Alphafold, the Brawndo of Modern Science
In a classic scene from Idiocracy, we see the citizens of a fictional United States from year 2500 repeatedly insisting that Brawndo is the “thirst mutilator” because it contains electrolytes, while water belongs only in the toilet. The slogan is repeated so many times by the Brawndo company that it has become deeply embedded in their thinking.
Well, the world of modern science already arrived there in 2025 with Alphafold. Alphafold, if you do not know, is a computational algorithm that uses transformer-based neural network for protein-folding. It is an excellent piece of work.
Alphafold is also owned by a company in the business of making slogans. For years, it used the slogan “Don’t be evil”, while giving all of your private data to NSA. It also heavily censored all discussions on Covid virus coming from a lab.
Thus it was no surprise to find Messrs Hassabis and Jumper create the slogan - “Alphafold solved protein folding”. Scientific American wrote - “One of the Biggest Problems in Biology Has Finally Been Solved”.
Google DeepMind CEO Demis Hassabis explains how its AlphaFold AI program predicted the 3-D structure of every known protein
Another popular science magazine wrote -
Demis Hassabis and John Jumper have developed an AI model to solve a 50-year-old problem: predicting proteins’ complex structures.
In 2020, Demis Hassabis and John Jumper presented an AI model called AlphaFold2. With its help, they have been able to predict the structure of virtually all the 200 million proteins that researchers have identified.
This second pop-sci magazine happens to be the Nobel prize committee, who likely only read Scientific American and diligently repeated the slogan.
This had a rather deleterious effect on all kinds of scientific and social commentary around the world. Even people critical of AI-generated slops unanimously agree that Alphafold is a success story. Why? “Because it got the Nobel prize.”
I also fell for this con until this summer and sat with three students to work on a bunch of proteins from a unicellular eukaryote. The structures generated by Alphafold turned out to be AI slops. So, I looked into the literature and came across this paper from Henry Childs.
AlphaFold 3 Fails to Predict D-peptide Chirality, Fold, and Binding Pose in Heterochiral Complexes
Due to their favorable therapeutic properties, including improved stability, bioavailability, and membrane permeability, D-peptides that bind biological L-proteins represent an important class of systems in computational drug design. A reliable in silico workflow for these systems must correctly preserve stereochemistry while predicting fold and binding pose. The AlphaFold 3 (AF3) model reported by Abramson et al. (2024) enforces a strict chirality violation penalty to maintain chiral centers from model inputs and is reported to have a low chirality violation rate of only 4.4% on a PoseBusters benchmark containing diverse chiral molecules. Herein, we report the results of 3,255 black-box experiments with AF3 to evaluate its ability to predict the fold, chirality, and binding pose of D-peptides in heterochiral complexes. Despite inputs specifying explicit D-stereocenters, we report that the AF3 chirality violation rate for D-peptide binders is much higher at 51% across all evaluated predictions; on average the model is as accurate as chance (random chirality choice, L or D, for each peptide residue). Increasing the number of seeds failed to improve this violation rate. The AF3 predictions exhibit incorrect folds and binding poses, with D-peptides commonly oriented incorrectly in the L-protein binding interface. Confidence metrics returned by AF3 also fail to distinguish predictions with low chirality violation and correct docking vs. predictions with high chirality violation and incorrect docking. We conclude that AF3 is a poor predictor of D-peptide chirality, fold, and binding pose and propose solutions to address these limitations.
In the meanwhile, now that Alphafold “solved protein folding”, all kinds of silly ideas are popping up in the literatures about how to use LLMs to “solve the cell” or cure cancer. AI will finally reveal the meaning of life in the same way as Brawndo solved the thirst problem for plants.
