Limits...
Joint analysis of differential gene expression in multiple studies using correlation motifs.

Wei Y, Tenzen T, Ji H - Biostatistics (2014)

Bottom Line: The motifs provide the basis for sharing information among studies and genes.The approach has flexibility to handle all possible study-specific differential patterns.It improves detection of differential expression and overcomes the barrier of exponential model complexity.

View Article: PubMed Central - PubMed

Affiliation: Department of Biostatistics, Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD, USADepartment of Statistics, The Chinese University of Hong Kong, Shatin NT, Hong Kong.

Show MeSH
Results for the model assumption-based simulations (simulations 1 and 4). Also see supplemental Figure A.1 available at Biostatistics online for simulations 2 and 3. (a) and (g) True motif patterns for simulations 1 and 4. The  of the true motifs is shown. Each row indicates a motif pattern and each column represents a study. The actual number of genes belonging to each motif (i.e. ) is displayed at the right end of each row. The gray scale of the cell  demonstrates the probability of differential expression in study  for pattern . Black means 1 and white means 0. (b) and (h) The estimated  from the learned motifs with  annotated at the end of each row. (c) and (i) BIC plots. It can be seen that motif patterns reported by CorMotif under the minimal BIC are similar to the true underlying motif patterns. (d)–(f) and (j)–(l) Gene ranking performance of different methods in simulations 1 and 4. , the number of genes that are truly differentially expressed in study  among the top  ranked genes by a given method, is plotted against the rank cutoff . For each simulation, results for a few representative studies are shown. Each plot is for one study.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4263229&req=5

KXU038F2: Results for the model assumption-based simulations (simulations 1 and 4). Also see supplemental Figure A.1 available at Biostatistics online for simulations 2 and 3. (a) and (g) True motif patterns for simulations 1 and 4. The of the true motifs is shown. Each row indicates a motif pattern and each column represents a study. The actual number of genes belonging to each motif (i.e. ) is displayed at the right end of each row. The gray scale of the cell demonstrates the probability of differential expression in study for pattern . Black means 1 and white means 0. (b) and (h) The estimated from the learned motifs with annotated at the end of each row. (c) and (i) BIC plots. It can be seen that motif patterns reported by CorMotif under the minimal BIC are similar to the true underlying motif patterns. (d)–(f) and (j)–(l) Gene ranking performance of different methods in simulations 1 and 4. , the number of genes that are truly differentially expressed in study among the top ranked genes by a given method, is plotted against the rank cutoff . For each simulation, results for a few representative studies are shown. Each plot is for one study.

Mentions: We first tested CorMotif using simulations. In simulation 1, we generated 10 000 genes and four studies according to the four differential patterns in Figure 2(a): 100 genes were differentially expressed in all four studies (); 400 genes were differential only in studies 1 and 2 (); 400 genes were differential only in studies 2 and 3 (); 9100 genes were non-differential (). Each study had six samples: three cases and three controls. The variances s were simulated from a scaled inverse distribution , where and . Given , the expression values were generated using ). Whenever , we drew from where , and was then added to the expression values of the three cases (i.e. s).


Joint analysis of differential gene expression in multiple studies using correlation motifs.

Wei Y, Tenzen T, Ji H - Biostatistics (2014)

Results for the model assumption-based simulations (simulations 1 and 4). Also see supplemental Figure A.1 available at Biostatistics online for simulations 2 and 3. (a) and (g) True motif patterns for simulations 1 and 4. The  of the true motifs is shown. Each row indicates a motif pattern and each column represents a study. The actual number of genes belonging to each motif (i.e. ) is displayed at the right end of each row. The gray scale of the cell  demonstrates the probability of differential expression in study  for pattern . Black means 1 and white means 0. (b) and (h) The estimated  from the learned motifs with  annotated at the end of each row. (c) and (i) BIC plots. It can be seen that motif patterns reported by CorMotif under the minimal BIC are similar to the true underlying motif patterns. (d)–(f) and (j)–(l) Gene ranking performance of different methods in simulations 1 and 4. , the number of genes that are truly differentially expressed in study  among the top  ranked genes by a given method, is plotted against the rank cutoff . For each simulation, results for a few representative studies are shown. Each plot is for one study.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4263229&req=5

KXU038F2: Results for the model assumption-based simulations (simulations 1 and 4). Also see supplemental Figure A.1 available at Biostatistics online for simulations 2 and 3. (a) and (g) True motif patterns for simulations 1 and 4. The of the true motifs is shown. Each row indicates a motif pattern and each column represents a study. The actual number of genes belonging to each motif (i.e. ) is displayed at the right end of each row. The gray scale of the cell demonstrates the probability of differential expression in study for pattern . Black means 1 and white means 0. (b) and (h) The estimated from the learned motifs with annotated at the end of each row. (c) and (i) BIC plots. It can be seen that motif patterns reported by CorMotif under the minimal BIC are similar to the true underlying motif patterns. (d)–(f) and (j)–(l) Gene ranking performance of different methods in simulations 1 and 4. , the number of genes that are truly differentially expressed in study among the top ranked genes by a given method, is plotted against the rank cutoff . For each simulation, results for a few representative studies are shown. Each plot is for one study.
Mentions: We first tested CorMotif using simulations. In simulation 1, we generated 10 000 genes and four studies according to the four differential patterns in Figure 2(a): 100 genes were differentially expressed in all four studies (); 400 genes were differential only in studies 1 and 2 (); 400 genes were differential only in studies 2 and 3 (); 9100 genes were non-differential (). Each study had six samples: three cases and three controls. The variances s were simulated from a scaled inverse distribution , where and . Given , the expression values were generated using ). Whenever , we drew from where , and was then added to the expression values of the three cases (i.e. s).

Bottom Line: The motifs provide the basis for sharing information among studies and genes.The approach has flexibility to handle all possible study-specific differential patterns.It improves detection of differential expression and overcomes the barrier of exponential model complexity.

View Article: PubMed Central - PubMed

Affiliation: Department of Biostatistics, Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD, USADepartment of Statistics, The Chinese University of Hong Kong, Shatin NT, Hong Kong.

Show MeSH