Changes in 89dafb6: add ref

				Supplemental_Methods_and_Results.md
			
          @@ -60,12 +60,12 @@ We used another approach to assess the reproducibility of the Zhang study. In th

           ## Potential contribution of rare germline variants

          -As detailed in Supplemental Table S5, many of the studies contributing to our gene lists did not have matched germline DNA from every sample. It has been shown that some genes have disproportionately higher numbers of rare germline variants. The genes with the highest number of rare variants overall have been described in a recent study as FrequentLy mutAted GeneS (FLAGS)(@shyrFLAGSFrequentlyMutated2014). These genes should be considered with caution in any study that associates them with a phenotype. Insufficient removal of germline variants in the studies explored herein could explain some of the genes in Tiers 2 and 3. To evaluate this, we used the ranking of the frequency of rare mutations in all genes covered by common exome panels. This "FLAGS rank" assignes the lowest rank to genes with the highest number of rare variants (e.g. TTN = 1, MUC16 = 2, OBSCN = 3, etc). We used the distribution of ranks across the three tiers of DLBCL genes to evaluate the relative extent to which the genes on each list may be explained, in part, by inadequate removal of rare germline variants. If there is no influence on genes, the FLAGS_rank is expected to follow a uniform distribution. Accordingly, we compared each of the four gene lists to a uniform distribution using a negative binomial regression. Interestingly, the distributions of FLAGS_rank for Tier 2 and Tier 3 and the CGC genes were significantly different than the uniform distribution. In contrast, the distribution for Tier 1 genes was not significantly different from a uniform distribution. This is consistent with the Tier 1 DLBCL genes having minimal contamination of genes due to rare germline variants. To evaluate whether the separation of genes into Tiers 2 and 3 offers some benefit in this regard, we compared the Tier 2 DLBCL genes to the other lists. Indeed, Tier 2 showed a significant difference when compared to Tier 3 (P=3.86e-06). Based on this, we conclude that the Tier 3 gene list is dominated by FLAGS whereas Tier 2 represents a mixture of true DLBCL genes and FLAGS. 

          +As detailed in Supplemental Table S5, many of the studies contributing to our gene lists did not have matched germline DNA from every sample. It has been shown that some genes have disproportionately higher numbers of rare germline variants. The genes with the highest number of rare variants overall have been described in a recent study as FrequentLy mutAted GeneS (FLAGS)@shyrFLAGSFrequentlyMutated2014. These genes should be considered with caution in any study that associates them with a phenotype. Insufficient removal of germline variants in the studies explored herein could explain some of the genes in Tiers 2 and 3. To evaluate this, we used the ranking of the frequency of rare mutations in all genes covered by common exome panels. This "FLAGS rank" assignes the lowest rank to genes with the highest number of rare variants (e.g. TTN = 1, MUC16 = 2, OBSCN = 3, etc). We used the distribution of ranks across the three tiers of DLBCL genes to evaluate the relative extent to which the genes on each list may be explained, in part, by inadequate removal of rare germline variants. If there is no influence on genes, the FLAGS_rank is expected to follow a uniform distribution. Accordingly, we compared each of the four gene lists to a uniform distribution using a negative binomial regression. Interestingly, the distributions of FLAGS_rank for Tier 2 and Tier 3 and the Cosmic Cancer Gene Census (CGC)@futrealCensusHumanCancer2004 genes were significantly different than the uniform distribution. In contrast, the distribution for Tier 1 genes was not significantly different from a uniform distribution. This is consistent with the Tier 1 DLBCL genes having minimal contamination of genes due to rare germline variants. To evaluate whether the separation of genes into Tiers 2 and 3 offers some benefit in this regard, we compared the Tier 2 DLBCL genes to the other lists. Indeed, Tier 2 showed a significant difference when compared to Tier 3 (P=3.86e-06). Based on this, we conclude that the Tier 3 gene list is dominated by FLAGS whereas Tier 2 represents a mixture of true DLBCL genes and FLAGS. 

           <figure>

           <img src="FLAGS_rank_boxplot.png" alt="flags1" style="width:500px;"/>

           <figcaption><strong>Supplemental Figure S1. Comparing FLAGS rank in the DLBCL genes.</strong> <br> 

          -The FLAGS rank of DLBCL genes in Tier 1, 2 and 3 is compared. The genes in the COSMIC Cancer Gene Census list are included for comparison. Values closer to 1 indicate the gene has a higher number of rare germline variants.  

