mmquant
VERONIKA CHALUPOVÁ & HANA BOHÁČOVÁ


mmquant
•  A tool to quantify gene expression
•
  Published on: 15 September 2017 by Matthias Zytnicki
  Last upload: 8 months and 28 days ago
   version: 1.3
  Language used: C++
  Operating system: Linux; Mac OS X

How does it work
 - this tool counts with duplicated genes: if read maps to different positions, corresponding genes
are duplicated -> this tool then creates a merged gene





 - by default, the method supposes that reads have been sorted beforehand
 - if not: genes are sorted into a vector, cutted into non-overlapping bins and index is given to
the first gene in bin; then for each read genes are scaned starting from first gene in bin

1. step of genes quantification
 = searching for reads matching genes
 The way a read R is mapped to a gene A depends on the -l n value set by user:







 - if read is mapped to several locations, the tool sets NH tag of SAM/BAM file to value >1
htseq-count: union / intersection-strict / intersection-nonempty
mmquant:      -l 1    /              -l -1              / no alternative (ambiguous reads are
discarded)

2. step of genes quantification
 = resolving ambiguities
 - when read matches several genes, some can be discarded depending on number of overlapping base
pairs
 -d n computes the differences of overlapping nucleotides (N_A, N_B). If N_A ≥ N_B + n, then the
read will be attributed to gene A only.
 -D m compares the ratio of overlapping nucleotides. If N_A / N_B ≥ m, then the read will be
attributed to gene A only.

◦- featureCounts: option largestOverlap (assigns to the gene read with largest number of
overlapping bases)
◦- mmquant: emulates this strategy by –d and –D parameters


Fig. 3 Fig. 3

Input    Output
 Compulsory options:
◦annotation file in GTF format
◦reads in BAM/SAM format
 The output is a tab-separated file. It also provides output stats on hits.

Comparison with other tools
 • time:
-the fastest: featureCounts
-also fast: mmquant
-slowest: htseq-count

 • number of expressed genes given by each tool is comparable
-but multi-mapping genes could provide up to 25% of new genes
-without them the results could be biased

Thank you for your attention