CAKE 3.7

New in Version 3.6

• Presentational tweaks to the Summary with Graphs report (15200).
• Corrected issues with flow fraction reporting in summary export reports, particularly for implicit flow fractions (15419).
• Tweaks to flow fraction normalisation algorithm to provide better fit quality in cases where complex formation pathways are combined with sharp initial parent decay and relatively high desired formation into metabolites (15420).
• The build date for CAKE is now reported on the splash screen (15493).
• Regulatory documentation is now referenced more explicitly (15494).
• Fixes a crash on reporting certain niche bad fits (15526).
• Fixes the UI switching back to the first data group when the topology is modified but the Data tab has never been opened (15568).
• Fixes a theoretical possible crash during fitting where optimisation fails in a very niche way (15577).
• Removed a scenario in which it was possible to export a blank fit summary report (15662).
• Minor layout changes to the fit reports tab (15668).

New in Version 3.5

Usability Enhancements

• CAKE is now more consistent in terms of how it handles non-UK/US locales for rendering and inputting dates and numbers throughout the application. Changes have been made to provide better adherance to international norms, whilst flexing to give CAKE users necessary additional flexibility where deemed necessary. Most are subtle, with the most prominent change being that fit reports now output consistently in the user's regional locale rather than outputting numbers following US/UK norms and dates following regional norms; this option can be overriden in the Application Preferences if desired. For more details, please see the user guide (CU-166).
• Users can now explicitly view and export diagnostic logs from the Help menu, showing the underlying CAKE R input/output and other information. We encourage users to provide a copy of these logs when reporting issues or queries. They will also be of interest to CAKE power users (CU-184).
• CAKE now supports installs to the user profile. This will not require admin privileges, so long as the appropriate .NET Framework is already installed on the machine (will be true for the vast majority of users) and there is not a previous per-machine install that needs to be uninstalled (CU-196).
• CAKE fit reports now utilise formal RTF headings to make it easier to navigate through them, particularly for long reports of multiple fits (CU-200).
• It is now possible for users to define different topologies and initial parameter sets for different data sheets by opting in to put them into distinct data sheet groupings (CU-202).
• New controls for the graphs on the first tab of the CAKE application so that users can more easily and explicitly control which data to display, as well as newly overlay multiple fits and datasets on a single graph. Previously limited parameter control from this tab has now been removed to designate this tab entirely to this purpose, and it has been renamed accordingly (CU-203).
• CAKE can now create summary reports of multiple fits in various tabular formats (CU-204):
  • Summary with Graphs
  • Detailed Statistical Summary
  • Simple Statistical Summary
• A new tab provides a high-level summary of all current fit reports in the current CAKE Study File (CU-205).
• When users schedule multiple fits across multiple datasets and models, CAKE will now keep the fits for each dataset together in the order, and will proceed with models in the order SFO, FOMC, DFOP, HS, IORE, for better alignment with FOCUS normal practices (CU-206).
• Added a new ICAMA option for the Kinetics Guidelines settings in Application Preferences, which will be the default for new users in China; renamed the "Both" setting to "FOCUS + NAFTA" (CU-207).
• New functionality allowing users to export compartment graphs to the clipboard or save them to disk (CU-210).
• New functionality allowing users to copy an image of the topology to the clipboard (CU-211).
• CAKE can now correctly display Chinese dates in fit reports (CU-215).
• CAKE reports now contain an image of the topology that was used instead of a textual description (CU-221).
• When compartments are removed from the topology (by whatever means), data associated with those compartments is now deleted after warning the user. This removes the previous confusing but common scenario of having data in a CSF file for a data sheet that was not visible in the application because the compartment was not selected in the topology (CU-226).
• New Data tab button to duplicate an existing data sheet (CU-227).

