10:00 am – 8:20 pm, 9:20pm – 12:00am
MERFISH
- reply to Quake lab with raw data
- working on fixing lib12 build
- found bug in my matlab BLAST — allhits field not being populated if no varargin’s passed. fixed.
- updating MERFISH team on discussions with collaborators, 12:00pm-12:40pm
- adjusting Library building script to allow buffering of total probe numbers 1:30pm – 3:00pm
- need to shuffle pri-secondary combos and secondaries when changing to new secondaries. (primary encoding sequences will be the same).
- running script to assemble library 12, 3:00pm – 4:00pm
adding remove probes so as to balance bits 4:00pm – 7:00 pm
- better to do based on bit combos
- potential bug in probe assembly: probe names are non-unique. Not certain targeting sequences selected at random are being selected without replacement… Need to investigate this.
- Investigated: these are from different 1kb sections, that happen by chance to have the same base end. Hence our sequence IDs are non-unique. should fix this for ID purposes, but makes no difference for probes.
Running probe building (again), 7:00pm – 12:00am
- serious problems with FAST libraries losing whole bits or large number of bits when fused to primers in the final BLAST.
- wrote loop to run through primers if we don’t keep at least 75% of the least abundant bit combo (since everyone will be pruned down to the least abundant).
- FAST secondaries + lib5 primers are getting destroyed by BLAST to abundant genes and rRNA/tRNA.
- memory is getting destroyed by BLAST
- launching on Morgan, 10:15 pm
- Writing to TSTORMdata2 is at least 10x slower than writing straight to disk, even from Morgan. Nope. 10x was too optimistic. Monet lapped Morgan in printing libraries. TSTORMdata2 is probably closer to 100x slower.
- On Monet, added breakpoint to start writing code to ID primers and secondaries that are highly repeated and killed by BLAST to abundant genes.
- BLAST is extremely slow for this purpose. (30 min per attempted primer and secondary set). Investigating BLAT alternative.
- probe names are non-unique.
- should have recorded which pt number of the sequence the probe was from and this would not have been a problem
- tried adding Tm. – no luck, still non-unique
- adding instead random 12-number integer id per probe.
- everything is unique. BLAST still refuses to build a database. darn it.
checking probe numbers:
- variable length 4490 > 1, 5374 > .1, Tm 66-76
- 35-55% GC content gives 4057 of 8750 with FPKM > 1
Multiplex chromatin
- discussion with Bogdan on strategy 11:30 am 12:00 pm
Discussion with XZ
- discussed setup distribution
Ph project
- reran simulations on cluster size in Ph-ML to match expression range. Finished much faster this time.
- zoomed in on sub-panels