Kun:LabNotes/MONOD/2014-8-4

From ZhangLabWiki
Revision as of 18:16, 4 August 2014 by KunZhang (talk | contribs) (Created page with "==Round 5: screening of all Blood LMS clusters== *What we learned from the previous rounds: **Primary tumor tissues are quite different from cancer cell lines, including the l...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Round 5: screening of all Blood LMS clusters

  • What we learned from the previous rounds:
    • Primary tumor tissues are quite different from cancer cell lines, including the level of heterogeneity (variable fraction of cancer cells), and the biological variability. Therefore, the previous lists of targets compiled based on the published RRBS data in cancer cell lines could be incomplete. The recent batches of RRBS data that we generated on primary tumors might provide additional targets. However, there could be additional targets outside of RRBS targets.
    • A different strategy to identify cancer markers is to start with a set of genomic regions that have no or low methylation in whole blood, and systematically screen these regions in all tumors. This could be more efficient, because (i) we want to do non-invasive detection in ctDNA in the first place, so other tumor DMRs that have some methylation in blood cannot be use anyway; (ii) we have identified a list of Blood LMS (Low Methylation Sites) clusters (16,208 regions, ~4.3Mb in total size, or 0.14% of the genome). These clusters represents a subset of Blood UMRs that Dinh called. Performing target sequencing of these regions is much more efficient than screening the entire genome or even RRBS.
    • Cell-free DNA in serum contains a large amount of large DNA fragments from lysed white blood cells. Cell-free DNA in plasma are more enriched for ctDNA, but the yield is quite low. There are ~10ng of ctDNA from 1mL of whole blood. Therefore, the capturing assay needs to be very efficient. Otherwise, pre-amplification would be necessary to increase the number of starting molecules.
    • Based on the UMI, padlock capture on genomic DNA tends to be ~10x more efficient (or sensitive) than BSPP. Noi has shown that she can improve the sensitivity of RRBS by ~10 folds using Klenow/N9 pre-amplification. However, it is unclear whether the pre-amplification is effective on small ctDNA fragments that are <200bp in length.
    • It is a good time to evaluate different methods for screening and targeted sequencing of ctDNA. Here are some comparisons:
      • BSPP: PROS(highly optimized, easy to implement, single-molecule counting); CONS (efficiency is generally 1% or lower)
      • RRBS: PROS(simple workflow, works for even 1ng of input DNA); CONS (a fixed set of targets; no enrichment for small ctDNA fragments; high sequencing cost)
      • Selector(or Agilent Haloplex): PROS(capture prior to bisulfite conversion, could be more specific and uniform, higher efficiency than BSPP??); CONS (difficult to implement UMI; target selection is less flexible than BSPP unless many restriction enzymes are used, which increase the requirement for input DNA and protocol complexity)
      • NimbleGen SeqCap EPI: PROS (commercially optimized, capturing post-conversion DNA, which means it is compatible to pre-amplification); CONS (very expensive; slightly more complicated workflow compared with BSPP)