Alistair's Lab Notebook

(Wed and Thur in NYC for Dale Frey interview)

9:30 am – 7:00 pm

sequencing data processing

• bowtie finished running on all the datasets (20)
• Need to run cufflinks.
• first need to sort sam data
• using sam tools (should have set this up to run overnight, it’s quite slow)
• also need to copy the data
• (this RNA-seq pipeline I have sucks — 6 hours to download the data, 4 hours to unzip it, overnight to run bowtie (probably finished in less than 1 hour since I did that multicore 20 threads), 2 hours to resort the data with samtools (longer because I was messing around with trying to multicore this), then XX hours to copy the data to RC and XX hours to run cufflinks.

Sample organization:


Sample Name Tube Tindex IndexSeq NEB Description
Ph1 A 1 ATCACG 1 original Ph KD sample
M1 A 2 CGATGT 2 original WT control sample
2-4P A 3 GCCAAT 6 KD performed on day 0 and day 2, extracted on day 4
4W A 4 CAGATC 7 latest WT sample, extracted on day 4
4P A 5 ACTTGA 8 latest Ph-KD sample, extracted on day 4
10-22#1 B 1 ATCACG 1 (WT) sample extracted on day 4 (need to check ID)
10-22#2 B 2 CGATGT 2 (Ph-KD) sample extracted on day 4 (need to check ID)
2W B 3 TTAGGC 3 WT sample, extracted on day 2
2P B 4 TGACCA 4 Ph-KD sample, extracted on day 2
2-4W B 5 ACAGTG 5 WT for KD performed on day 0 and day 2, extracted on day 4


bash CufflinksArray1.bash

repo issues

• somehow my bt2 files got put in my repo and accidently committed last night
• BFG is an excellent tool for quickly and easily scrubbing these from the repo (https://rtyley.github.io/bfg-repo-cleaner/), which has previously been a horrible pain.

9:00 am – 7:30 pm

Library building code

• fixed bugs in library design — can’t call TRDesigner with parallel pool with large off-target libraries.
  • it takes a ton of time and a ton of memory to replicate these large OTTables
  • this is blazingly fast in non-parallel mode.
• No need to use parallelization in building OTTables — launching the parallel matlab cores is slow, uses more memory, and doesn’t speed things up for this code (at least on the scale of my current library design).
• worked out primer building
• don’t actually need lots of index primers, part of the beauty of the new design is to move away from having tons of individually indexed regions.
• requested readout probes sequences from SW, stick with 30mers, these are working well.

Updated scripts

• AssembleChrLib7_151103
• S151103_PlotCutSitesAndChIPdata
  • (finished creating coverage vectors for all ChIP-seq data).

Presentations

• seminar on interviews (see notes)
• practicing talk for DF interview tomorrow.

9:15 am – 6:00pm, 8:00pm – 8:55 pm

Library building

• goal: building 10 kb tiling regions of BX-C probes against genome
• approach: adapt new OligoArray-free pipeline from transcriptome libraries to design genome libraries

Problems

• crashed Morgan when all 256 GB of RAM got paged during penalty calculations
• this may have happened because the target sequences are all included in the off-target table (the whole genome)
• tested this hypothesis by running against chr2L only (currently constructing for BX-C on 2R)
  • this runs quickly without any sign of crashing.
  • (might have been a mis-interpretation — code is written to load an existing trDesigner rather than rerun. May not have attempted to use this off-target library at all.)
  • indeed this appears to have been the case, now rerunning.
• wrote new function to remove regions probed from the genome sequence and create a new offTarget fasta file database of the genome for this library construction
  • writing whole-genome fasta files is not fast, even for Drosophila (nor is loading them, though the fly fasta loads pretty quickly)
  • building OTtable to whole genome duplicated to plus and minus strands, (with BXC removed) goes pretty fast — less than 400 seconds
• this goes MUCH faster with local databases, read-write speed pretty important, doesn’t work well over network.
  • moved to new folder on MorganData (ChromatinLibraries/Library7)
  • (originally had data on Monet in Chromatin/OligoLibraries/Library7)
• calculating penalties is pretty slow
  • this is the step that caused the memory to destroy the computer running (maxed out).
  • currently have 29 GB paged, now calc’ penalities with the chr2L_plus strand only. We’ll see how much it eats and whether this destroys the system again. (started 8:45 pm).
  • even without BX-C this does indeed use tons of memory. aside from that it’s not killer slow.