          +The FLAGS rank of DLBCL genes in Tier 1, 2 and 3 is compared. The genes in the COSMIC CGC list are included for comparison. Values closer to 1 indicate the gene has a higher number of rare germline variants.  

           </figcaption>

           </figure>

				morinlab.bib
			
          @@ -1,3 +1,22 @@

          +@article{futrealCensusHumanCancer2004,

          +	title = {A census of human cancer genes},

          +	volume = {4},

          +	issn = {1474-175X},

          +	doi = {10.1038/nrc1299},

          +	abstract = {A central aim of cancer research has been to identify the mutated genes that are causally implicated in oncogenesis (‘cancer genes’). After two decades of searching, how many have been identified and how do they compare to the complete gene set that has been revealed by the human genome sequence? We have conducted a ‘census’ of cancer genes that indicates that mutations in more than 1\% of genes contribute to human cancer. The census illustrates striking features in the types of sequence alteration, cancer classes in which oncogenic mutations have been identified and protein domains that are encoded by cancer genes.},

          +	language = {eng},

          +	number = {3},

          +	journal = {Nature Reviews. Cancer},

          +	author = {Futreal, P. Andrew and Coin, Lachlan and Marshall, Mhairi and Down, Thomas and Hubbard, Timothy and Wooster, Richard and Rahman, Nazneen and Stratton, Michael R.},

          +	month = mar,

          +	year = {2004},

          +	pmid = {14993899},

          +	pmcid = {PMC2665285},

          +	keywords = {Genes, Genome, Human, Humans, Mutation, Neoplasms, Oncogenes},

          +	pages = {177--183},

          +	file = {Accepted Version:/Users/rmorin/Zotero/storage/CCYI8YXU/Futreal et al. - 2004 - A census of human cancer genes.pdf:application/pdf},

