Perhaps a first approach for single base substitutions may be to subtract from the cancer patient variants those already reported in the 1000 Genomes Project, which are germline.

Very good idea! Thank you very much!

Due to the high variations between individuals per se I fear you will have great problems to obtain suitable results even when comparing your data set with the data of the 1000 Genomes Project. Even if you find mutations (and according to the literature you will find hundreds if not more individual changes) you can´t be sure whether they will be the reason for the transformation. I wouldn´t invest the money in sequencing this tumor if you don´t have additional DNA, save the money and time for a better suited patient.

Little point to do the sequencing then, unless you are looking for large structural variations using whole genome sequencing with deeper depth (40X plus), which costs a lot currently.

Large seq databases such as 1000G can only help to remove about 95% of your single base substitutions, plus there is no reliable function prediction tools for  the rest hundreds or thousands (if you sequence the whole genome) substitutions, plus somatic substitutions tend to give low quality scores, the real thing will be unseen.

Hi Julia,

although the suggestion of Albino is ok (you can also try the newest dbSNP, which incorporates the 1000g data), the problem raised by Michael is critical. Even filtering with dbSNP you will likely end up with a long list of variants. Then, discriminating between rare SNPs, somatic drivers and somatic passengers will be extremely challenging/impossible, unless you're only interested in the identification of already known oncogenic lesions. Are cancer cells available or just DNA? is it a myeloid or lymphoid leukemia?

Dear all, first of all, thank you to all of you for replying! I feared already that it would be basically impossible.

However, it's JMML we would like to sequence, so it affects very young children. From previous studies we know that the gene lists are usually quite small, as there are virtually no age-dependent passenger mutations. Furthermore, we know in quite a bit of the cases (ca 80%) of a somatic mutation in the RAS pathway (PTPN11, NRAS, KRAS, CBL), as the kids have been sequenced by Sanger for these genes previously (and again, from previous studies it is known that these are truly somatic aberrations). 

Rocco, we only have genomic DNA, unfortunately...

Hi Julia,

I'm familiar with MDS/MDS as I work, at least in part, on the same field (aCML/CMML mostly). You can still do NGS if you want to characterize your JMML for the quite extensive list of somatic variants that have been already identified in these disorders. Although it doesn't cover JMML, you can have a look at the excellent review written by Nick Cross and K. Zoi (Int J Hematol (2015) 101:229–242). However, if your final goal is to identify new driver mutations, unless you can contact kids/parents and obtain at least some buccal swab gDNA, I see this as a no go. :(

Dear Rocco,

Thank you again for your help, and also the literature suggestion! I will have a look at it. We might now get some saliva samples, so in the end it should be feasible to sequence at least part of the patients with a germline control :)

Have a nice week-end, best, Julia

For the remainder of samples that do not have controls, it would be useful to use control BAMs from unmatched samples. This will filter artifacts from sequencing chemistry, reagents, alignment, etc... Pediatric samples should be fairly silent, so I hope you get to your answer without having to sift through a million SNPs.

Dear Venkata, that's our experience as well! Thank you for the good suggestion, will give it a try! Best, Julia

Hi Julia,

that's great news: fingers crossed for your study! Consider also that if the patient is highly hyperleukocytotic you may end-up with a small but significant contamination of leukemic cells DNA in the buccal gDNA. This may be important for the subsequent bioinformatic analysis.

Best,

Rocco

I would NOT filter out known SNPs because many rare SNPs are turning out to be pathogenic. Moreover, the SNP databases contain many mutations that are also reported in COSMIC. Further, multiple variants associated with hematological cancer are germline (e.g., PIK3CD). I would perform the NGS using the cancer cell DNA, then prioritize candidate variants using ANNOVAR to query various SNP databases, COSMIC, FATHHM (which predictes driver and passenger mutations), and ClinVar (http://annovar.openbioinformatics.org/). I like Venkata's idea to use control BAMs from unmatched samples to rule out false positives, as we also do that in our lab. Lastly, you can sequence the top candidate variants using gDNA from the parents to test for germ line mutations, though a negative result would not be informative.

Dear Julie and Rocco,

Thank you for your answers and all the input. Julie, that sounds good! We are partly using your suggestions in our DNASeq analysis pipeline, but never used e.g. FATHHM to my knowledge - I will talk to the bioinformatics here about it.

Thank you all again for the input and the many helpful suggestions!

Best, Julia

Hello again,

back to Julie observation: among the data present in the dbSNP database you can also extract the MAF and the Clinical vs Non Clinical field. What we typically do when trying to identify somatic variants is to give low prioriity to the SNPs that have a MAF > 0.01 and are not clinical.

All the best,

Rocco