Modelling Enhancements

• Major updates to underlying optimisation libraries and overhaul of CAKE optimisation code, resulting in significant performance improvements in most tested use cases and improved optimiser performance on some fits with complex topologies (CU-163, CU-172).
• The step-wise approach to building the optimisation for multi-compartment topologies means that different metabolites of a compartment are first optimised independently of one another. This can lead to proposed initial values for the first step where the compartments are combined such that the formation fractions from a parent compartment sum to more than 1. Where this happens, CAKE now normalises the initial formation fraction values for the combined step back to sum to 1 so far as this is permitted by given parameter constraints (fixed parameters and parameter bounds) (CU-178).
• The pre-fitting algorithm used to determine a "best guess" for the initial value for the breakpoint on Hockey Stick models has been refined to improve accuracy, resulting in better fit quality on a handful of "difficult" Hockey Stick fit scenarios (CU-197).
• The default IRLS parameters for new studies have been changed: "Maximum Reweightings" now defaults to 200, with "Error Variance Tolerance" now defaulting to 1E-10. This gives greater assurance of IRLS convergence in a rare minority of cases that did not converge under the previous defaults. Existing CAKE Study Files will retain their current settings (CU-201).
• CAKE will now highlight to users if the likely cause for missing statistics on fits for higher order models is that the fit has degenerated to SFO (e.g. DFOP k values are close to equal) (CU-217).

Bug Fixes

• Corrected fit penalties not always being levied correctly when formation fractions summed to more than one. This previously meant that CAKE could sometimes favour fits with formation higher than one more often than was intended. This should now be much rarer (CU-178).
• Options to transfer fitted parameters into the initials are now aware of the "Automatic" options (CU-185).
• Corrected a rare scenario in which graphs for fits with the extra solver could be truncated before the end of the data range (CU-187).
• Corrected an unexpected error condition when certain types of badly formatted data were pasted into CAKE (CU-192).
• Fit reports now render correctly within the bounds of a page in Microsoft WordPad again (CU-193).
• "Additional statistics" are now correctly reported if a model is run with all of the parameters fixed (CU-194).
• Resolved some minor glitching of tooltips in the Fit Configuration dialog (CU-195).
• There is no longer potential data loss in some scenarios where compartments are given new names containing brackets (CU-199).
• Pasting data into CAKE that contains a blank line no longer results in no response from the application. Instead, blank lines are simply ignored and the rest of the data is pasted (CU-208).

New in Version 3.4

New User Interface Functionality

• New option to swap two compartments in the topology (rather than just columns of data) (CU-17).
• Increased explanation in reports when values are missing/uncalculated (CU-87, CU-144).
• CAKE now renders better on high-density displays and other configurations with Windows DPI settings other than "100%"/"Small" (CU-101).
• New option to disable sinks and transfer fitted values to initials when formation fractions optimised to (sum to) 1 (CU-112).
• The behaviour of the Data grid has been tweaked to make it easier to paste in data under some circumstances (CU-121).
• It is now possible to copy graphs to the clipboard from reports without pre-highlighting them (CU-122).
• New options to customise the footer in saved CAKE RTF reports, both at the application and file level (CU-125).
• Formation fraction initial values are now automatically changed more intelligently as the topology is adjusted (CU-129).
• It is now possible to display multiple extra metabolite curves at the same time (CU-143).
• Sum of squared residuals (or fit cost) is now directly provided in CAKE reports (CU-153).
• The CAKE CSV importer now allows you to proceed (with a warning) with a missing header and provides more detailed information when there are problems with your CSV (CU-162).
• New option to transfer fitted values to initials from reports (CU-164).
• Small tweaks to plot colours (CU-167).
• Various other minor tweaks.

New Modelling Capabilities

• Extra metabolite curves are now additionally supported for the following previously unsupported scenarios (CU-79a):
  • DFOP metabolites. DT50=1000 curves are for the k2 parameter.
  • Metabolites with non-zero initial concentrations.
  • Metabolites with more than one parent compound (so long as no parent is an ancestor of another parent).
  • Metabolites where the reported DT50 was ">10,000".
  • The final metabolite of a parent without a sink.
• The DFOP DT50 and DT90 calculation now succeeds more often when faced with very large k values (CU-85). N.B. This change has led to the removal of the old "SANN Iterations" optimiser option, which is no longer required.
• IORE models now pre-fit using an FOMC model, which aids fit stability and optimality (CU-119).
• In interim steps of fits on large topologies with some sinks disabled, CAKE sometimes needs to introduce a temporary formation fraction. This temporary formation fraction is now being configured more intelligently (CU-123, CU-176).
• As a result of work on CU-141a, below, the extra solver will now occasionally produce slightly more optimal fits (CU-141b).
• Minor tweak to the way CAKE handles optimal values that are on or very close to specified parameter limits.

