I am trying to find orfs with Glimmer. At the end, Glimmer gives me a .predict file with the orfs and a corresponding raw score.

When I look at the documentation, it says about the score, "This is 100 times the per-base log-odds ratio of the in-frame coding ICM score to the independent (i.e., non-coding) model score. It gives a rough quantification to how well an orf scores that can be compared between any two orfs." You can look here for the documentation-


But what does the score really mean in an absolute manner? How do I realize for an orf, if the score is significant or not?


The score is the log odds ratio, times 100, that the ORF that you input is a coding region. For example, if Glimmer thinks the log odds are 10 to 1 that your region is coding, then your score will be 10x100 = 10000. Log odds of 10:1 imply more than a million to one certainty.

| improve this answer | |
  • 1
    $\begingroup$ Thanks. But the documentation says log-odds ratio. But even then, my question is what is a good odds ratio absolutely. More abstractly, when can I be confident that this orf is really a coding region. $\endgroup$ – nafizh Aug 22 '16 at 21:45
  • $\begingroup$ You can't really be sure without experimentation and research, because glimmer can screw up. But a score of 300 or more suggests better than 90% odds that it is a coding region according to Glimmer. 10000+ implies a million to one odds. $\endgroup$ – D J Sims Aug 22 '16 at 21:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.