Open h-ahl opened 2 months ago
Thank you for bringing this issue to our attention. We will be working on updating the bioconda version asap. To help us better understand the issue, could you provide us with the sequence(s) you used?
Hi, @xinyu-g. Having some issues sharing the sequences in question due to legal restrictions. This should be a general issue though; you can evaluate it by generating a pareto front e.g. using LinearDesign and DERNA (Zuker) using the same codon table
@h-ahl That might make it a little challenging on my side to debug the potential issue. I have not seen such problem with the sequences I have run. For instance, using the P15421 sequence given in the sweep example in the README, I got the attached results, where lambda = 1e-05 gave me a CAI of 1.
I also tried using fixed lambda on another sequence (O14880) with the following command
./derna -i ../data/uniprotSeq/O14880.fasta -o O14880.fasta_fixed_lambda.txt -m 1 -s 2 -l 1e-5
And derna returned lambda: 1e-05,O: -0.2035,cai: 0,cai_s: 1,mfe: -20350,combined: -0.2035
which also gave a CAI of 1.
@h-ahl Hi, I wanted to follow up on the issue. Have you had a chance to run the sequences using your own codon table, now that the issue with loading a custom codon table has been resolved? It’s possible that the default codon table derna uses is different from the one you’re using to evaluate the sequences, which could be causing the problem.
Hi, @xinyu-g - I've asked my colleague to follow up on this. The issue shouldn't be codon table related, but thank you very much for resolving that issue
The bioconda version (1.0.2) is obsolete, as has issues in it preventing it from optimising CAI as intended. E.g. maximal CAI optimisation does not lead to a CAI of 1.