Bug Fixes

• The CAKE user interface no longer gets "stuck" if a fit is stopped by the user at a particularly unfortunate time (CU-54).
• Situations where formation on some extra metabolite curves for second- or higher-order metabolites could be suppressed have been corrected (CU-79b).
• MCMC tested model count no longer counts one past the configured numbers of tested models (CU-111).
• Warnings about fitted DT50 being larger than 1000 when enabling extra metabolite curves now pop up only once per session/report. The text has also been tweaked (CU-127).
• Fixed a crash scenario that could eventually occur after the application had been used over an extremely long session (CU-128).
• Fixed a crash that could occur when renaming a tab when the user had previously been working in a different file that contained more tabs (CU-131).
• Fixed an error on creating an Advanced NAFTA Spreadsheet from a run with fixed parameters (CU-132).
• Fixed a disagreement between main graphs and residual plots/tables that could very rarely occur in fits that used the extra solver (CU-141a).
• Fixed the DT90/3.32 value being miscalculated in a German Windows locale (CU-145).
• The "Observed vs Predicted" tables in reports are now presented in a more consistent order (CU-146).
• Corrected double accounting of the g factor in formation from DFOP metabolites into their child metabolites (CU-150).
• Fixed parameters missing from tables in reports where configured initial values matched the configured M0 initial value (CU-160).
• CAKE can now load files on a Norwegian locale (CU-161).
• Corrected some inconsistent selection of which formation fraction should be considered "implicit" in some large topologies with disabled sinks (CU-171).
• CAKE will no longer advise the user to disable a sink that is already disabled (CU-174).
• IORE fits with metabolites now consistently re-use the k value from interim steps (CU-188).

New in Version 3.3

• Fixed a launch issue affecting some locales (CU-140).

New in Version 3.2

• Entering a new values on the Parameters tab and then starting a new fit immediately without first clicking outside the text box could cause the old parameter value to be reported at the top of the fit report. This has been fixed.
• It was previously possible to paste values from the clipboard containing tab characters into the field for compartment names. This could lead to a crash and data corruption. This has now been fixed by filtering tab characters out from values pasted from the clipboard.
• Clicking on a sink for an inactive compartment will no longer change the selected compartment to that compartment.
• Some of the internal optimisation calculations could use incorrect values at interim steps if a metabolite initial concentration was unfixed. This could throw the final fit step off by providing unreasonable initial values to it. This has been fixed.
• Implicit formation fractions (expressing formation from compartments without a sink to their final metabolite) are now explicitly stated in fit reports.
• If a fit sequence completes with formation fractions optimised to 1 from a compartment with a sink, CAKE will now draw attention to this and advise the user to consider re-running with better initial values and the sink disabled.
• Fitted data point values that are smaller than the specified ODE solver tolerance are now rounded down to 0.
• Data tab names can no longer be duplicated by double-clicking on a tab to rename it.
• Hockey stick fits and plots are now more accurate where breakpoints and/or time ranges are small.
• For all graphs, the number of data points used to produce lines has been increased to improve accuracy.
• Scrolling the mouse wheel when the Extra Metabolite Curves or Graphs for each Compartment drop-downs on the Current Fit tab have focus will no longer change their values.
• Under the right circumstances, models with DFOP primary metabolites could be marked as in need of saving when the user had not changed anything. This has been fixed.
• The mathematical model used for DFOP metabolites in versions 3.0 and 3.1 did not match that prescribed by FOCUS guidelines. This has been corrected.
• NAFTA Process fit sequences no longer use SFO fitted parameters as the basis for initial values for the DFOP and IORE stages. The IORE N value is still estimated automatically based on results from the SFO and/or DFOP stages where possible.
• The process for recommending/highlighting a representative half-life on the NAFTA Process spreadsheet has been updated to match the latest guidelines.
• DFOP models with long decay times sometimes had lines missing on some metabolite graphs. This has been fixed.
• Using spaces in names for compartments will no longer cause odd behaviour and missing data.
• DFOP models will now always report k₁ as the faster rate by convention, transforming parameters after the optimisation process if required.
• Ctrl+X can now be used to "cut" on the Data tab.
• Changing the Graphs for each Compartment drop-down or the Generate Advanced Reports check box on the Current Fit tab will now reload any active report with the chosen option.
• Initial values for metabolite initial concentrations are now stated in fit reports.
• Fitted values for unfixed metabolite initial concentrations are now used as initial values in subsequent steps, in common with other fitted parameters.
• Study files that dated from version 2.2 or earlier and were opened in version 3.0 or newer could end up misbehaving with the IORE model until the file was saved and re-loaded (specifically, the IORE model was internally stuck on a parent-only topology and with default display names for compartments). This has been fixed.
• The representative half-life for FOMC models, t_FOMC, has been renamed in reports to DT90 / 3.32.
• Long names for data sets and topology compartments are now handled better throughout the application and name length limits are applied more consistently.
• It is no longer possible to use the Generate Advanced NAFTA Spreadsheet option while fits are running.
• Fixed a case where it was possible to crash the option when pasting data on the data tab in a very particular way.
• Minor user interface improvements.

