Groups keyboard shortcuts have been updated
Dismiss
See shortcuts
Skip to first unread message
Chris Khoury
Dec 26, 2023, 10:17:06 PM12/26/23
to Biociphers
Happy Holidays San!
Quick query - has outlier detection been implemented again in MAJIQ 2.5?
Many thanks,
Samantha/Chris.
San Jewell
Jan 3, 2024, 9:46:00 AM1/3/24
to Biociphers
Hi Samantha/Chris,
There was no major revision to any feature with this kind of functionality that I'm aware of. Could you please give a more specific example of the feature that you are referring to?
Thanks,
-San
Chris Khoury
Jan 3, 2024, 7:12:33 PM1/3/24
to Biociphers
Happy New Year San!
In older posts (https://groups.google.com/g/majiq_voila/c/TEj_bCSLYm4/m/KFNV-DksAgAJ), it was noted the argument 'weights', was an option that could be called, following the team's 2018 Bioinformatics research paper by Norton et al (https://academic.oup.com/bioinformatics/article/34/9/1488/4721782).
In our neurodegenerative disease biomarker study, we have a vast array of RNA sequencing depth across all subjects. To obviate erroneous reporting of potential biomarker LSVs, prior to MAJIQ we remove samples which are dissimilar in depth. If we do include them in the future, we are thinking how best to manage these outlier samples "which might be exhibiting disproportionately large deviations in exon inclusion levels compared with other samples from the same experimental condition (biological replicates)" when compared to control subjects. (*outlier detection as defined in aforementioned paper). As such, it would be useful for us to also know how/if outlier detection in the current MAJIQ algorithm can help contribute towards that goal.
In older posts (https://groups.google.com/g/majiq_voila/c/TEj_bCSLYm4/m/KFNV-DksAgAJ), it was noted the argument 'weights', was an option that could be called, following the team's 2018 Bioinformatics research paper by Norton et al (https://academic.oup.com/bioinformatics/article/34/9/1488/4721782).
In our neurodegenerative disease biomarker study, we have a vast array of RNA sequencing depth across all subjects. To obviate erroneous reporting of potential biomarker LSVs, prior to MAJIQ we remove samples which are dissimilar in depth. If we do include them in the future, we are thinking how best to manage these outlier samples "which might be exhibiting disproportionately large deviations in exon inclusion levels compared with other samples from the same experimental condition (biological replicates)" when compared to control subjects. (*outlier detection as defined in aforementioned paper). As such, it would be useful for us to also know how/if outlier detection in the current MAJIQ algorithm can help contribute towards that goal.
i.e is this something MAJIQ psi, dpsi or het deals with now? Does MAJIQ automatically run this outlier detection now and downweight the sample and the junction? Is there an option to report back to the user which sample and LSV exhibited vastly different exon inclusion levels compared to others of the same condition?
Many thanks,
Samantha/Chris
Reply all
Reply to author
Forward
0 new messages