Limits...
Multiple criteria optimization joint analyses of microarray experiments in lung cancer: from existing microarray data to new knowledge.

Camacho-Cáceres KI, Acevedo-Díaz JC, Pérez-Marty LM, Ortiz M, Irizarry J, Cabrera-Ríos M, Isaza CE - Cancer Med (2015)

Bottom Line: These data, however, are stored and often times abandoned when new experimental technologies arrive.This work reexamines lung cancer microarray data with a novel multiple criteria optimization-based strategy aiming to detect highly differentially expressed genes.In the analysis, groups of samples from patients with distinct smoking habits (never smoker, current smoker) and different gender are contrasted to elicit sets of highly differentially expressed genes, several of which are already associated to lung cancer and other types of cancer.

View Article: PubMed Central - PubMed

Affiliation: Bio IE Lab, The Applied Optimization Group, Industrial Engineering Department, University of Puerto Rico, Mayaguez, Puerto Rico.

No MeSH data available.


Related in: MedlinePlus

Globally‐optimal Pareto‐efficient frontier consisting of RAGE and SPP1 genes.
© Copyright Policy - creativeCommonsBy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4940807&req=5

cam4540-fig-0007: Globally‐optimal Pareto‐efficient frontier consisting of RAGE and SPP1 genes.

Mentions: For the first analysis the group of never smokers was considered and the comparison was between controls and cancer samples. There were fifteen controls (HNS) and sixteen cancer (CNS) samples. The absolute value of the differences of means and medians for each gene were calculated. The analysis in MatLab tool was run in a computer with 4 GB of memory RAM and 2.66 GHz CPU. Due to this memory constraint, the Pareto‐efficient frontier was found in a tournament fashion 32 as explained next. The 22,283 genes were divided into three groups: two groups of 7500 and one of 7283 genes. The MatLab tool was used to find the locally efficient frontier in each group. Finally, the genes in each one of the three efficient frontiers were analyzed together to find the global Pareto‐efficient frontier. It is important to point out that the order of the partition and input of the data does not affect the final efficient frontier, as this is a case of explicit full comparison. In one criterion, the process would be similar to finding the tallest person in a room by picking the tallest one in different subgroups and comparing the local winners in the end to find the global winner. With enough computing memory, partitioning the data is not necessary. For each group, the locally nondominated subset was identified (Fig. 6). Then the locally nondominated subsets were used to obtain the globally optimal Pareto‐efficient frontier, as seen in Figure 7. For this first analysis RAGE and SPP1 are the genes in the global Pareto‐efficient frontier. It is important to recall that the user does not need to normalize or use a threshold value to achieve this result.


Multiple criteria optimization joint analyses of microarray experiments in lung cancer: from existing microarray data to new knowledge.

Camacho-Cáceres KI, Acevedo-Díaz JC, Pérez-Marty LM, Ortiz M, Irizarry J, Cabrera-Ríos M, Isaza CE - Cancer Med (2015)

Globally‐optimal Pareto‐efficient frontier consisting of RAGE and SPP1 genes.
© Copyright Policy - creativeCommonsBy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4940807&req=5

cam4540-fig-0007: Globally‐optimal Pareto‐efficient frontier consisting of RAGE and SPP1 genes.
Mentions: For the first analysis the group of never smokers was considered and the comparison was between controls and cancer samples. There were fifteen controls (HNS) and sixteen cancer (CNS) samples. The absolute value of the differences of means and medians for each gene were calculated. The analysis in MatLab tool was run in a computer with 4 GB of memory RAM and 2.66 GHz CPU. Due to this memory constraint, the Pareto‐efficient frontier was found in a tournament fashion 32 as explained next. The 22,283 genes were divided into three groups: two groups of 7500 and one of 7283 genes. The MatLab tool was used to find the locally efficient frontier in each group. Finally, the genes in each one of the three efficient frontiers were analyzed together to find the global Pareto‐efficient frontier. It is important to point out that the order of the partition and input of the data does not affect the final efficient frontier, as this is a case of explicit full comparison. In one criterion, the process would be similar to finding the tallest person in a room by picking the tallest one in different subgroups and comparing the local winners in the end to find the global winner. With enough computing memory, partitioning the data is not necessary. For each group, the locally nondominated subset was identified (Fig. 6). Then the locally nondominated subsets were used to obtain the globally optimal Pareto‐efficient frontier, as seen in Figure 7. For this first analysis RAGE and SPP1 are the genes in the global Pareto‐efficient frontier. It is important to recall that the user does not need to normalize or use a threshold value to achieve this result.

Bottom Line: These data, however, are stored and often times abandoned when new experimental technologies arrive.This work reexamines lung cancer microarray data with a novel multiple criteria optimization-based strategy aiming to detect highly differentially expressed genes.In the analysis, groups of samples from patients with distinct smoking habits (never smoker, current smoker) and different gender are contrasted to elicit sets of highly differentially expressed genes, several of which are already associated to lung cancer and other types of cancer.

View Article: PubMed Central - PubMed

Affiliation: Bio IE Lab, The Applied Optimization Group, Industrial Engineering Department, University of Puerto Rico, Mayaguez, Puerto Rico.

No MeSH data available.


Related in: MedlinePlus