[bioinfo] Fusion detection software FusionMap analysis process and report results

Image source: https://en.wikipedia.org/wiki/Fusion_gene

written in front

The main content below is about the fusion of RNA-seq data analysis, and the software used is FusionMap [ FusionMap reference ].

Which software is used for fusion analysis and which software performs better? I found a question and answer in Biostarts that listed some software ( see here ). There are more than 30 fusion analysis software such as STAR-Fusion, STAR-Fusion, deFuse, FusionCatcher, etc., of which about 20 The literature of multiple software was published in 2011-2013, and the literature of FusionMap software was also published in 2011. There are also several software comparison literatures, and the pros and cons of each analysis software will also be mentioned in the literature, and the literature published later will also be compared with the previously published software.

In addition, the FusionMap software should not be updated very early, and it is maintained in the Oshell toolkit .

FusionMap fusion detection principle

Fusion Reads: Seed readsand Rescued reads

Fusion Direction: Figure Source

FusionMap vs other soft comparisons

Image source literature: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4797269/

The red mark in the above table is that, compared to FusionMap, the result is worse than FusionMap. From the above table given in the literature, it can be seen that FusionMap performs well on the three groups and structured data sets, but performs poorly on the sample data of breast cancer and melanoma. The comprehensive evaluation of it belongs to the medium level, but One of its biggest advantages is that it is written in C#, and its running speed is faster than other software.

For the soft comparison literature , what kind of data is used specifically, the parameters used in the analysis of each software, and the scoring criteria for comparison may have an impact on each software.

FusionMap analysis process

Software analysis process pipeline: Figure source

Fusion detection process:

Among them, sequence alignment is based on some improvements made on the basis of GSNAPsoftware . Introducing GSNAP

(1) Analysis process configuration oscript file example :

http://www.arrayserver.com/wiki/index.php?title=OmicScript_example_for_RNA-Seq_data_analysis_pipeline

（2）Example of software usage:

mono oshell.exe --runscript Base_Dir Script_path/buildIndex.oscript Temp_Dir Mono_Path

Description of FusionMap result file

Result file report : http://www.arrayserver.com/wiki/index.php?title=Fusion_SE_report

header name	meaning
FusionID	Fusion ID information, the format is: FUS_Start_ENDNote ${}^{Note\left[1\right]}$
Bam.UniqueCuttingPositionCount	The number of Uniq reads, which is equivalent to the deduplication of Seed Reads+Rescued reads
Bam.SeedCount	As shown in the figure above, assume $\alpha$ is the minimum length of softclip at one end, and SeedCount issoftclip长度>=αthe number of Reads. (If the value is relatively small, it may not be comparable at all.) The purpose is that these Reads can be used as a seed sequence to expand into a longer fusion sequence, and then use the extended fusion sequence as a self-constructed ref, and compare the marginal fusion sequence to the self-constructed ref. Build ref. If it is PE150bp,$a=25$ , which can be expanded into a fusion sequence of 125+125=250bp at most
Bam.RescuedCount	The equivalent softclip长度<αnumber of reads, the reads on the comparison are carried out through the self-constructed ref of SeedReads
Strand	chain direction
Chromosome1	breakpoint 1 chromosome
Position1	breakpoint 1 location
Chromosome2	breakpoint 2 chromosome
Position2	breakpoint 2 position
KnownGene1	breakpoint 1 gene
KnownTranscript1	Breakpoint 1 transcript
KnownExonNumber1	Breakpoint 1 exon number
KnownTranscriptStrand1	Breakpoint 1 gene strand direction
KnownGene2	breakpoint 2 gene
KnownTranscript2	Breakpoint 2 transcript
KnownExonNumber2	Breakpoint 2 exon number
KnownTranscriptStrand2	Breakpoint 2 gene strand direction
FusionJunctionSequence	Fusion breakpoint upstream and downstream (30bp) sequences
FusionGene	fusion gene
SplicePattern	Fusion Splicing Mode[1]
SplicePatternClass	Types of fusion splicing patterns[1]
FrameShift	The format in which frameshift occurs [2]
FrameShiftClass	Type of frameshift[2]
Distance	Distance between fusion breakpoints (-1 if not the same chromosome)
OnExonBoundary	Whether it is on the Exon boundary, None: neither breakpoint is on; Both: both breakpoints are on; Single: one breakpoint is on.
Filter	Information can be filtered, including: InFamilyList (family gene list)/InBlackList (blacklist list)

in:

