Thanks for the comment! You're absolutely right - MAFFT does use FFT, and it's a great tool. But there's a key difference: MAFFT uses FFT to accelerate alignment (finding the optimal alignment path). Our approach uses FFT to compare sequences without alignment at all - it's a spectral resonance score between phase portraits, not a sequence alignment. In other words, MAFFT answers "how well do these sequences align?", while TEES answers "how similar are their spectral structures?". We see TEES as complementary, not a replacement. It could be useful for: Rapid screening of large datasets (where alignment is too slow) Detecting structural similarities that survive mutations (which might break alignment) Creating irreversible "passports" (SHA-256 hashes of windows) for privacy-preserving identification That said, we'd love to run a proper benchmark against MAFFT on a standard dataset. If you have a recommendation (e.g., a set of orthologs or a benchmark from BAliBASE), we're happy to run it and share results. Thanks again for the historical context - it's helpful to know where we stand!
Has anyone tried using spectral resonance (FFT-based coherence) for alignment-free DNA comparison? We've developed an open-source Python tool that does exactly that. It computes a resonance score (0–1) between two sequences based on their phase portrait, without alignment.
Benchmark (HBA1 & INS):
Comparison Score (chaos mode)
Human HBA1 - Human INS 0.64
Human HBA1 - Mouse HBA1 0.56
Human HBA1 - Chicken HBA1 0.50
Mouse HBA1 - Human INS 0.25
Speed: ~0.5 sec for 4 sequences. Caching (SQLite) makes repeated queries instant.
Repository: http://github.com/Dimius0/spectravortex
Direct script: dna_tees_analyzer_v5_absolute.py
Would appreciate any feedback, testing on real datasets, or suggestions for improvement.
Thanks!
1 answer
I don't know if this is exactly the same thing you're asking about but tools like MAFFT (https://pmc.ncbi.nlm.nih.gov/articles/PMC135756/) have existed for a long time, and as the name suggests, FFTs are the underlying mechanism.
Though they were quick, as far as I know the methods have been largely superseded. I haven't used MAFFT for 10 years and it wasn't exactly new then.
There may be some mileage in this if one method can truly be better in all of these different areas, but you will be facing slightly different 'competitors' in each space.
Rapid screening of large datasets (where alignment is too slow)
This, for example, would be where kmer approaches currently reside and they're pretty good by all accounts. You would just have to do the head to head comparisons for each use case I think.
There's probably not much value in benchmarking against MAFFT specifically, as I don't think its the most widely used anymore. You're probably looking more at Clustal Omega or maybe MUSCLE. There might even be newer ones I'm not aware of.
Thanks for the professional insight, Joe. We're not trying to compete with established tools or claim a niche - that's not our goal. We're exploring a fundamentally different approach to structural analysis: deterministic, resonance-based, rather than probabilistic or heuristic. We see that the combination of computational speed and our deterministic quantum-like algorithms may produce emergent effects useful across various fields - including genomic data screening and synthetic construct verification. We don't yet know how relevant this approach might be in your specific field. We've run tests on available data, and we believe the results are worth attention. We're open to dialogue and ready to adapt the method to real-world tasks if we see meaningful applications. Thanks for taking the time and pointing out the benchmarks - it really helps us better understand the context.
It sounds interesting, but ultimately it only matters if it is a material improvement on the existing approaches to these problems (or does something nothing else can).
Unfortunately, it doesn't matter how clever it is under the hood if it doesn't outperform whatever the standard in that space is (regardless of how you are actually positioning it).
Log in to answer this question.
I took a glance at this repository. Frankly speaking, looks like a work of a schizophrenic armed with AI. Did you try to publish your work in a reputable journal, and if you did, what was the reviewers' response?
If answering this question, please don't generate the answer entirely with AI, like you did when responding to Joe's comment.