mschilli87 / dropseq_barcodes

Collapsing barcodes

barcodes

Collapsing barcodes generated by faulty dropseq beads. Implemented in Python.

Scripts

• deletionsAnalysis, collapses barcodes according to existence of Ts in the end of UMIs. Assumes the existence of a single tab separated file containing each barcode and its UMI.
• checkHamming, collapses barcodes from bulk onto top ones. Only mutations are currently considered. Assumes the existence of two files, topBarcodes.txt and restBarcodes.txt, which are lists of barcodes sorted by number of reads.

To-do list

High priority

• collapse a set of top barcodes among themselves.

Medium priority

• introduce dynamically thresholds for number of reads and/or percentages of Ts.
• introduce check for percentage of Gs in the last positions of barcodes when Ts are found in the UMIs.
• introduce computeKnee function to have an estimate of the number of tops.

Low priority

• refactor the deletionsAnalysis search to incorporate arbitrary number of bases to check.
• improve the structure of identifying mutations function in the checkHamming.
• possibly trear Ns in the checkHamming.

dropseq_barcodes's Issues

introduce computeKnee function to have an estimate of the number of tops

taken from @nukappa's To-do list in initial README.md (priority: medium)

standardize/document command line argument parsing

taken from To-do list in initial README.md for the C implementation (priority: low)

possibly trear Ns in the checkHamming

taken from @nukappa's To-do list in initial README.md (priority: low)

improve the structure of identifying mutations function in the checkHamming

taken from @nukappa's To-do list in initial README.md (priority: low)

collapse at least 2 deletions before checking mutations

taken from To-do list in initial README.md for the C implementation (priority: high)

heisenbug (manifests with -O2) in clearing hit_lists

Commit db935d1 introduced a bug that causes a segmentation fault when clearing the hit_lists during top-barcode collapsing.

This bug does not occur with -O1 optimization.

Even with -O3, the bug disappeared when introducing fputs statements within the clear_hit_list definition.

refactor the deletionsAnalysis search to incorporate arbitrary number of bases to check

taken from @nukappa's To-do list in initial README.md (priority: low)

introduce check for percentage of Gs in the last positions of barcodes when Ts are found in the UMIs

taken from @nukappa's To-do list in initial README.md (priority: medium)

add UMI correction

taken from To-do list in initial README.md for the C implementation (priority: high)

collapse a set of top barcodes among themselves

taken from @nukappa's To-do list in initial README.md (priority: high)

allow for more deletions

taken from To-do list in initial README.md for the C implementation (priority: low)

introduce dynamically thresholds for number of reads and/or percentages of Ts

taken from @nukappa's To-do list in initial README.md (priority: medium)