biotmle 1.37.1 (BioC 3.24):
- Removal of the `superheat` dependency, which has been archived on CRAN. The
  heatmap drawn by `heatmap_ic()` is now built directly with `ggplot2`.
- `heatmap_ic()` gains a `row_dendrogram` argument, which restores the row
  dendrogram previously available through `superheat`. Biomarkers are ordered
  by hierarchical clustering whether or not the dendrogram is drawn, so this
  argument affects only the display. It defaults to `FALSE`, matching the
  default of the `superheat` argument it replaces. Drawing the dendrogram
  introduces dependencies on `ggdendro` and `patchwork`.
- `heatmap_ic()` now returns a `ggplot` object, which can be stored and further
  modified in place, rather than drawing to the active graphics device as a
  side effect.
- `heatmap_ic()` gains a `scale` argument controlling whether each biomarker's
  contributions are standardized across observations before plotting; this
  replaces the argument of the same name previously forwarded to `superheat`.
- Arguments passed to `heatmap_ic()` via `...` are now forwarded to
  `ggplot2::geom_tile()` rather than to `superheat::superheat()`. Calls that
  relied on `superheat`-specific arguments (e.g., `left.label`,
  `clustering.method`, `row.dendrogram`) must be updated.

---

biotmle 1.18.0 (BioC 3.14) through 1.37.0 (BioC 3.24):
- No significant updates; these version numbers reflect the automatic version
  increments applied by Bioconductor at each release cycle.

---

biotmle 1.17.0:
- Removal of `future` and `doFuture` for simplification of parallelization. All
  control of parallel computation now done through `BiocParallel`.

---

biotmle 1.16.0 (BioC 3.13):
- No significant updates.

---

biotmle 1.15.0:
- No significant updates.

---

biotmle 1.14.0 (BioC 3.12):
- No significant updates.

---
biotmle 1.13.0:
- No significant updates.

---

biotmle 1.12.0:
- No significant updates.

---

biotmle 1.11.0 (BioC 3.11):
- Change of estimation backend from the `tmle` package to the `drtmle` package.
- Removal of option to have repeated subjects since unsupported in new backend.
- Adds argument `bppar_debug` to facilitate debugging around parallelization.

---

biotmle 1.10.0 (BioC 3.10):
- No significant updates.

---

biotmle 1.8.0 (BioC 3.9):
- No significant updates.

---

biotmle 1.6.0 (BioC 3.8):
- No significant updates.

---

biotmle 1.4.0 (BioC 3.7):
- An updated release of this package for Bioconductor 3.7, released April 2018.
- This release primarily implements minor changes, including the use of colors
  in the plots produced by the visualization methods.

---

biotmle 1.3.0 (BioC 3.6):
- An updated release of this package for Bioconductor 3.6, released in October
  2017.
- An option for applying this methodology to next-generation sequencing data has
  been added, based on the popular "voom" transform of the limma R package.
- Facilities for parallelized computation have been completely re-implemented:
  current routines favor a combination of future and BiocParallel.
- The method for estimating biomarkers based on an observed outcome has been
  removed (temporarily). Inference based on this method requires re-thinking.
- A full suite of unit tests have been added, covering most package functions.

---

biotmle 1.0.0 (BioC 3.5):
- The first release of this package was made as part of Bioconductor 3.5, in
  2016.

---

The biotmle R package provides routines for statistical methodology first
described in the technical manuscript [1] and the software paper [2]:

1. Nima S. Hejazi, Sara Kherad-Pajouh, Mark J. van der Laan, Alan E. Hubbard.
   Variance stabilization of targeted sstimators of causal parameters in
   high-dimensional settings. https://arxiv.org/abs/1710.05451

2. Nima S. Hejazi, Weixin Cai, Alan E. Hubbard. biotmle: Targeted Learning for
   Biomarker Discovery. The Journal of Open Source Software, 2(15), 2017.
   https://dx.doi.org/10.21105/joss.00295