New in Version 3.1

• The calculation of the number of degrees of freedom used when determining t-test statistics and confidence intervals was incorrect in version 3.0 when there were replicate time 0 observations for a compartment with a fixed initial concentration. This has been fixed.
• CAKE previously displayed blank cells in reports where values could not be calculated for some reason. This has now been changed to "nd" (not determined) and "N/A" not applicable as appropriate and there is additional guidance in the user documentation.
• The user can no longer change metabolite model selections while a fit is running.
• Decay times in reports now save and load consistently in a situation where some of them could not be determined during a fit.
• Under very particular circumstances, the parameters tab in version 3.0 could show no metabolite model selection after loading a file. This has been fixed.

New in Version 3.0

• The CAKE configurable topology is now more complex. Metabolites C1 and A3 have been added and more links are configurable, including horizontal links that can be configured in either direction at each level.
• Advanced users can now configure primary metabolites to be fitted to the DFOP model instead of the SFO model. Here, "primary metabolite" means metabolites that have the overall parent compound as their only parent. This choice can be made on the Parameters tab.
• The IORE model has been added to the available models list. The "N" parameter initial value is auto-estimated by default; this can be changed from the Fit Options dialog.
• A new "NAFTA Process" fit is available. This runs SFO, DFOP and IORE in sequence with estimations of initial values for parameters applied at each step. The fitting process is similar to the EPA's PestDFV tool.
• A post-processing spreadsheet intended for "NAFTA Process" fits has been added. Users can export their "NAFTA Process" fits to the spreadsheet to view summary information, graphs and statistics in a form similar to the output from PestDFV. Additionally, advanced users can choose to combine any three parent-only fit reports of any type that share a common data set and put the combined information onto the spreadsheet. The template for the spreadsheet is also user-editable: full details in the built-in application help.
• CAKE now provides a regional setting on the Application Preferences dialog. In "FOCUS" mode, the SFO, DFOP, FOMC and HS models are available; in "NAFTA" mode, the SFO, DFOP and IORE models are available as well as the "NAFTA Process" fit; in "Both" mode, all features are available.
• Splash/about screen updated to recognise Ian Hardy (Battelle) and Simon Ford (JSCi) as scientific advisors.
• The back-calculated DT50 for FOMC models recommended by the FOCUS guidelines has been added to CAKE reports.
• Improved the behaviour when the "rerun selected fits that exist already" option is disabled so that partially completed reports (from cancelling a run) are discarded and run again.
• The "Run Fit" menu option is now disabled while a fit is running.
• The option to not save extra metabolite curves into files has now been removed. Extra metabolite curves will now be saved into CAKE Study Files whenever they have been produced.
• Improved the parameters grid so that parameters displayed as "automatic" provide a shortcut to the Fit Options dialog via double click (so that the user can easily change these values to be manually entered).
• Fixed a rare bug where chi squared values would be missing from reports if the fitted model produced an incalculably small concentration at a relevant time point.
• Fixed the CAKE window sometimes losing focus as soon as a file is opened.
• The calculation of the number of degrees of freedom used when determining t-test statistics and confidence intervals has been changed to exclude time zero observations if and only if the corresponding time zero parameter is fixed, per latest industry advice. This means that many pre-existing CAKE fits will show slightly different statistics in this release.
• Fixed an issue where copying and pasting whole data sheets within the CAKE application could fail.
• Fixed an extremely rare crash that could occur in the middle of a fit operation.
• Removed extra metabolite curves from CAKE reports where the metabolite either has a non-zero initial concentration or multiple parents: curves in these circumstances were previously being calculated incorrectly. (Extra metabolite curves are also not available for DFOP metabolites.)
• Improved hockey-stick auto-estimation of initial t_b value so that it will not choose values outside bounds specified by the user.
• Added additional validation to prevent users from entering invalid parameter configurations (e.g. an initial value that falls outside the bounds they have specified). In addition, accidental non-numeric input is now handled better.
• Improved the behaviour of the big graph on the Properties tab. It will now handle custom compartment names correctly and more consistently select a fit line to display when multiple suitable reports are present.
• Added the missing page on application preferences to the documentation (it was accidentally excluded from the CAKE 2.2 installer).
• 64 bit CAKE will now use the 64 bit R interpreter in preference to the 32 bit one.
• Fixed the MCMC optimiser putting additional redundant columns in the decay time tables in reports.
• Fixed loaded CAKE study files claiming in reports that t_b selection was automatic when it was not.
• Fixed incorrect graphs at intermediate steps when running the OLS optimiser in combination with a metabolite with non-fixed initial concentration.
• Various minor user interface improvements.