[1] SplicePatternClassincludes:

• CanonicalPatter[Major]: GT-AG SplicePattern
• CanonicalPatter[Minor]: GC-AG and AT-AC SplicePattern
• NonCanonicalPatter: all other detected di-nucleotides

[2] FrameShiftClassincluding:

• FrameShift: A frameshift has occurred at the fusion.
• InFrame: The fusion point is the whole code (the base of the gene at the breakpoint is a multiple of 3).

FrameShiftThe format of the corresponding value is: [0{0,1}1{0,1}2{0,1}->0{0,1}1{0,1}2{0,1}(python regular expression 0{0,1}means that the 0 character appears 0-1 times)

• (1) If the value is ->or ->0{0,1}1{0,1}2{0,1}or 0{0,1}1{0,1}2{0,1}->(for example, ->0// ), it may be that the two breakpoints or one end breakpoint of the fusion alignment are not in the coding region ->01? ->012【InFrame】
• (2) If the value is 0->1or 1->2or 2->0(below), it means that after the two genes are fused, no frameshift occurs【InFrame】
• (3) If the value is 0->2/ 0->0, 1->0/ 1->1or 2->1/ 2->2, it means that after the two genes are fused, a frame shift occurs【FrameShift】
• (4) If the value is ->left or right, there are multiple modes, for example 02->2, 012->1. When multiple patterns are included in case (3), it is [ FrameShift], for example: 0->02, 01->0, 02->2, 1->01, 12->1, 2->12; When at least one of the multiple patterns belongs to situation (2), it is [ InFrame】, for example: 0->012, 0->01, 01->01, etc.

The source of the figure also introduces the fusionMap detection fusion:

Suggested filter : graph source

SeedCount>= 3; SplicePatternClass=CanonicalPattern[Major] or CanonicalPattern[Minor]; Filter=Empty
More stringent conditions: FrameShiftClass=InFrame; OnExonBoundary=Both

Gene expression results of single-end/double-end fusion : source of graph (set analysis expression step in oscript configuration)

---

FusionMap mono CUP settings

High CPU usage, how to set it? 【Not sure yet】

• FusionMap usage documentation

• Some installation instructions about FusionMap mentioned that in the control document example, some parameter settings of mono are mentioned: (but I don’t know where to set this file?)
  

There are related parameters in Mono's help documentation :

parameter	illustrate
–aot
Environment variables:
MONO_CPU_ARCH	Overrides the automatic CPU detection mechanism. Currently only for arm, eg:MONO_CPU_ARCH="armv4 thumb" mono ...
MONO_THREADS_PER_CPU	Generally the maximum number of threads in a thread pool will be 20 + (MONO_THREADS_PER_CPU * number of CPUs). The default value of this variable is 10
MONO_TLS_SESSION_CACHE_TIMEOUT	The amount of time, in seconds, that the SSL/TLS session cache will keep its entries to avoid new negotiations between the client and server. Negotiation is very CPU intensive, so application-specific custom values ​​may prove useful for small embedded systems. The default is 180 seconds.

Other references:

• Mono related issues MONO_THREADS_PER_CPU=100 reference on github
• Special attention must be paid to using mono to run c# programs on linux. The problem of while (true) (using sleep) refers to
• Configure supervisor to manage mono program reference
• https://www.mono-project.com/docs/
• Fusion mechanism and detection method

---