Transcript Document
RNA surveillance and degradation: the Yin Yang of RNA AAAAAAAAAAA RNA Pol II production RNA destruction AAA Ribosome MODEL: Mtr4 Trf4p Polyadenylation AAAAA by Trf4p * * * * Hypomodified tRNAiMet * * * * * AAAAA Rrp46p Csl4p Rrp43p Rrp44p Rrp45p Rrp42p Mtr3p Rrp41p * * * * Rrp6p Rrp40p Exosome Rrp4p Degradation of hypomodified tRNAiMet *- Hypothetical diagram of the exosome Workflow Knockdown mMtr4 Connect& Compare Collect Library Construction PolyA-Seq Mapping Normalize Aggregate Remove Internal A Visualize Next Gen sequencing PolyA-Seq TRAMP Complex AAAA AAAA Papd5 ZCCHC7 Mtr4 AAAA AAAA siRNA knockdown Library creation for NGS Map paired end reads to genome • BWA (Burrows-Wheeler Aligner) Algorithm used to map each pair of reads to the genome • Report each pair of reads as a single nucleotide position within the genome where polyadenylation detected in an RNA sample • Average insert size 300 – Read size ~45 3’-A AAAA-3’ TTTp-5’ Raw reads vs Mapped reads Data type/kd type Raw reads Mapped reads positions Mtr4 15,135,078 10,853,534 651,551 Ctrl 16,348,780 11,708,310 652,128 Rrp6 15,971,926 12,388,266 705,173 Mtr4 ND 34,204,534 1,124,968 Ctrl ND 7,195,942 582,256 Rrp6 ND 8,241,505 597,672 Replicate Data Original data Normalization of data: reads per million (rpm) Analysis • Starting with refseq database – Raw read counts converted to reads per million • Reads at position/total reads in sample – Remove all non-coding RNAs – From each sample collect normalized reads mapping at the 3’ end +/- 50 bases of each refseq encoding protein – Dot Plot normalized reads on log scale, X axis=control and Y axis=mMtr4KD mRNA polyadenylation does not change between Mtr4 and control KD 10000 R2=0.95141 Normalized pASeq reads in mMtr4KD 1000 100 10 1 0.1 0.1 1 10 100 Normalized pASeq reads in mControlKD 1000 10000 Problems encountered • Sequencing read depth very different in the original data – 34 mil mapped reads in one sample 8 mil in other • Lack of 3 replicates for robust statistical analysis of data • Removal of internal A – Seq reads that map to a oligoadenylate track in the genome – Algorithm developed misses many – Manual removal takes too much time. Remove Internal A AAAAAAAA AAAAAAAA TTTTTTTTT TTTTTTTTT How to mine the data based on a hypothesis • Hypothesis: PolyA+ RNAs of unknown identity will accumulate upon depletion of mMtr4 vs. the control. – How can the transcriptome be queried? – How detailed should a query be? • Every pA position, or only those exhibiting greater than x number of raw/normalized reads? • How do we find significant differences with one sample, or possibly two? • How can repetitive elements be accounted for in the data? Custom annotation to remove bias from existing annotations • Data mapped with Bowtie to mouse genome mm10 build • Mapped data from KD and control compared using cufflinks to explore gene expression differences using a custom annotation • Custom annotation – 1000 base pair genes with 500 base pair overlap with next gene • This did not work well Problems with using custom annotation • First real problem was the no computing could handle more than 5000 genes of the custom annotation at a time – One chromosome had 147K genes • There was a problem with assignment when the reads overlapped – Cuffdiff would randomly assign the reads to only one of the genes. • Overlaps split into two fasta files, but we could not capture differences in the data that we knew exists. – cuffdiff collects data from the entire 1000 bp gene and compares between 2 samples – This method leads to false negatives for pA data where the focus is on one or a few positions as a pA event. What next? Mapping • Map raw reads against mm10 assembly with Bowtie2/ Tophat Strand and 3’ end selection • Select alignments on positive and negative strand • Select 3’ read of paired reads to define site of polyadenylation Custom annotation preparation and count • Run F-Seq to identify the mode of all peaks • Normalize data then collect reads at mode (+/-5-10 nucleotides) Statistical test with DESeq, an R package • Negative binominal model F-Seq • Tags to identify specific sequence features for different library preparations (ChIP-seq), (DNase-seq) and (pA-seq). • Will summarize and display individual sequence data as an accurate and interpretable signal, by generating a continuous tag sequence density estimation. Generating Peaks with FSeq • 1. Estimate kernel density to estimate pdf • 2. compute threshold – – – – nw=nw/L. xc, Repeat step 2 k times s SDs above the mean • 2.1 threshold output module is modifiable Magnitude of data: one sample both strands 51 million bases of Chromosome 12 12 thousand bases of Chromosome 12 Chromsome 12 is 121 million base pairs long rRNA workflow Mapping • Map raw reads against 13kb rDNA with Bowtie2 Strand and 3’ end selection • Select alignments on positive strand • Select 3’ read of a pair and 3’ end of a read Density estimation and visualization • Density estimation with F-Seq, a peak calling tool 1 316 631 946 1261 1576 1891 2206 2521 2836 3151 3466 3781 4096 4411 4726 5041 5356 5671 5986 6301 6616 6931 7246 7561 7876 8191 8506 8821 9136 9451 9766 10081 10396 10711 11026 11341 11656 11971 12286 12601 12916 13231 13546 pA reads intersecting 45S prerRNA 20000 18000 16000 14000 12000 10000 Ctrl 8000 Mtr4 6000 4000 2000 0 18S 5.8S 28S 1 309 617 925 1233 1541 1849 2157 2465 2773 3081 3389 3697 4005 4313 4621 4929 5237 5545 5853 6161 6469 6777 7085 7393 7701 8009 8317 8625 8933 9241 9549 9857 10165 10473 10781 11089 11397 11705 12013 12321 12629 12937 13245 13553 pA reads intersecting 45S prerRNA 100% 90% 80% 70% 60% 50% Mtr4 40% Ctrl 30% 20% 10% 0% 18S 5.8S 28S Accumulation of micro RNA processed 5’ leader upon depletion of Mtr4 • Comparison of Mtr4 V. Control KD • Abundant polyA found near 5’ end of annotated Mir322 • Confirmed using molecular technique