AF Cache: Efficient Pipeline for Running AlphaFold for High-Throughput Protein-Protein Interaction Prediction 1 AF Cache is a Nextflow pipeline designed to make large-scale protein–protein interaction (PPI) screening practical with AlphaFold2 and AlphaFold3 by cutting the dominant runtime bottlenecks: repeated MSA generation and (for AF2) repeated JAX recompilations. 2 The key idea is to generate MSAs once per unique monomer for the whole dataset, then reuse (“cache”) them across all pairs. For an all-against-all screen of N proteins, default multimer workflows can trigger ~N(N 1) chain-level MSA generations, while AF Cache reduces this to N by construction. 3 AF Cache accelerates MSA generation by using GPU-accelerated MMseqs2 and batching all proteins in a single dataset-level step (rather than per-target or small batches). It also overlaps CPU and GPU steps across UniRef and environmental database alignments to reduce GPU idle time. 4 For AlphaFold2 specifically, AF Cache adds sequence-length bucketing for multimers (bucketed by total chain length) and pads within each bucket so JAX compilation happens once per bucket instead of once per pair—saving ~1–2 minutes of GPU time for every additional multimer in that bucket. 5 Benchmark: 100 human mitochondrial proteins (40–1000 aa) were screened all-against-all including homodimers (5,050 pairs). Settings were streamlined for throughput (AF2: single model, 3 recycles, templates disabled; AF3: single seed, one diffusion sample). 6 Pre-prediction speedups (MSA stage): when comparing to a realistic 128-CPU-core baseline, AF Cache achieved ~13x faster MSA generation for AF2 settings using full BFD, and ~5x faster for AF3 settings using small BFD. The paper also reports very large raw GPU-vs-CPU core-hour reductions, emphasizing the benefit of GPU-based MSA search in high-throughput regimes. 7 End-to-end impact for AF2: caching bucketing reduced AF2 prediction compilation time from 253 to 125 GPU hours (>50% reduction), and made inference ~90 seconds faster per protein pair on average while preserving the same runtime scaling with pair length. 8 Output similarity was assessed via ipTM comparisons. Correlations between default vs AF Cache runs were moderate across all pairs (Pearson r ~0.70 for AF2; ~0.64 for AF3), but much higher for pairs where both proteins map to a shared PDB entry (r ~0.98 for AF2; ~0.94 for AF3), suggesting structurally supported cases are highly consistent across pipelines. 9 Practical deployment details: AF Cache provides a ready-to-use workflow for local and HPC environments, supports all-against-all or user-specified pair lists, automates AF3 JSON preparation, and can download/install dependencies (including databases; AF2 network parameters) as needed. 💻Code: github.com/clami66/AF_cache 📜Paper: arxiv.org/abs/2606.04566 #AlphaFold #ProteinInteractions #PPI #Bioinformatics #ComputationalBiology #StructuralBioinformatics #Nextflow #HPC #MMseqs2 #Proteomics

This is how you know tech bros are not going to solve major problems in biology.

KETO-CTA: Schrödinger's Plaque 🥓 In his latest lecture, @realDaveFeldman claims plaque progression was TOO RAPID to make sense. Yet, we were told "NO or MINIMAL plaque progression". Can progression be so great as to not make sense AND also "no or minimal"?

Im not sure I will ever overcome how furious I get reading reviewer comments. You spend 3 years of your life working on something just for 3 people to skim your paper and point out a few things that were already well explained in an incredibly rude way and then reject it

For the past 70 years, modern biology has been built on a static objects worldview: DNA = information, proteins = structures, cells = nanomachines, disease = broken parts, drugs = part repair/update. I'm honestly excited for a “new worldview” and “Dynamic Biology” to emerge.

Ok we are closer to agreement than it first appeared. N=1 experiments that are not randomized are equivalent to case reports, especially if documented to a similar degree. N=1 randomized experiments are the highest tier evidence for the individual in whom they are done provided that the time course is feasible and relevant to the outcome of interest. N=1 randomized experiments cannot be used for population-level guidelines because they don't apply outside the individual. However, since all information from any source is imperfect, and since many areas of interest have enormous gaps in knowledge, it is very valuable to use findings from N=1 experiments as potentially relevant to other individuals especially when those individuals are similar in genetics, history, and constitution in the variables thought to be relevant to inter-individual differences. Randomized self-experiments are practical on short time scales, especially when effects are acute over the course of hours. They are reasonably practical for time horizons of one day or several days. They are entirely feasible for time horizons that take weeks or longer but are difficult enough at those timescales that one would have to be extremely motivated to do it, either because the stakes seemed high or because one is a scientist in the classical sort, which is now almost absent from science. The classical scientist, the most genuine of all scientists, loved to experiment out of curiosity and loved to experiment on himself. There are many such scientists in history but they are a rare breed. Barry Marshall most famously with ulcers, Carl Wilhelm Scheele most notoriously for smelling and tasting every chemical he could, William Brown who working with George Burr tried to give himself an EFA deficiency. A R Berger 1, H H Schaumburg, C Schroeder, S Apfel, and R Reynolds taught us a lot about vitamin B6 toxicity by purposefully inducing it in themselves in self-experiments that lasted YEARS. They published their report in 1992. So the answer to your question is that no it is not practical but it is the sign of a genuine scientist of the highest breed.

Multimodality non-invasive cardiovascular screening reduces 5-year all-cause mortality in both sexes! doi.org/10.1093/eurheartj/eh… #multimodality #screening #cardiotwitter @ESC_Journals @escardio

Water is 70–80% of a cell, with our mspSA SGAG affinity grid, we capture a near atomic (<2 Å) view showing water actively driving gene transcription inside RNA Pol II. Not just background,water is part of the machinery. Now published in Molecular Cell.doi.org/10.1016/j.molcel.202…

I’m thrilled to announce that my new comment titled “How to design effective scientific figures” is now released from @NatureHumBehav This provides five keys to make your graphical items attractive 🎨 Hope this is useful for your project! nature.com/articles/s41562-0…