Hello

The human genome has been annotated automatically as well as manually: https://www.ensembl.org/info/genome/genebuild/index.html

There are several different 'biotypes' apart from protein coding genes: https://www.ensembl.org/info/genome/genebuild/biotypes.html

You can explore the supporting evidence that was used for annotation of a specific transcript in the browser, e.g.: http://www.ensembl.org/Homo_sapiens/Transcript/SupportingEvidence?db=core;g=ENSG00000140538;r=15:87876789-88256153;t=ENST00000394480

The MANE transcripts are flagged in the transcript table. We do suggest to work with the MANE transcript if you have to choose one. Please note that they are not avaible for all genes yet. Here is an FAQ that should help you to decide which transcript to use: http://www.ensembl.org/Help/Faq?id=276

Best wishes

Astrid

Ensembl helpdesk

Not all Ensembl transcripts are predictions. The human genome annotation process is described here. For higher confidence in the transcripts, you may want to consider those part of the CCDS set. For support for a particular, transcript, you can look at the associated flags.
Note also that not all transcripts of a gene are always expressed at the same time and place so a gene may be associated with many transcripts but in an given experimental system only one or a few may be relevant.

I suggest you also check out the MANE project. This set includes only one selected transcript per gene and that transcript matches end to end including UTRs with an Ensembl transcript.

As far as evidence goes, you can find that on the flat files for RefSeqs. For example, you can see the following for NM_002826

        ##Evidence-Data-START##
        Transcript exon combination :: U97276.2, SRR3476690.853351.1
                                       [ECO:0000332]
        RNAseq introns              :: single sample supports all introns
                                       SAMEA1965299, SAMEA1966682
                                       [ECO:0000348]
        ##Evidence-Data-END##