Package ubic.gemma.core.analysis.preprocess

Class TwoChannelMissingValuesImpl

java.lang.Object

ubic.gemma.core.analysis.preprocess.TwoChannelMissingValuesImpl

All Implemented Interfaces:

TwoChannelMissingValues

@Component
public class TwoChannelMissingValuesImpl
extends Object
implements TwoChannelMissingValues

Computes a missing value matrix for ratiometric data sets.

Supported formats and special cases:

• Genepix: CH1B_MEDIAN etc; (various versions)
• Incyte GEMTools: RAW_DATA etc (no background values)
• Quantarray: CH1_BKD etc
• F635.Median / F532.Median (genepix as rendered in some data sets)
• CH1_SMTM (found in GPL230)
• Caltech (GPL260)
• Agilent (Ch2BkgMedian etc or CH2_SIG_MEAN etc)
• GSE3251 (ch1.Background etc)
• GPL560 (*_CY3 vs *CY5)
• GSE1501 (NormCH2)

The missing values are computed with the following considerations with respect to available data

1. If the preferred quantitation type data is a missing value, then the data are considered missing (for consistency).
2. We then do additional checks if there is 'signal' data available.
3. If there are background values, they are used to compute signal-to-noise ratios
4. If the signal values already contain missing data, these are still considered missing.
5. If there are no background values, we try to compute a threshold based on a quantile of the signal
6. Otherwise, values will be considered 'present' unless the signal values are zero or missing.

Author:

pavlidis

Field Summary

Fields inherited from interface ubic.gemma.core.analysis.preprocess.TwoChannelMissingValues

DEFAULT_SIGNAL_TO_NOISE_THRESHOLD

Constructor Summary

Constructors

Constructor	Description
TwoChannelMissingValuesImpl()

Method Summary

Modifier and Type	Method	Description
Collection<RawExpressionDataVector>	computeMissingValues(ExpressionExperiment ee)
Collection<RawExpressionDataVector>	computeMissingValues(ExpressionExperiment ee, double signalToNoiseThreshold, Collection<Double> extraMissingValueIndicators)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

TwoChannelMissingValuesImpl

public TwoChannelMissingValuesImpl()

Method Detail

computeMissingValues

@Transactional
public Collection<RawExpressionDataVector> computeMissingValues(ExpressionExperiment ee)

Specified by:

computeMissingValues in interface TwoChannelMissingValues

Parameters:

ee - The expression experiment to analyze. The quantitation types to use are selected automatically. If you want more control use other computeMissingValues methods.

Returns:

collection of raw expression data vectors

computeMissingValues

@Transactional
public Collection<RawExpressionDataVector> computeMissingValues(ExpressionExperiment ee,
                                                                double signalToNoiseThreshold,
                                                                Collection<Double> extraMissingValueIndicators)

Specified by:

computeMissingValues in interface TwoChannelMissingValues

Parameters:

ee - The expression experiment to analyze. The quantitation types to use are selected automatically.

signalToNoiseThreshold - A value such as 1.5 or 2.0; only spots for which at least ONE of the channel signal is more than signalToNoiseThreshold*background (and the preferred data are not missing) will be considered present.

extraMissingValueIndicators - Values that should be considered missing. For example, some data sets use '0'. This can be null or empty and it will be ignored.

Returns:

DesignElementDataVectors corresponding to a new PRESENTCALL quantitation type for the experiment.