New in Version 2.2

• Degrees of freedom for chi squared computations will now exclude observations at time 0 if and only if the corresponding parameter is fixed.
• Fixed a crash that could occur when running fits with the "keep existing fits" option enabled but the "rerun selected fits that exist already" option disabled.
• By default, CAKE will now choose the initial value for the breakpoint for hockey stick models automatically. This solves problems of convergence to sub-optimal fits when the initial value of the breakpoint is not chosen well, an issue that is inherent to the hockey stick model for the numerical solvers that CAKE uses. Users can still elect to choose an initial value manually if they turn this option off in the "Fit Options".
• Fixed a crash that could occur under some circumstances when a long-running fit was stopped in progress and then another fit started in its place.
• Under certain circumstances, improved the convergence of the iteratively reweighted least squares process when the extra solver is disabled.
• A sink being disabled is now correctly read in from saved CAKE Study Files.
• Reduced the "as an image" size of graphs placed on the clipboard. Pasting into Office applications "as an image" will now render at a sensible size.
• Improved the efficiency of chi squared calculations, in line with a change in mkin version 0.9-22.
• Prevented a crash that could occur when attempting to view reports that contained incomplete fit information.
• Improved the wording of a couple of error messages.
• Added an option for the user to specify their preferred paper size for reports as either A4 or US Letter (relevant when opening in or copying to external programs such as Microsoft Word). This option is set on the new Application Preferences dialog, accessible from the File menu. When you first run this version of CAKE, it will automatically select US Letter for users with a USA or Canada Windows regional setting and A4 for all other users.

New in Version 2.1

• The use of the "extra solver" is no longer controlled from the "Fit Options" dialog and CAKE will no longer prompt users about using the "extra solver" in the middle of fits. Instead, users now make a choice about using the extra solver when starting a run: "Never", "If Required" or "Always (where possible)".
• Confidence intervals for decay parameters have been reintroduced into CAKE in line with updated guidance from the European Food Safety Authority.
• Fixed a bug where 95% confidence interval upper limits were not being loaded correctly from saved CAKE Study Files.
• Fixed a bug where graphs would sometimes not be plotted if the decay parameter for a metabolite was fixed.
• Fixed a bug where statistics produced from the R library for the "extra solver" were sometimes wildly erroneous. We have worked around this issue in CAKE by no longer relying on the "extra solver" R library for these statistics.
• Fixed an issue where the iteratively reweighted least squares method would sometimes fail to iterate and reweight when running with the "extra solver" turned on.
• Fixed a crash that could occur if the user right clicked on the report area while it was empty.
• Fixed an error where chi squared information was reported with one too few degrees of freedom when running a model with every parameter fixed.
• Fixed incorrect DT50/DT90 values for the DFOP model that could occur when the correct values were very large (e.g. due to very slow decay in the asymptotically dominant term in the model). DT50/DT90 values cannot be determined analytically for the DFOP model, but the robustness of the numerical calculation has been improved. Additionally, all DT50/DT90 values larger than 10,000 days now simply report as ">10,000" in CAKE reports.
• Fixed a bug with pasting data into the data tab in CAKE where empty cells could get lost if pasting individual columns.
• Increased the default value of t_b for Hockey Stick models from 1 day to 10 days.
• Changed the CSV import process so that it will not store any data within CAKE files that it cannot positively associate with a substance in the application.
• Fits for metabolites with no associated observation data are now plotted in CAKE.
• Fixed bugs with using custom names for compartments where data could apparently go missing from the data tab (particularly when switching columns around) and old custom names could remain internally within the CAKE study file (but not be visible within the application).
• Improved the automatic naming of new data sheets so that it does not duplicate existing names.