe @intrepidsci]

This right here is what I've been waiting for. https://t.co/5hxXBb8djh

[Keenan Crane @keenanisalive]

Entropy is one of those formulas that many of us learn, swallow whole, and even use regularly without really understanding.

(E.g., where does that “log” come from? Are there other possible formulas?)

Yet there's an intuitive & almost inevitable way to arrive at this expression.

[Loïc A. Royer 💻🔬⚗️ @loicaroyer]

PI trying to finish a paper despite incessant problems, meetings, and a million other distractions.

[Albe van der Merwe @intrepidsci]

@leptograph @Fabiteam1 A fumble of forest pathologists?

[Albe van der Merwe @intrepidsci]

When an R becomes a P and changes the entire meaning.

[Albe van der Merwe @intrepidsci]

We are incredibly small and insignificant. But we are intrepidly exploring the cosmos with technology made with human hands and human brains. First #JWST deep field shows thousands of galaxies and gravitational lensing. Things with lens flares = stars that are much closer to us.

[President Biden Archived @POTUS46Archive]

The first image from the Webb Space Telescope represents a historic moment for science and technology. For astronomy and space exploration.

And for America and all humanity.

[Albe van der Merwe @intrepidsci]

@planet4589 @hollybik Are the different wavelengths for e.g. Oxygen because of different excitation states?