12:00 pm – 6:30 pm, working on applications
Submitted all apps with Nov 1st deadlines.

9:30 am – 5:00 pm

Applications

• working on Harvard MCB application
• applications won’t submit requests to letter writers until application is complete.
• Princeton application also won’t submit requests to letter writers
• given that a motivated candidate would like to have a tailored, carefully edited application to each school which has the most recent updates of publications, it is quite convenient to submit the final application at the last moment before it will be read. This is however highly inconvenient to letter writers, who get a minimum of advanced notice for no especially good reason.

9:00 am

Deep seq final prep

• running tape-station

samplesname, conc pg/uL, index, tube

• Ph1, 496, 1, A
• M1, 442, 2, A
• M2, 665, 4, –
• 10-22 #1, 652, 1, B
• 10-22 #2, 516, 2, B
• 2W, 617, 3, B
• 2P, 708, 4, B
• 2-4W, 588, 5, B
• 2-4P, 686, 6, A
• 4W, 565, 7, A
• 4P, 631, 8, A

final

• A mixed 10 uL of each (as marked above). conc ~2.82 nM, call it 3.0 to be safe
• B mixed 10 uL of each (as marked above). conc ~3.2 nM, call it 3.5 nM to be safe. 50 uL
• sent for sequencing.

Interview prep

• working on slides for phone interview.

Manuscript

• shrinking figures to fit limitations
• implemented reference trimming as discussed yesterday.
• still need BB’s rearranged fig 2 with added legends.

9:30 am – 7:00 pm, 9:00 pm – 11:00 pm

Deep seq prep

• continued with deep-seq prep
• finished second strand synthesis
• ran clean up on DNA-XP beads from Bauer
• ran adaptor ligation. premixed adaptor and enzyme mix (2 min on ice), this is bad it will increase adaptor dimer.
  • we’ll quantify this on the tape-station and see how much trouble it will cause.

RNAi quantification by qPCR

• set up QPCR plate, same organization as last time.
• cDNA row order: 4W, 4P, 2-4W, 2-4P
• ran tape-station (only 4 lanes left, ran ladder and lanes 1-3). looks decent.
• ran kapa quantification

seq prep

• previously used 1,2,4
• setup 1 & 2 as repeats with 6,7,8 as multiplexed. (Illumina 1,2,6,7,8)
• run new 1-5 multiplexed together (Illumina 1,2,3,4,5)
• submit for 20 uL, 2x single lanes.
  

References to cut (sad, I like these papers)

• Killian 2015
• Kharchenko 2011
• Ernst 2011 (ENCODE)
• Ram 2010 (ENCODE)
• Li 2015
• Galupa 2015, (Heard Lab)
• Rivera 2013 (Ren Lab)
• Gomez-Diaz (Corces lab)
• Rosa 2014

maybe cut

• Gorkin (Ren Lab)
• Gibcus (Dekker Lab)
• Bernstein & Lander 2007 (~idea founder but old)
• Simon and Kingston 2013
• Boyle, Yamada, and Beliveau 2012 to supplement, Chen 2015 to supplement?

notes

• Bickmore 2013;Gibcus 2013;Levine 2014;Gorkin 2014;Sexton 2015;Galupa 2015;
• 8-10 Bernstein 2007;Rivera 2013;Bickmore 2013;
• Bickmore 2013;Gibcus 2013;Levine 2014;Gorkin 2014;Sexton 2015;Galupa 2015;Bickmore 2013 (1);

8:30 pm – 11:15 pm

RNA deep-seq prep

• sample order (8 samples): 10-22 n1, 10-22 n2, 2W, 2P, 2-4W, 2-4P, 4W, 4P
• running sample isolation by magnetic polyT beads, following manufacturer’s protocol.
• 3 samples dehydrated in last elution step. I added more elution/fraction buffer to resupsend the dried out beads and reheated to 94C for 1 min.
  • this probably incurred some sample loss though. We’ll see in the quantification steps.
• running 1st strand synthesis reaction overnight

cDNA prep for QPCR validation of RNAi

• running RNaseH digestion