          +}

          +

           @article{shyrFLAGSFrequentlyMutated2014,

           	title = {{FLAGS}, frequently mutated genes in public exomes},

           	volume = {7},

...	...	@@ -60,12 +60,12 @@ We used another approach to assess the reproducibility of the Zhang study. In th
60	60
61	61	## Potential contribution of rare germline variants
62	62
63		-As detailed in Supplemental Table S5, many of the studies contributing to our gene lists did not have matched germline DNA from every sample. It has been shown that some genes have disproportionately higher numbers of rare germline variants. The genes with the highest number of rare variants overall have been described in a recent study as FrequentLy mutAted GeneS (FLAGS)(@shyrFLAGSFrequentlyMutated2014). These genes should be considered with caution in any study that associates them with a phenotype. Insufficient removal of germline variants in the studies explored herein could explain some of the genes in Tiers 2 and 3. To evaluate this, we used the ranking of the frequency of rare mutations in all genes covered by common exome panels. This "FLAGS rank" assignes the lowest rank to genes with the highest number of rare variants (e.g. TTN = 1, MUC16 = 2, OBSCN = 3, etc). We used the distribution of ranks across the three tiers of DLBCL genes to evaluate the relative extent to which the genes on each list may be explained, in part, by inadequate removal of rare germline variants. If there is no influence on genes, the FLAGS_rank is expected to follow a uniform distribution. Accordingly, we compared each of the four gene lists to a uniform distribution using a negative binomial regression. Interestingly, the distributions of FLAGS_rank for Tier 2 and Tier 3 and the CGC genes were significantly different than the uniform distribution. In contrast, the distribution for Tier 1 genes was not significantly different from a uniform distribution. This is consistent with the Tier 1 DLBCL genes having minimal contamination of genes due to rare germline variants. To evaluate whether the separation of genes into Tiers 2 and 3 offers some benefit in this regard, we compared the Tier 2 DLBCL genes to the other lists. Indeed, Tier 2 showed a significant difference when compared to Tier 3 (P=3.86e-06). Based on this, we conclude that the Tier 3 gene list is dominated by FLAGS whereas Tier 2 represents a mixture of true DLBCL genes and FLAGS.
	63	+As detailed in Supplemental Table S5, many of the studies contributing to our gene lists did not have matched germline DNA from every sample. It has been shown that some genes have disproportionately higher numbers of rare germline variants. The genes with the highest number of rare variants overall have been described in a recent study as FrequentLy mutAted GeneS (FLAGS)@shyrFLAGSFrequentlyMutated2014. These genes should be considered with caution in any study that associates them with a phenotype. Insufficient removal of germline variants in the studies explored herein could explain some of the genes in Tiers 2 and 3. To evaluate this, we used the ranking of the frequency of rare mutations in all genes covered by common exome panels. This "FLAGS rank" assignes the lowest rank to genes with the highest number of rare variants (e.g. TTN = 1, MUC16 = 2, OBSCN = 3, etc). We used the distribution of ranks across the three tiers of DLBCL genes to evaluate the relative extent to which the genes on each list may be explained, in part, by inadequate removal of rare germline variants. If there is no influence on genes, the FLAGS_rank is expected to follow a uniform distribution. Accordingly, we compared each of the four gene lists to a uniform distribution using a negative binomial regression. Interestingly, the distributions of FLAGS_rank for Tier 2 and Tier 3 and the Cosmic Cancer Gene Census (CGC)@futrealCensusHumanCancer2004 genes were significantly different than the uniform distribution. In contrast, the distribution for Tier 1 genes was not significantly different from a uniform distribution. This is consistent with the Tier 1 DLBCL genes having minimal contamination of genes due to rare germline variants. To evaluate whether the separation of genes into Tiers 2 and 3 offers some benefit in this regard, we compared the Tier 2 DLBCL genes to the other lists. Indeed, Tier 2 showed a significant difference when compared to Tier 3 (P=3.86e-06). Based on this, we conclude that the Tier 3 gene list is dominated by FLAGS whereas Tier 2 represents a mixture of true DLBCL genes and FLAGS.
64	64
65	65	<figure>
66	66	<img src="FLAGS_rank_boxplot.png" alt="flags1" style="width:500px;"/>
67	67	<figcaption><strong>Supplemental Figure S1. Comparing FLAGS rank in the DLBCL genes.</strong> <br>
68		-The FLAGS rank of DLBCL genes in Tier 1, 2 and 3 is compared. The genes in the COSMIC Cancer Gene Census list are included for comparison. Values closer to 1 indicate the gene has a higher number of rare germline variants.
	68	+The FLAGS rank of DLBCL genes in Tier 1, 2 and 3 is compared. The genes in the COSMIC CGC list are included for comparison. Values closer to 1 indicate the gene has a higher number of rare germline variants.
69	69	</figcaption>
70	70	</figure>
71	71

...	...	@@ -1,3 +1,22 @@
	1	+@article{futrealCensusHumanCancer2004,
	2	+ title = {A census of human cancer genes},
	3	+ volume = {4},
	4	+ issn = {1474-175X},
	5	+ doi = {10.1038/nrc1299},
	6	+ abstract = {A central aim of cancer research has been to identify the mutated genes that are causally implicated in oncogenesis (‘cancer genes’). After two decades of searching, how many have been identified and how do they compare to the complete gene set that has been revealed by the human genome sequence? We have conducted a ‘census’ of cancer genes that indicates that mutations in more than 1\% of genes contribute to human cancer. The census illustrates striking features in the types of sequence alteration, cancer classes in which oncogenic mutations have been identified and protein domains that are encoded by cancer genes.},
	7	+ language = {eng},
	8	+ number = {3},
	9	+ journal = {Nature Reviews. Cancer},
	10	+ author = {Futreal, P. Andrew and Coin, Lachlan and Marshall, Mhairi and Down, Thomas and Hubbard, Timothy and Wooster, Richard and Rahman, Nazneen and Stratton, Michael R.},
	11	+ month = mar,
	12	+ year = {2004},
	13	+ pmid = {14993899},
	14	+ pmcid = {PMC2665285},
	15	+ keywords = {Genes, Genome, Human, Humans, Mutation, Neoplasms, Oncogenes},
	16	+ pages = {177--183},
	17	+ file = {Accepted Version:/Users/rmorin/Zotero/storage/CCYI8YXU/Futreal et al. - 2004 - A census of human cancer genes.pdf:application/pdf},
	18	+}
	19	+
1	20	@article{shyrFLAGSFrequentlyMutated2014,
2	21	title = {{FLAGS}, frequently mutated genes in public exomes},
3	22	volume = {7},